The Persistent Effectsof Peru’s Mining Mita∗

Melissa DellDepartment of Economics, Massachusetts Institute of Technology

May, 2009.

Abstract

This study utilizes regression discontinuity to examine the long run impacts of the mita,an extensive forced mining labor system in effect in Peru and Bolivia between 1573 and1812. Results indicate that a mita effect lowers household consumption by around onethird in subjected districts today. Using data from the Spanish Empire and PeruvianRepublic to trace channels of institutional persistence, I show that the mita’s influencehas persisted through its impacts on land tenure and public goods provision. Mita dis-tricts historically had fewer large landowners and lower educational attainment. Today,they are less integrated into road networks, and their residents are substantially morelikely to be subsistence farmers.

Keywords : forced labor, land tenure, public goods.

JEL Classification: H41, N26, O43

∗I am grateful to Daron Acemoglu, Bob Allen, Josh Angrist, Abhijit Banerjee, John Coatsworth,Knick Harley, Nils Jacobsen, Alan Manning, Ben Olken, James Robinson, Peter Temin, Gary Urton,Heidi Williams, Jeff Williamson, and seminar participants at Harvard, MIT, and Oxford for helpfulcomments and suggestions. I also thank Javier Escobal and Jennifer Jaw for assistance in accessingdata. Research funding was provided by the George Webb Medley Fund (Oxford University). Contactemail: mdell@mit.edu.

1 Introduction

The role of historical institutions in explaining contemporary underdevelopment has

generated significant debate in recent years.1 Studies find quantitative support for an

impact of history on current economic outcomes (Nunn, 2008; Glaeser and Shleifer, 2002;

Acemoglu et al., 2001, 2002; Hall and Jones, 1999) but have not focused on channels of

persistence. Existing empirical evidence offers little guidance in distinguishing a variety

of potential mechanisms, such as property rights enforcement, high inequality, ethnic

fractionalization, barriers to entry, and public goods. This paper uses variation in the

assignment of historical institutions in Peru to identify land tenure and public goods as

channels through which their effects persist.

Specifically, I examine the long run impacts of the mining mita, a forced labor system

instituted by the Spanish government in Peru and Bolivia in 1573 and abolished in 1812.

The mita required over 200 indigenous communities to send one seventh of their adult

male population to work in the Potosı silver and Huancavelica mercury mines (Figure 1).

The contribution of mita conscripts changed discretely at the boundary of the subjected

region - on one side all communities sent the same percentage of their population to

the mines, while on the other side all communities were exempt. This discrete change

suggests a regression discontinuity (RD) approach for evaluating the long term effects

of the mita.

The validity of the RD approach requires all relevant factors besides treatment to

vary smoothly at the mita boundary, and detailed quantitative and historical evidence

suggest that this restriction is plausible. I implement the RD approach by focusing

exclusively on the portion of the mita boundary that transects the Andean range in

southern Peru. While much of the boundary tightly follows the steep Andean precipice -

and hence has elevation and the ethnic distribution of the population changing discretely

at the boundary - elevation and the ethnic distribution are identical across the segment of

the boundary on which this study focuses. Moreover, specification checks using detailed

census data on local economic capacity, local institutions, and demography - collected

just prior to the mita’s institution in 1573 - do not find differences across this segment

before the mita’s enactment. All regressions reported in this paper include an RD

polynomial in geographic distance (to either the Potosı mines or mita boundary).

1See for example Coatsworth, 2005; Glaeser et al., 2004; Easterly and Levine, 2003; Acemoglu etal., 2001, 2002; Sachs, 2001; Engerman and Sokoloff, 1997.

1

Using the RD approach and household survey data, I show that a long-run mita ef-

fect lowers equivalent household consumption by around one third in subjected districts

today. Moreover, data from a height census of six to nine year old school children in

the region document that the mita’s persistent impact increases childhood stunting by

around five percentage points. These estimates remain stable as I limit the sample to

fall within increasingly narrow bands of the mita boundary and are robust to a num-

ber of alternative specifications. In addition, they are consistent with recent empirical

studies, such as Acemoglu et al. (2002). These authors document a reversal in the

world income ranking between 1500 and today, hypothesizing that colonial rulers es-

tablished persistent extractive institutions in historically prosperous regions to exploit

dense indigenous populations and natural resources. The present study offers direct

quantitative evidence for a long run impact of extractive institutions. More generally,

it provides microeconomic evidence consistent with a number of studies that establish a

relationship between historical institutions and contemporary economic outcomes using

aggregate data at the within-country (Banerjee and Iyer, 2005) or cross-country (Nunn,

2008; Glaeser and Shleifer, 2002) level.

After establishing this result, the remainder of the paper uses data from the Spanish

Empire and Peruvian Republic, combined with the RD approach, to investigate channels

of persistence. Three principal findings emerge. First, using district level data collected

in 1689, I document that haciendas - rural estates with an attached labor force - devel-

oped primarily outside the mita catchment. In order to minimize the competition the

state faced in accessing scarce mita labor, colonial policy restricted the formation of ha-

ciendas in mita districts, promoting communal land tenure there instead (Garrett, 2005;

Larson, 1988; Sanchez-Albornoz, 1978). The mita’s effect on hacienda concentration re-

mained negative and significant in 1940. Second, mita districts may have historically

achieved lower levels of education, and today they remain less integrated into road net-

works. Finally, data from the most recent agricultural census document that residents

of mita districts are substantially more likely to be subsistence farmers.

Based on quantitative and historical evidence, I hypothesize that the long-term pres-

ence of large landowners in non-mita districts provided a stable land tenure system that

encouraged public goods provision. In Peru, the alternative to large landowners was not

relative equality between small, enfranchised landowners. While the property rights of

large landowners remained consistently secure, the institutional structures in place did

not respect the rights of the peasants who held most of the land in mita districts. Soon

2

after the mita ended, the Peruvian government abolished the communal land tenure

that had predominated in mita districts during the colonial period, but did not replace

it with a system of enforceable peasant titling (Jacobsen, 1993; Dancuart and Rodriguez,

1902, vol. 2, p. 136). As a result, extensive confiscation of peasant lands and numerous

responding peasant rebellions were concentrated in mita districts during the late 19th

and early 20th centuries (Bustamante Otero, 1987, p. 126-130; Flores Galindo, 1987,

p. 240; Ramos Zambrano, 1985, p. 29-34). Widespread banditry and livestock rustling

remained endemic there well into the 20th century (Jacobsen, 1993; CVR, 2003). Be-

cause established landowners in non-mita districts enjoyed more secure title to their

property, it is probable that they received higher returns from investing in public goods

than did most individuals inside the mita catchment. Moreover, landowners controlled

a substantial percentage of the productive factors, both land and labor, so they could

expropriate much of the surplus that public goods provided. Finally, historical evidence

indicates that well-established landowners possessed the political connections required

to secure public goods (Stein, 1980).

Roads offer an important example of the role landowners played in securing public

goods. The hacienda elite lobbied successfully for roads, constructed to pass through as

many haciendas as possible (Stein, 1980, p. 59). Recent empirical evidence links roads

to increased market participation and higher household income (Escobal, 2001; Agreda

and Escobal, 1998). Roads constructed during the era of haciendas remain - although

haciendas were dissolved by agrarian reform in the 1970’s - and they continue to impact

market participation and household income.

The positive association between the historical presence of large landowners and

contemporary levels of economic development that this study documents contrasts with

the well-known hypothesis that historically high land inequality is at the heart of Latin

America’s poor long run growth performance (Engerman and Sokoloff, 1997). Engerman

and Sokoloff argue that high historical inequality lowered subsequent investments in

public goods, leading to worse outcomes in areas with high land concentration. Peru

provides an important example where the Engerman-Sokoloff logic does not hold: large

landowners could more easily defend their property rights and lobby for public goods.

While non-mita districts were not organized for the masses, more extensive public goods

do benefit the general populace today.2

2Also note that this paper’s findings are consistent with evidence from other Latin American coun-tries. Acemoglu et al. (2008) document that in Cundinamarca, Colombia, municipalities with a higher

3

The rest of this paper is organized as follows: In the next section, I provide an

overview of the mita, and Section 3 tests whether the mita affects contemporary living

standards. Section 4 examines channels empirically. Finally, Section 5 offers concluding

remarks.

2 The Mining Mita

2.1 Historical Introduction

The mining mita, instituted by the Spanish government in 1573 and abolished in 1812,

required over 200 indigenous communities in Peru and Bolivia to send one seventh of

their adult male population to work in the Potosı silver and Huancavelica mercury

mines (Figure 1). The Potosı mines, discovered in 1545, provided the largest deposits of

silver in the Spanish Empire, but production collapsed in the 1560’s due to severe labor

shortages and the exhaustion of high-grade ores (Cole, 1985, p. 4). The development in

1557 of a mercury amalgamation process for refining low-grade ores and the discovery of

mercury at Huancavelica in 1563 created hopes for revitalizing Potosı silver production.

Beginning in 1573, villages located within a contiguous region were required to provide

rotating mita laborers to Potosı or Huancavelica, and the region subjected remained

constant for most of the mita’s existence.3 The mita assigned 14,181 conscripts from

southern Peru and Bolivia to Potosı and 3,280 conscripts from central and southern Peru

to Huancavelica (Bakewell, 1984, p. 83).4 Using population estimates from the early

17th century (Cook, 1981), I calculate that around 3% of adult males living within the

current boundaries of Peru were conscripted to the mita at a given point in time. Since

all adult males in mita districts were supposed to serve one out of every seven years,

land gini in the 19th century are more developed today. Coatsworth (2005) cites evidence that in-equality in Mexico did not became high relative to the U.S. until the early 20th century, after almosta century of very poor economic performance relative to the U.S. had already occurred.

3In 1578, eighteen districts subjected to the mita in the previous 1575 list do not reappear. Theywere districts with small populations that do not appear in later censuses, which suggests that they wereincorporated into nearby districts (Bakewell, 1984, p. 83). Moreover, several districts in Condesuyos(now part of Arequipa) were briefly required to contribute in 1578, but were subsequently re-exempt.There were no other changes in the set of subjected districts.

4Individuals could attempt to escape mita service by fleeing their communities, and a number pursuedthis strategy (Wightman, 1993). Yet fleeing had costs - giving up access to land, community, and family;facing severe punishment if caught; and either paying additional taxes in the destination location as a’foreigner’ (forastero) or attaching oneself to an hacienda.

4

the percentage of males who at some point participated in the mita was considerably

higher. It is also of relevance to note that mita districts contain a non-trivial fraction of

Peru’s contemporary population. Calculations using the most recent population census

reveal that 17% of the Peruvian population (and one quarter of the population living

outside the Lima metropolitan area) reside in districts that contributed to the mita

(INEI, 1993).

Historical evidence indicates that authorities enforced the mita throughout the sub-

jected region. Local native elites were responsible for collecting conscripts, delivering

them to Potosı or Huancavelica, and ensuring that they reported regularly for mine

duties (Cole, 1985, p. 15; Bakewell, 1984). If community leaders were unable to pro-

vide their allotment of conscripts, mita captains required them to pay in silver the sum

needed to hire wage laborers instead. Historical documents and scholarship suggest that

this rule was strictly enforced (Zavala, 1980; Sanchez- Albornoz, 1978).5 Some com-

munities did commonly meet mita obligations through payment in silver, particularly

communities who because of their proximity to mining centers in present-day Bolivia

had relatively easy access to coinage (Cole, 1985). Detailed records of mita contribu-

tions from the 17th, 18th, and early 19th centuries indicate that communities in the

region that this paper examines contributed primarily in people (Tandeter, 1993, p. 56,

66; Zavala, 1980, II, p. 67-70). This is corroborated by comprehensive population data

collected in a 1689 parish census (Villanueva Urteaga, 1982), described in more detail in

the data appendix, which shows that the male-female ratio was 22% lower in mita dis-

tricts (a difference that is significant at the 1% level). Moreover, demographic evidence

suggests greater population decline in mita districts than in nearby non-mita districts,

particularly during the seventeenth century (Wightman, 1990, p. 72).

While colonial observers highlighted the deleterious effects of the mita on demography

and well-being in subjected communities,6 there are some features of the institution that

could have promoted relatively better outcomes for mita districts. For example, mita

conscripts sold locally produced goods in Potosi, generating trade linkages with this

important economic center. Moreover, to the extent that conscripts or family members

who accompanied them earned monetary income that exceeded subsistence demands,

5Community leaders faced confiscation of oftentimes substantial personal property, removal fromoffice, physical abuse, and imprisonment if they failed to comply (Garrett, 2005, p. 126; Cole, 1985, p.44).

6See for example the volumes of formal complaints housed at Archivo General de Indias in Sevilla,Spain.

5

this could have provided a source of funds for investment upon return to their home

communities. Despite such potential benefits, this paper documents that the persistent

effects of the mita on living standards have been large and negative and sheds light on

why this has been the case.

With silver deposits depleted, the Court of Cadiz abolished the mita in 1812.7 At

the date of its dissolution, the mita had been in force for almost 240 years. Sections 3

and 4 will discuss historical and empirical evidence showing divergent histories of mita

and non-mita districts.

2.2 The Mita ’s Assignment

Why did colonial authorities require only a portion of districts in Peru to contribute

to the mita, and how did they determine which districts to subject? There is little

doubt that the mita’s objective was to increase royal silver revenues through lowering

mining wages (Levillier, 1921 [1572], 4, p. 108). However, coercing labor also imposed

costs – administrative and enforcement costs, compensation to conscripts for traveling

to and from the mines (as much as two months each way from the region I examine),

and the risk of decimating Peru’s indigenous population, as had occurred in earlier

Spanish mining ventures in the Caribbean (Tandeter, 1993, p. 61; Cole, 1985, p. 3, 31;

Arzans, 1975, vol. 1, p. 43-46; Canete, 1973 [1794]). To establish the minimum number

of conscripts required to revive silver production to previous levels, Peruvian Viceroy

Francisco Toledo commissioned a detailed inventory of mines and production processes

in Potosı and elsewhere in Peru in 1571 (Bakewell, 1984, p. 76-78; Levillier, 1921

[1572], 4). These numbers were used, together with census data collected in the early

1570’s, to enumerate the mita assignments. The limit that the mita subject no more

than one seventh of a community’s adult male population at a given time was already

an established rule that regulated local labor drafts in Peru (Glave, 1989). Together

with estimates of the required number of conscripts, this rule roughly determined what

fraction of Andean Peru’s districts would need to be subjected to the mita.

Historical documents and scholarship reveal two criteria used to assign the mita:

distance to the mines at Potosı and Huancavelica and elevation. Important costs of ad-

ministering the mita, such as travel wages and the probability that conscripts would die

7Peru gained independence from Spain in 1821.

6

or desert en route, were increasing in distance to the mines (Tandeter, 1993, p. 60; Cole,

1985, p. 31). Moreover, Spanish officials believed that only highland peoples could sur-

vive intensive physical labor in the Potosı and Huancavelica mines, both located at over

4000 meters (13,000 feet) (Golte, 1980). The geographic extent of the mita is consistent

with the application of these two criteria. Figure 1 shows that an elevation constraint

was binding along the eastern and western mita boundaries, which tightly follow the

steep Andean precipice. The southern Potosı mita boundary was also constrained, by

the border between Peru and the Viceroyalty of Rio de la Plata (Argentina) and by the

geographic divide between agricultural lands and an uninhabitable salt flat. This study

focuses on the portion of the mita boundary that transects the Andean range (Figure

2). The districts along this portion of the boundary are termed the study region.8 Here,

exempt districts were the ones located furthest from the mining centers given the road

networks at the time (Hyslop, 1984).9

The key RD identifying assumption is that all relevant factors besides treatment

vary smoothly at the mita boundary. Distance to the mines clearly varies smoothly,

and I focus on the segment of the boundary that transects the Andean range; hence

elevation does not jump. Concerns remain that factors not mentioned in the historical

literature may have been used to determine the mita boundary and may also impact

economic outcomes today. I examine the following potentially important characteristics:

elevation, terrain ruggedness, soil fertility, rainfall, ethnicity, pre-existing settlement

patterns, variation in Inca institutions, and local economic prosperity and institutions

just prior to the mita’s enactment. To be conservative, the statistics that follow, and

8To provide context, it is useful to note that the study region as a whole is poorer and more ruralthan the Peruvian average. Equivalent household consumption in the study region - as measured bythe nationally representative Encuesta Nacional de Hogares (ENAHO, see section 3.1) - is 57% of thePeruvian mean (ENAHO, 2001). In the ENAHO sample, 11% of households in the study region live intowns with more than 4,000 inhabitants, as compared with 48% of households nationally.

9This discussion suggests that exempt districts were those located relatively far from both Potosı andHuancavelica. The correlation between distance to Potosı and distance to Huancavelica is -0.996, makingit impossible to separately identify the effect of distance to each mine on the probability of receivingtreatment. Thus, I divide the sample into two groups - municipalities to the east and those to thewest of the dividing line between the Potosı and Huancavelica mita catchment areas. When consideringdistricts to the west (Potosı side) of the dividing line, a flexible specification of mita treatment on a cubicin distance to Potosı, a cubic in elevation, and their linear interaction shows that being 100 additionalkilometers from Potosı lowers the probability of treatment - evaluated at the mean - by 0.873, with astandard error of 0.244. Being 100 meters higher increases the probability of treatment by 0.061, witha standard error of 0.027. When looking at districts to the east (Huancavelica side) of the dividingline and using the same specification with a polynomial in distance to Huancavelica instead of distanceto Potosı, the marginal effect of distance to Huancavelica, evaluated at the mean, is negative but notstatistically significant.

7

all results in this paper, exclude metropolitan Cusco. Metropolitan Cusco is composed

of seven non-mita and two mita districts located along the mita boundary. It was the

capital of the Inca Empire, with an estimated population of around 100,000 at the date of

Spanish conquest (Cook, 1981, p. 212-214; Cieza de Leon, 1959, p. 144-148). I exclude

Cusco because part of its relative prosperity today likely relates to its pre-mita heritage

as the Inca capital. When Cusco is included, the impacts of the mita are estimated to

be even larger.

To examine elevation - the principal determinant of climate and crop choice in Peru

- as well as terrain ruggedness, I use 30 arc second (one kilometer) resolution data

produced by NASA’s Shuttle Radar Topography Mission (SRTM, 2000). I divide the

study region into twenty by twenty kilometer grid cells, approximately equal to the

mean size of the districts in my sample, and calculate the mean elevation and slope

within each grid cell (Figure 2).10 These grid cells then act as the unit of observation.

Spatial correlation is a serious concern with means comparisons of geographic measures,

and hence I report standard errors corrected for spatial correlation in square brackets.

Following Conley (1999), I allow for spatial dependence of an unknown form.11 For

comparison, I report robust standard errors in parentheses.

Means comparisons are reported in Table 1. The first set of columns restricts the

sample to fall within 100 kilometers of the mita boundary and the second, third, and

fourth set of columns restrict it to fall within 75, 50, and 25 kilometers, respectively.

Row 1 shows that elevation is statistically identical across the mita boundary.12 I next

look at terrain ruggedness, using the SRTM data to calculate the mean uphill slope

in each grid cell. In contrast to elevation, there are some statistically significant, but

relatively small, differences in slope. These disappear as one moves closer to the bound-

ary. Although potentially concerning, any differences that exist are not likely to pose a

threat to identification. Available empirical evidence suggests that the direct effects of

terrain ruggedness would tend to reduce economic prosperity, biasing estimates of the

10All results are similar if the district is used as the unit of observation instead of using grid cells.11Specifically, the Conley covariance matrix is a weighted average of spatial autocovariances, where

the weights are the product of Bartlett kernels in two dimensions (North/South and East/West). Theystart at one and decline linearly to zero when a pre-specified cut point is reached. I choose the cutoff inboth dimensions to be one degree (approximately 100 kilometers); choosing other cut points producessimilar estimates.

12Elevation remains identical across the mita boundary if I restrict the sample to inhabitable areas(<4800 m) or weight by population, rural population, or urban population, using 30 arc second (onekilometer) resolution population data (SEDAC, 2007).

8

mita’s impact downwards since mita districts are less rugged (Nunn and Puga, 2007).

Moreover, terrain ruggedness is observable. I control for it, and it is rarely statistically

significant.13

Row 3 examines ethnicity using data from the 2001 Peruvian National Household

Survey, which is described in greater detail in the data appendix. A household is defined

as indigenous if the primary language spoken in the household is an indigenous language

(usually Quechua). Results show no significant differences in ethnic identification across

the mita boundary.

Spanish authorities could have based mita assignment on settlement patterns, insti-

tuting the mita in densely populated areas and claiming land for themselves in sparsely

inhabited regions where it was easier to usurp. Evidence does not suggest differential

settlement patterns across the mita boundary at the date of enactment. A detailed

review by Bauer and Covey (2002) of all archaeological surveys in the region surround-

ing the Cusco basin indicates no large differences in settlement density at the date of

Spanish Conquest. Moreover, there is no evidence suggesting differential rates of pop-

ulation decline in the forty years between Spanish conquest and the enactment of the

mita (Cook, 1981, p. 108-114).

The term mita was first used by the Incas to describe the system of labor obligations

that supported the Inca state. While the Spanish co-opted this phrase to describe

their system of mine labor, the Inca and Spanish mitas were very different institutions,

and historical evidence strongly supports independent assignment.14 Centrally, the Inca

13I also examined data on district soil quality and rainfall (results available upon request). ThePeruvian Institute for Natural Resources (INRENA, 1997) has assigned each district an agriculturepotential score on a scale from 0 to 300, with greater values denoting higher agricultural potential.INRENA determined these scores through district field surveys that examined soil depth, texture, anddrainage. The INRENA scores reveal higher soil quality in mita districts, which is not surprising givenresults on land tenure discussed in Section 4. I do not emphasize soil quality because it is endogenous toland usage. On the other hand, while climate is exogenous, high resolution data are not available andinterpolated climate estimates are notoriously inaccurate for the mountainous region examined in thisstudy (Hijmans et al., 2005). Temperature is primarily determined by altitude (Golte, 1980; Pulgar-Vidal, 1950), and the similarities in altitude imply that it is quite unlikely that there are substantialdifferences in temperature across the mita boundary. To examine precipitation, I use station datafrom the Global Historical Climatology Network, Version 2 (Peterson and Vose, 1997), which are notavailable for all districts in the sample. Using all available data (from stations in 50 districts locatedwithin 100 km of the mita boundary), mita districts appear to receive somewhat higher average annualprecipitation, and these differences disappear when comparing districts closer to the mita boundary.When using only stations with at least twenty years of data (to ensure a long-run average), whichprovides observations from twenty different stations (eleven outside the mita catchment and nine inside),the difference declines somewhat in magnitude and is no longer statistically significant.

14It should also be noted that the Inca and Spanish mitas served different purposes. The Inca m’ita

9

m’ita required every married adult male in the Inca Empire (besides leaders of large

communities), spanning an area far more extensive than the study region, to provide

several months of labor services for the state each year.15

Row 4 examines detailed data on local tribute (tax) rates, collected just prior to the

mita’s 1573 enactment. The Peruvian viceroy, Francisco Toledo, blamed demographic

collapse on excessive, unregulated rates of tribute extraction by local Spanish elites.

Thus, he coordinated an in depth inspection of modern Peru, Bolivia, and Ecuador in

the early 1570’s to evaluate the maximum tribute that could be demanded from local

groups without threatening subsistence. Based on their assessment of ability to pay,

authorities assigned varying tribute obligations at the level of the district - socioeco-

nomic group, with each district containing one or two socioeconomic groups.16 These

per capita contributions reflect Spanish authorities’ best estimates of local economic

prosperity, with more prosperous groups paying greater tribute. The assignments have

been preserved for all districts in the study region (Miranda, 1975 [1583]), and summary

statistics are provided in Appendix Table A1. Row 4 of Table 1 shows average tribute

contributions per adult male (women, children, and those over age fifty were not taxed).

Simple means comparisons across the mita boundary do not find statistically significant

differences, documenting that economic prosperity, as measured by Spanish officials, was

similar prior to the mita’s enactment.

Rows (5) through (8) examine district level data on how Spanish authorities allo-

cated these tribute revenues. Revenues were divided between rents for Spanish nobility

(encomenderos, row (5)), salaries for Spanish priests (row (6)), salaries for local Spanish

administrators (justicias, row (7)), and salaries for indigenous mayors (caciques, row

provided public goods, such as maintenance of road networks and sophisticated irrigation and croppingsystems that required inter-community coordination of labor (D’Altoy, 2002). The majority of Incasubjects performed their m’ita obligations in or near their home communities, often in agriculture;service in mines was extremely rare (D’Altoy, 2002, p. 266; Rowe, 1946, 267-269). In contrast, theSpanish mita acted as a subsidy to private mining interests and the Spanish state, which used taxrevenues from silver production largely to finance European wars (Cole, 1985, p. 20).

15The Incas used knot records to keep track of communities’ m’ita obligations, and archaeologistshave located some of these. The labor services they record are proportional to the number of householddwellings (D’Altoy, 2002, p. 266). This corroborates colonial chronicles, which report that the Incam’ita applied uniformly to all Inca subjects (Cieza de Leon (1967 [1551]).

16Teams of surveyors were ordered to list the ages and occupations of residents; inspect the com-munities’ grain storage facilities; uncover the tribute that residents provided in the past; investigate aseries of geographic and economic questions relating to natural resources and agricultural production;record the tribute, labor services, and land received by indigenous leaders and Spanish administrators;and investigate a variety of other questions.

10

(8)). The data on tribute revenue allocation are informative about the financing of local

government, about the extent to which Spanish nobility extracted local revenues, and

about the relative power of competing local administrators to obtain tribute revenues.

Table 1 reveals some modest differences: when the sample is limited to fall within 100

km or 75 km from the mita boundary, we see that Spanish nobility received a slightly

lower share of tribute revenue inside the mita catchment than outside (60% versus 64%),

whereas Spanish priests received a slightly higher share (21% versus 19%). All differ-

ences disappear as the sample is limited to fall closer to the mita boundary. These data

will be examined further in Section 3.

3 The Mita and Long Run Development

3.1 Data

I examine the mita’s long run impact on economic development by testing whether it

affects living standards today. A list of districts subjected to the mita is obtained from

Saignes (1984) and Amat y Junient (1947) and matched to modern districts using sources

discussed in the data appendix. Peruvian districts are in most cases small political units

that consist of a population center (the district capital) and its surrounding countryside.

Mita assignment varies at the district level.17

I measure living standards using two independent datasets, both geo-referenced to

the district. Household consumption data are taken from the 2001 Peruvian National

Household Survey (ENAHO) collected by the National Institute of Statistics (INEI).

To construct a measure of household consumption that reflects productive capacity, I

subtract the transfers received by the household from total household consumption, and

normalize to Lima metropolitan prices using the deflation factor provided in ENAHO. I

also utilize a micro census dataset, obtained from the Ministry of Education, that records

the heights of all six to nine year old school children in the region. Following international

standards, children whose heights are more than two standard deviations below their

age-specific median are classified as stunted, with the medians and standard deviations

calculated by the World Health Organization from an international reference population.

Because stunting is related to malnutrition, to the extent that living standards are lower

17A complete list of modern districts and the sources used to match each one to colonial districts isgiven in the data appendix.

11

in mita districts, we would also expect stunting to be more common there.18 While

the height census includes only children enrolled in school, 2005 data on primary school

enrollment and completion rates do not show statistically significant differences across

the mita boundary, with primary school enrollment rates exceeding 95% throughout the

region examined (MINEDU, 2005b).

Geospatial data are used to construct controls for exogenous geographic characteris-

tics, and these controls are subsequently included in all regressions. I calculate the mean

area weighted elevation of each district by overlaying a map of Peruvian districts on the

SRTM data, discussed in the previous section, and employ a similar procedure to obtain

each district’s mean area weighted slope.19 I also utilize soil type data, at a scale of

one to five million, produced by the Soil Terrain Database for Latin America (SOTER-

LAC).20 I construct a series of soil type dummies equal to one for the soil type(s) that

predominate over the greatest percentage of the district’s landmass area. Results are

similar when I use the INRENA soil quality measure, discussed in footnote 13, instead

of soil type dummies.

3.2 Estimation Framework

Mita treatment is a deterministic and discontinuous function of a known covariate: geo-

graphic location. This suggests estimating the mita’s long run impacts using a regression

discontinuity design. Specifically, I run regressions of the following form:

yidb = α + γ mitad + X ′β + f(distd) + φb + εidb (1)

where yidb is the outcome variable of interest for observation i in district d along

segment b of the mita boundary, and mitad is an indicator equal to 1 if district d con-

tributed to the mita and equal to zero otherwise. X is a vector of covariates that includes

soil type indicators, the mean area weighted elevation and slope for district d, and (in

18To the extent that height is influenced by birthweight and birthweight influenced by mother’s height,stunting may reflect past malnutrition as well as (or instead of) current undernutrition. The extent ofintergenerational transmission of stunting is controversial (Deaton and Dreze, 2008).

19I choose not to weight by population because it may be endogenous to the mita. When I insteadweight by population and employ these controls in the analysis that follows, results are unchanged.

20SOTERLAC uses the standard FAO soil type categories. The FAO classifies soils based on theirmineral properties (i.e. the degree to which they are saline, lithic, or stony), as well as on the size andmixture of mineral and clay particles.

12

regressions with equivalent household consumption on the lefthand side) demographic

variables giving the number of infants, children, and adults in the household. f(distd)

is the RD polynomial, which controls for smooth functions of distance (various forms

will be explored), and φb is a set of boundary segment fixed effects that denote which

of four equal length segments of the boundary is the closet to the observation’s district

capital.21 To be conservative, all analysis excludes metropolitan Cusco.22

I also report results from the following specification, which includes additional inter-

action terms between the RD polynomial and treatment dummy:

yidb = α + γ mitad + X ′β + f(distd) + f(distd) ∗mitad + φb + εidb (2)

where all variables are as defined above.

The following identifying assumptions are necessary for γ to estimate the causal

impact of the mita. First, and most centrally, identification requires all relevant factors

besides treatment to vary smoothly at the mita boundary. That is, letting y1 and

y0 denote potential outcomes under treatment and control, identification requires that

E[y1|dist = d] and E[y0|dist = d] are continuous in d at the boundary. We need this

assumption for individuals located just outside the mita catchment to be an appropriate

counterfactual for those located just inside it. The evidence discussed in Section 2.2

supports its plausibility.

A second identifying assumption is no selective sorting across the mita boundary. A

small direct mita effect could provoke high rates of out-migration of the most productive

individuals, leading to a much larger indirect effect. If this were the case, it would still

suggest a long run impact of the mita, but the interpretation would be different. Section

3.3 examines data on population and migration, from 1876 through the most recent

census. These data - combined with historical evidence - suggest that population in

the region has been quite immobile over the past 130 years. Section 3.3 also discusses

evidence on migration during the colonial period.

To simplify and shorten the exposition, I focus throughout the paper on specifications

that use distance as the running variable. (Table 3 (living standards) and Appendix Ta-

ble A2 (all main outcome variables) report results from an alternative strategy that uses

21Results (available upon request) are robust to allowing the running variable to have heterogeneouseffects by including a full set of interactions between the boundary segment fixed effects and f(distd).

22When Cusco is included, the mita’s impact is estimated to be even larger. See Table 4.

13

latitude, longitude, and their interactions to control for smooth functions of geographic

location.) Polynomials in distance can be used to control for characteristics that change

smoothly at the mita boundary, a discontinuity whose location is given by coordinates in

x-y space. Because distance is one-dimensional, there is more than one plausible way to

use distance polynomials to control for smoothly changing unobservables: two obvious

options are distance to Potosı and distance to the mita boundary. Distance to Potosı is

the underlying determinant of treatment and is a clearly meaningful economic variable,

given Potosı’s position as the largest city in the Western Hemisphere during much of the

colonial period. An alternative model holds that any omitted variables change smoothly

as the mita boundary is approached, suggesting distance to the mita boundary as the

running variable. While I focus on distance to Potosı as a main specification because of

its more direct mapping into economic variables, it would be problematic if the choice

of running variable mattered substantially for the results. Throughout the paper, I thus

report two specifications, each using one of these running variables. In addition, Table

3 and Appendix Table A2 examine robustness to RD functional form assumptions for

both of the running variables.

Another choice in specifying the RD polynomial is whether or not to interact it with

the mita dummy, with prominent papers in the existing literature pursuing both routes

(Chay et al., 2005; Lee et al., 2004; van der Klaauw, 2002; Angrist and Lavy, 1999) and

limited reasons for preferring one over the other (Angrist and Pischke, 2009, p. 255).

Interacting yields a more flexible specification, but also tends to lead to noisier estimates

and may make estimates more sensitive to outliers near the boundary. Throughout

the paper, I report a parsimonious specification that includes a quadratic polynomial

in Euclidean distance to the mines at Potosı,23 and a more flexible specification that

includes a quadratic polynomial in Euclidean distance to the mita boundary, interacting

this polynomial with the mita dummy and evaluating the effect at the boundary.24

The latter specification is the most flexible one that can be estimated with reasonable

precision across all samples in the paper.25 Table 3 and Appendix Table A2 document

23I include only a polynomial in distance to Potosı, rather than polynomials in both distance to Potosıand distance to Huancavelica, because the correlation between the two is -0.996.

24Estimates are robust to calculating distance in a number of different ways: distance accounting forchanges in elevation, distance according to a least cost algorithm that imposes penalties for ascent anddescent, and distance averaged over the district’s entire surface area rather than calculated from thedistrict capital.

25Coefficients in the household consumption regressions become large and imprecisely estimated whenexamining specifications that interact cubic or higher order polynomials in distance to the boundarywith the mita dummy.

14

that the paper’s results are generally robust to using a number of other functional forms

for the RD polynomial, reporting specifications that interact distance to Potosı with the

mita dummy and that do not interact distance to the boundary with the mita dummy.

I begin by estimating the long-run impact of the mita on living standards today.

First, I test for a mita effect on household consumption, using the log of equivalent house-

hold consumption, net transfers, in 2001 as the dependent variable.26 Panel A reports

results from the specification that includes a quadratic polynomial in Euclidean distance

to the mines at Potosı, and Panel B reports the specification that uses a quadratic poly-

nomial in Euclidean distance to the mita boundary, interacted with the mita dummy.

The first column of Table 2 limits the sample to districts within 100 kilometers of the

mita boundary, and columns (2) through (4) restrict it to fall within 75, 50, and 25

kilometers, respectively. Columns (5) through (9) repeat this exercise, using as the de-

pendent variable a dummy equal to one if the child’s growth is stunted and equal to

zero otherwise. Column (5) limits the sample to districts within 100 kilometers of the

mita boundary, and columns (6) through (8) restrict it to fall within 75, 50, and 25

kilometers, respectively. Column (9) limits the sample to only those districts bordering

the mita boundary.27 When combined with the inclusion of boundary segment fixed

effects, this ensures that I am comparing observations in close geographic proximity.

3.3 Estimation Results

The estimates in columns (1) through (4) support a large long run mita effect on house-

hold consumption. Choosing a relatively conservative estimate from Panel A, Column 2,

we see that a long-run mita effect lowers household consumption by around 32 percent in

subjected districts. The point estimates are fairly similar across samples, and I cannot

reject that the mita coefficients estimated by the two specifications are the same. How-

ever, while three of the estimated mita coefficients in Panel A are statistically significant

at the 1% or 5% level, the point estimates from the more flexible specification in Panel

B are not statistically significant.28

26Following Deaton (1997), I assume that children aged 0 to 4 are equal to 0.4 adults and childrenaged 5 to 14 are equal to 0.5 adults.

27This is not feasible with the household consumption data, since limiting the sample to districtsbordering the mita boundary leaves only sixteen clusters.

28The results in Panel B are unreliable at 25 km from the boundary, given the low number of clustersand observations and large number of parameters.

15

Columns (5) through (9) of Table 2 examine census data on stunting. This dataset

contains substantially more observations and clusters than the consumption sample, pro-

viding more power for estimating the specification in Panel B.29 Moreover, childhood

stunting is an important outcome associated with detrimental impairments including

poor health and diminished capacity to learn (Victora et al., 2008). When using only

observations in districts that border the mita boundary, point estimates of the mita

effect on stunting range from 5.1 (s.e. = 0.024) to 10.2 (s.e. = 0.051) percentage points.

This compares to a mean prevalence of stunting of 40% throughout the region exam-

ined. The mita dummy is always statistically significant, is similar in magnitude across

samples, and is somewhat larger and more imprecisely measured when I use a quadratic

polynomial in distance to the mita boundary interacted with the mita dummy.30 I can-

not reject at the 10% level that the estimates of the mita coefficient are the same in

Panels A and B. Overall, the large, robust, and statistically significant impact of the

mita on stunting, combined with its consistently large but, in Panel B, more noisily es-

timated impact on household consumption, provide strong evidence for a long-run mita

effect on living standards.

The results from Table 2 can be seen graphically in Figure 3, Panels A and B. Panel

A plots household consumption, and Panel B plots the percentage of children in each

district who have stunted growth. Circles to the left of the vertical line fall outside the

mita catchment and circles to the right fall inside. The thick line gives the predicted

outcome from a regression that includes a second order polynomial in distance to Potosı

and the mita dummy, and the thin lines are 95% confidence bands.

Table 3 documents the robustness of the mita effect to various RD functional form

assumptions. I show estimates for the same samples examined in Table 2, and to conserve

space, present just the mita coefficients. The first five rows report results from alternative

specifications using an RD polynomial in distance to Potosı. Rows (1), (2), and (3)

include a linear, cubic, or quartic polynomial, respectively, and rows (4) and (5) interact

the mita dummy with a linear or quadratic polynomial.31 The mita coefficients for both

29An additional concern with the consumption data is that while they are representative at thedepartmental level, they may not be representative at the district level. The height census suggeststhat it is unlikely that the mita effect is primarily driven by non-representative sampling.

30A similar picture emerges when I use height in centimeters as the dependent variable and includequarter x year of birth dummies, a gender dummy, and their interactions on the righthand side.

31The mita effect is evaluated at the mean distance to Potosı for observations very near (<10 kmfrom) the mita boundary. Results are broadly robust to evaluating the mita effect at different averagedistances to Potosı - i.e. for districts < 25 km from the boundary, for bordering districts, or for all

16

household consumption and stunting in rows (1) through (3) are similar in magnitude to

those in Table 2 and all but one of the 27 point estimates are statistically significant. The

coefficients in rows (4) and (5) tend to be similar to those from the specifications that

do not interact the polynomial with the mita dummy. An exception is the specification

that interacts a quadratic polynomial with the mita dummy, which estimates smaller

and statistically insignificant mita impacts on stunting. Next, rows (6) through (9)

examine the robustness of specifications that use distance to the mita boundary as the

running variable. Row (6) includes a linear polynomial in distance to the mita boundary,

interacted with the mita dummy, estimating (imprecise) coefficients similar in magnitude

to those in Panel B of Table 2.32 Rows (7) through (9) then report specifications that

do not interact distance to the boundary with the mita dummy, using linear, quadratic,

and cubic polynomials, respectively. All 27 estimated effects are statistically significant

and of similar magnitude to the estimates in Table 2.33

Another reasonable approach is to include the polynomials from both of the baseline

specifications - a quadratic polynomial in distance to Potosı and an interacted quadratic

polynomial in distance to the mita boundary, and Row (10) reports this specification.

Estimates are of similar magnitude and statistical significance to those from the specifi-

cation including only the interacted quadratic polynomial in distance to the boundary.

On a different note, since the treatment threshold reflects a discontinuity in x-y space,

an alternative strategy is to control for polynomials in latitude, longitude, and their in-

teractions. Row (11) reports results from a regression including a quadratic polynomial

in latitude and a quadratic polynomial in longitude, as well as their interaction and

interaction squared. Estimates are again similar in magnitude to those in Table 2, and

statistically significant in all columns. Finally, row (12) presents estimates using ordi-

nary least squares, reporting similar statistically significant mita effects. Results (not

shown) are also robust to including higher order polynomials in elevation and slope. In

sum, Table 3 provides robust evidence that a persistent mita effect lowers equivalent

household consumption by around one third and increases the prevalence of stunting by

around five percentage points.

districts.32Estimates for the mita’s impact on household consumption become large and noisily measured

when cubic or higher order splines in distance to the mita boundary are included. Estimates of themita effect on stunting are robust to including cubic and quartic quintic splines in distance to the mitaboundary (though the specifications are somewhat more noisily estimated).

33A specification with a quartic polynomial - omitted to save space - also estimates similar results.

17

Given this robustness to functional form assumptions, Table 4 reports a number of

additional robustness checks using the two baseline specifications of the RD polynomial:

a quadratic in distance to Potosı and a quadratic in distance to the mita boundary,

interacted with the mita dummy. To conserve space, I report estimates only from the

sample that contains districts within 50 km of the mita boundary. Columns (1) through

(7) examine the household consumption data and columns (8) through (12) the stunting

data. For comparison purposes, columns (1) and (8) present the baseline estimates from

Table 2 for consumption and stunting respectively. Column (2) shows that adding a

control for ethnicity, equal to one if an indigenous language is spoken in the household

and zero otherwise, does not substantially change the coefficient on mitad, now equal

to -0.270 (s.e. = 0.103) when a quadratic in distance to Potosi is included and equal to

-0.329 (s.e. = 0.250) when I interact a quadratic polynomial in distance to the boundary

with the mita dummy. While ethnic identification is probably endogenous to income,

complicating the interpretation of the coefficients, this result nevertheless suggests that

the mita’s impact cannot be explained primarily by an effect on ethnic identification.

Next, columns (3) and (9) include metropolitan Cusco, and the mita coefficients remain

large and of similar statistical significance.

In response to the potential endogeneity of the mita to Inca landholding patterns,

columns (4) and (10) exclude districts that contained estates of Inca royalty.34 Simi-

larly, columns (5) and (11) exclude districts falling along portions of the mita boundary

formed by rivers, to account for one way in which the boundary could be endogenous

to geography. Finally, column (6) estimates consumption equivalence flexibly, using log

household consumption as the dependent variable and controlling for the ratio of children

to adults and the log of household size. In all cases (except the Inca estate result using

the more flexible specification), point estimates and significance levels remain similar.35

A small direct mita effect could provoke out-migration of the most productive indi-

viduals, leading to a much larger indirect effect. Data from the 1993 Population Census

34The Inca chose their estates to serve sacred as opposed to productive purposes (some existed onlyafter they were reclaimed from swampland) (Niles, 1987, p. 13). According to Spanish law, the SpanishCrown inherited personal possession of these properties upon Inca defeat.

35Excluding the Inca estate districts implies dropping five non-mita districts located near the mitaboundary. The flexible specification estimates the trend relation separately on either side of the bound-ary, and thus dropping these observations decreases the non-mita sample size sufficiently to lead toclearly unreliable estimates. When I estimate the flexible specification using all districts within 100km of the boundary - besides those containing Inca estates - the mita coefficient equals -0.359 (s.e. =0.302).

18

reveal that there are no statistically significant differences in rates of out-migration be-

tween mita and non-mita districts, though rates of in-migration are around 4.8% higher

outside the mita catchment.36 Columns (7) and (12) present a (conservative) test of

whether differential rates of migration in recent years could be primarily responsible for

living standards differences between mita and non-mita districts, by omitting the 4.8%

of the non-mita sample with the highest equivalent household consumption and least

stunting, respectively. Estimates remain of similar magnitude and statistical significance

to the baseline, documenting that migration today is not the primary force responsible

for the mita effect. In understanding why individuals do not arbitrage income differ-

ences between mita and non-mita districts, it is useful to note that much of the region’s

population lives in formally recognized indigenous communities. For example, in the

largely agricultural region examined, the 1994 Agricultural Census documents that over

half of households obtained their plots from an indigenous community. It tends to be

difficult to gain membership and land in a different indigenous community, making ma-

jor cities the primary feasible destination for most immigrants (INEI, 1993). Cities have

various disamenities, and individuals living in indigenous communities typically forfeit

their claims to land held in usufruct if they migrate.

Historical migration over the past 130 years also appears to have been low. Data

from the 1876, 1940, and 1993 population censuses show a (quite high) district level pop-

ulation correlation of 0.87 between 1940 and 1993 for both mita and non-mita districts.

Similarly, the population correlation between 1876 and 1940 is 0.80 in mita districts and

0.85 in non-mita districts. While a constant aggregate population distribution does not

preclude extensive sorting, this is unlikely given the relatively closed nature of indigenous

communities and the stable linkages between haciendas and their attached peasantry

(Morner, 1978). The limited migration that has occurred since independence appears to

have been primarily directed towards Lima (Stein, 1980, p. 62).

On the other hand, 17th century population data - available for 15 mita districts

and 14 non-mita districts - provides evidence consistent with the hypothesis that indi-

viduals migrated disproportionately from mita to non-mita districts while the mita was

in force.37 To the extent that flight was selective and intelligence or other character-

36The 2005 Population Census was methodologically flawed and re-conducted in 2007. Results arenot yet available.

37The data are drawn from the 1689 Cusco parish reports, described in the data appendix. Theyprovide information on the number of males in the district whose ancestors were not assigned to thatdistrict by the Spanish in the late 16th century; these individuals were termed forasteros (foreigners),

19

istics are heritable enough to persist over several hundred years, differential historical

migration could contribute to the estimated mita effect.

If the RD specification is estimating the mita’s long run effect, as opposed to some

other underlying difference, being inside the mita catchment should not affect economic

prosperity, institutions, or demographics prior to the mita’s enactment. In a series of

specification checks, I first regress the log of the mean district 1572 tribute contribution

per adult male on the variables used in the stunting regressions in Table 2. The coeffi-

cient on mitad, reported in column (1) of Table 5, is small and statistically insignificant,

providing evidence that economic prosperity - as measured by the colonial government

- was similar across the boundary prior to the mita’s enactment. Columns (2) through

(5) then examine the shares of tribute revenues allocated to rents for Spanish nobility

(encomenderos, column (2)), salaries for Spanish priests (column (3)), salaries for lo-

cal Spanish administrators (justicias, column (4)), and salaries for indigenous mayors

(caciques, column (5)). There is little evidence for differences across the boundary, docu-

menting similarities in the power of Spanish nobility, the Church, and other local officials

to claim tribute revenues. Finally, columns (6) through (8) examine demographics, with

the population shares of tribute paying males (those aged 18 to 50), boys, and women as

the dependent variables. The colonial census data fail to indicate statistically significant

differences in population across the mita boundary.

To achieve credible identification, the above RD specifications exploit variation across

observations located close to the mita boundary. If the boundary is an unusual place,

these estimates may have little external validity. One way to shed light on whether

observations near the mita boundary are atypical is to limit the sample to districts more

than a threshold distance from the boundary. When I use ordinary least squares to

estimate the correlation between the mita and the main outcome variables (including

those that will be examined in Section 4) for districts located between 25 km and 100

km from the mita boundary, the estimates are quite similar to those obtained from the

RD specifications (results available upon request). This substantially alleviates concerns

that districts near the boundary are unusual. Nevertheless, it could still be the case that

the segment of the boundary that I examine is atypical. While I cannot use regression

and colonial policy prohibited them from holding land in usufruct in their resident communities. In the14 non-mita districts, 52.5% of individuals had ancestors who had not been assigned to their currentdistrict of residence, as compared to only 35% in the 15 mita districts. While we do not know thatthe higher percentage of forasteros in non-mita districts resulted from individuals fleeing mita service,qualitative evidence suggests that this was important (Wightman, 1990).

20

discontinuity to empirically identify the long-run impacts of the mita along the entire

mita boundary - since other characteristics change discontinuously - correlations between

the mita and living standards calculated along the entire mita boundary within Peru are

consistent in magnitude with the effects documented above.38 This evidence suggests

that the RD estimates are informative about the mita’s overall persistent impacts.

In summary, Tables 3 through 5 support a valid regression discontinuity design, show-

ing an economically meaningful impact of the mita on prosperity today. These results

are consistent with the cross-country literature on the long run impacts of institutions

(Nunn, 2008; Acemoglu et al., 2001, 2002) and provide microeconomic evidence that

historical institutions matter. The question remains: why would the mita affect eco-

nomic prosperity nearly 200 years after its abolition? To open this black box, I turn to

an investigation of channels of persistence.

4 Channels of Persistence

This section uses data from the Spanish Empire and Peruvian Republic to test the

channels through which the mita’s impacts persist.39 In all tables, Panel A reports the

specification that uses a quadratic polynomial in distance to Potosı, and Panel B reports

the specification that uses a quadratic polynomial in distance to the mita boundary,

interacted with the mita dummy. Twelve other specifications are reported in Appendix

Table A2. To shorten the exposition, I present only estimates where the sample is limited

to fall within 50 kilometers of the mita boundary, and metropolitan Cusco is excluded.

The main results can be seen graphically in Figure 3, Panels C through H, which plot

different outcome variables on distance to the mita boundary. Circles to the left of the

vertical line fall outside the mita catchment and circles to the right fall inside. The thick

lines give the predicted outcomes from regressions that include second order polynomials

in distance to Potosı and the mita dummy (with outcomes predicted separately on either

side of the boundary), and the thin lines are 95% confidence bands.

38When considering observations in Peru within 50 kilometers of any point on the mita boundary,being inside the mita catchment is associated with 28.4 percent lower equivalent household consumptionand an increase of 16.4 percentage points in the prevalence of stunting. When considering individualslocated within 15 kilometers of the boundary, being inside the mita catchment is associated with an 11.5percentage points higher prevalence of stunting, a correlation which falls modestly to 10.2 percentagepoints when I control for polynomials in elevation and slope.

39Descriptions of these sources are provided in the relevant subsections and the data appendix.

21

I focus on two channels suggested as particularly important by the historical liter-

ature: land tenure and public goods. The results document that the mita limited the

establishment of large landowners inside the mita catchment and, combined with histor-

ical evidence, suggest that land tenure has in turn affected public goods provision and

market participation.

4.1 Land Tenure and Labor Systems

This section examines the impact of the mita on the formation of haciendas - rural

estates with an attached peasant labor force permanently settled on the estate (Keith,

1971, p. 437). Critically, when authorities instituted the mita in 1573 (forty years after

the Spanish conquest of Peru), a landed elite had not yet formed. At the time, Peru was

parceled into encomiendas, pieces of territory in which appointed Spaniards exercised

the right to collect tribute and labor services from the indigenous population but did

not hold title to land (Keith, 1971, p. 433). Rivalries between encomenderos provoked

civil wars in the years following Peru’s conquest, and thus the Crown began to dismantle

the encomienda system during the 1570’s. This opened the possibility for manipulating

land tenure to promote other policy goals, in particular, the mita.40

Specifically, Spanish land tenure policy aimed to minimize the establishment of

landed elites in mita districts.41 Large landowners - who unsurprisingly opposed yielding

their peasants for a year of mita service - formed the state’s principal competition in ac-

cessing scarce mita labor (Larson, 1988; Sanchez-Albornoz, 1978). As Bolivian historian

Brooke Larson concisely articulates: “Haciendas secluded peasants from the extractive

institutions of colonial society” (1988, p. 171). Moreover, by protecting native access to

agricultural lands, the state promoted the ability of the indigenous community to subsi-

dize the mita. Subjected communities contained communal plots dedicated to sustaining

mita conscripts, who were paid substantially below subsistence wages (Garrett, 2005, p.

120; Tandeter, 1993, p. 58-60; Cole, 1985, p. 31). Similarly, authorities believed that

protecting access to land could be an effective means of staving off the demographic

collapse that threatened the supply of mita labor (Larson, 1982, p. 11; Cook, 1981,

40Throughout the colonial period, royal policy aimed to minimize the power of the (potentially re-bellious) landed class - landowners did not acquired the same political clout as mine owners, the mostpowerful colonial interest group (Tandeter, 1993; Cole, 1985).

41For example, land sales under Philip VI between 1634 and 1648 and by royal charter in 1654 playeda central role in hacienda formation and were almost exclusively concentrated in non-mita districts(Brisseau, 1981, p. 146; Glave and Remy, 1978, p. 1).

22

p. 108-114, 250; Morner, 1978). Finally, a policy of indigenous communal land tenure

provided opportunities for the colonial state to co-opt the support of the indigenous

elite in administering the mita. In return for ensuring the delivery of conscripts, local

authorities were permitted to extract surplus from their communities that would have

otherwise been claimed by large landowners (Garrett, 2005, p. 120; Sala i Vila, 1991, p.

65-66).42

I now test empirically whether the mita affected hacienda formation, and if so,

whether a mita effect persisted after the mita’s abolition. Specifically, I examine the con-

centration of haciendas in 1689, 1845, and 1940. The 1689 data are contained in parish

reports commissioned by Bishop Manuel de Mollinedo and submitted by all parishes

in the bishopric of Cusco, which encompassed most of the study region. The reports

list the number of haciendas located within each subdivision of the parish and were

compiled by Horacio Villanueva Urteaga (1982). For haciendas in 1845, I employ data

collected by the Cusco regional government on the percentage of the rural tributary

population (males between the ages of 18 and 50) residing in haciendas (Peralta Ruiz,

1991). Data from 1845, 1846, and 1850 are combined to form the c. 1845 dataset.43

Finally, data from the 1940 Peruvian Population Census are aggregated to the district

level to calculate the percentage of the rural population residing in haciendas.44

Table 6, column (1) estimates the mita’s impact on the number of haciendas per

district in 1689. The estimates in Panel A show that the mita lowered the number of

haciendas in subjected districts by 13.6 (with a standard error of 2.5), a very large effect

given that on average mita districts contained only one hacienda. Panel B estimates

an effect of -7.5 (s.e.= 3.8) that, while somewhat smaller, is still large and statistically

significant. Figure 3, Panel C clearly demonstrates the discontinuity. Column (2) repeats

the same regression using the number of haciendas per 1000 district residents in 1689,

and a broadly similar pattern emerges.45 The mita coefficient is large in both panels,

42Various evidence for this hypothesis is contained in David Garrett’s book on the indigenous colonialelite in the region. As Garrett states: “The establishment of haciendas and other rural properties createdcompetition for control of the rural economy and prevented rural Inca caciques [north of Cusco] fromacquiring the wealth of their peers to the south [in mita districts]” (Garrett, 2005, p. 115).

43When data are available for more than one year, figures change very little, and I use the earliestobservation.

44The number of observations in Table 6 varies primarily because the c. 1845 data are available fora somewhat more limited geographic region, and also because the number of districts increased overtime. See the data appendix.

45Given the mita’s role in provoking population collapse (Wightman, 1990, p. 72), this measure islikely endogenous, but nevertheless provides a useful robustness check.

23

highly significant in Panel A, and imprecisely estimated by the more flexible specification

in Panel B.46

Did the effect persist after the mita’s abolition? Column (3) examines the mita’s

impact on haciendas c. 1845, estimating that a persistent mita effect lowered the per-

centage of the rural tributary population in haciendas by a statistically significant 19 to

23 percentage points. Column (4), Panel A estimates that in 1940, the mita’s impact

on the percentage of the rural labor force attached to haciendas remained around 14

percentage points (s.e. = 0.048). The coefficient of -0.091 (s.e. = 0.111) estimated by

the flexible specification is broadly similar in magnitude.

While the gap between mita and non-mita districts persisted over time, the percent-

age of the rural population in haciendas nearly doubled between 1845 and 1940. This

parallels historical evidence documenting a rapid expansion in the concentration of ha-

ciendas in the late 19th and early 20th centuries. This expansion was spurred by a large

increase in land values due to late 19th century globalization and seems to have been

particularly coercive inside the mita catchment (Jacobsen, 1993, p. 226-237; Spalding,

1974; Favre, 1967, p. 243; Nunez, 1913, p. 11). No longer needing to ensure mita

conscripts, upon independence Peru abolished the communal land tenure predominant

in mita districts, but did not replace it with enforceable peasant titling (Jacobsen, 1993;

Dancuart and Rodriguez, 1902, vol. 2, p. 136). This opened the door to tactics such

as the interdicto de adquirir, a judicial procedure which allowed aspiring landowners

to legally claim “abandoned” lands that in reality belonged to peasants. Hacienda ex-

pansion also occurred through violence, with cattle hustling, grazing estate cattle on

peasant lands, looting, and physical abuse used as strategies to intimidate peasants into

signing bills of sale (Avila, 1952, p. 22; Roca-Sanchez, 1935, p. 242-43). The historical

literature documents that in response, numerous peasant rebellions engulfed mita dis-

tricts during the 1910’s and 1920’s (Ramos Zambrano, 1984; Hazen, 1974, p. 170-78). In

contrast, large landowners had been established since the early 17th century in non-mita

districts, and these districts remained relatively stable (Flores Galindo, 1987, p. 240).

By 1940, the hacienda expansion had ended, but indiscriminate banditry and livestock

rustling remained prevalent in many mita districts for decades (Jacobsen, 1993; Tamayo

Herrera, 1982).

In 1969, the Peruvian government enacted an agrarian reform bill mandating the

46Information on the size of the attached hacienda labor force is available for around 25 percent ofthe sample. Data show no differences across the mita boundary in the number of attached peasants.

24

complete dissolution of haciendas. As a result, the hacienda elite were deposed and

lands formerly belonging to haciendas were divided into “Agricultural Societies of Social

Interest” (SAIS) during the early 1970’s (Flores Galindo, 1987). In SAIS, neighboring

indigenous communities and the producers acted as collective owners. By the late 1970s,

attempts to impose collective ownership through SAIS had failed, and many of these

collective units were gradually divided and allocated to individual producers (Alvarez

and Caballero, 1980; Mar and Mejia, 1980). Differences in hacienda concentration at

the date of agrarian reform are suggested by the 1994 Agricultural Census. When

considering districts within 50 km of the mita boundary, 20% of households outside the

mita catchment received their land through the agrarian reform, as compared to only

9% of households inside the mita catchment. Column (5), using data from the 1994

Agricultural Census, documents somewhat lower land inequality in non-mita districts

today. This finding is consistent with those in columns (1) through (4), given that non-

mita districts had more large properties that could be distributed to smallholders during

the extensive agrarian reform process.

4.2 Public Goods

In this section I consider the mita’s long run effects on access to the two principal public

goods in Andean Peru: education and roads.47 Table 7 examines the mita’s impact

on education in 1876, 1940, and 2001, providing two sets of interesting results. First,

there is some evidence that the mita lowered access to education historically. In column

(1), the dependent variable is the district’s mean literacy rate, obtained from the 1876

Population Census. Individuals are defined as literate if they could read, write, or

both. Panel A estimates that a mita effect lowered literacy by around 2.4 percentage

points (s.e. = 0.006). While this suggests an absolutely small effect, it is useful to

note that literacy in mita districts was only 2.7% in 1876 (as compared to 5.2% in non-

mita districts). In contrast, the estimated effect is small and positive in Panel B. This

discrepancy is explained by two mita districts with relatively high literacy rates located

near the mita boundary, to which the more flexible specification is sensitive.48 When

47Education, roads, and irrigation are the three public goods traditionally provided in Peru (Por-tocarrero and Zimmerman, 1988). Irrigation has been concentrated along the coast, with almost noprojects in the study region.

48One of these districts is Urcos (Cusco), whose capital is one of the historically larger towns in theregion. The other is Lampa (Ayacucho), for which it is less clear why literacy would be particularlyhigh.

25

these two observations are dropped, the mita effect estimated by the specification in

Panel B is negative and statistically identical to that in Panel A.

Next consider column (2), where the dependent variable is mean years of schooling

by district, from the 1940 Population Census. Panel A estimates that the long run

impact of the mita lowered schooling by 0.22 years, with a standard error of 0.08. This

compares to a mean educational attainment in mita districts of 0.41 years. Panel B

produces a substantially smaller point estimate of 0.08 years (s.e.=0.17). While the

point estimate in Panel B is one third the magnitude of that in Panel A, I am unable

to reject at the 10% level that they are statistically identical. This suggests that while

there is some evidence for a mita effect on schooling attainment in 1940, it should be

interpreted cautiously.

Second, the evidence for a mita effect on years of schooling more recently is weak.

In column (3), the dependent variable is individual years of schooling, obtained from

ENAHO 2001.49 The mita coefficient is negative and not quite significant at the 10%

level in Panel A. It is much smaller and insignificant both in Panel B and when I use the

specification in Panel A limited to other samples (i.e. districts within 100 km or 75 km

of the boundary). Moreover, data collected by the Ministry of Education in 2005 reveal

no systematic differences in primary school enrollment rates, primary school completion

rates, secondary school enrollment rates, or secondary school completion rates (results

available upon request). Examination of data from a census of schools conducted by

the Ministry of Education in 2006 likewise showed little evidence for a causal impact

of the mita on the quality of school infrastructure or the student-to-teacher ratio. This

evidence is consistent with studies of the Peruvian educational sector, which emphasize

near-universal access and expenditures per student that are amongst the lowest in Latin

America (Saavedra, 2002; Portocarrero, 1988).

What about roads, the other principal public good? I estimate the mita’s impact

using a GIS road network map of Peru produced by the Ministry of Transportation

(2006). Roads are classified as paved, gravel, non-gravel, and trocha carrozable, which

translates roughly as “narrow path, often through wild vegetation . . . that a vehicle can

be driven on with great difficulty” (Real Academia Espanola, 2008). The total length

49Data from the 1981 Population Census, the first available source of education data after the 1940Population Census, likewise do not show a mita effect on years of schooling. The 2005 PopulationCensus was methodologically flawed, and district level results from the 2007 Population Census are notyet available. Hence, I have used the household survey data.

26

(in m) of the respective road type within each district is divided by the district surface

area (in km2) to obtain a road network density.

Several interesting findings emerge. First, column 1 of Table 8 suggests that the mita

does not impact local road networks, which consist primarily of non-gravel and trocha

roads. Care is required in interpreting this result, as the World Bank’s Rural Roads

program, operating since 1997, has worked to reduce disparities in local road networks

in marginalized areas of Peru. This result may reflect the successful targeting of the

World Bank program, as opposed to local institutional factors.

Second, there are significant disparities in regional road networks, which connect

population centers to each other. Column (2) of Panel A estimates that a mita effect

lowers the density of regional roads by 41 meters of roadway for every one kilometer

squared of district surface area (standard error = 10.2). In Panel B, the coefficient is

similar, at -52.5, and is significant at the 5% level. This large effect can be compared

to an average density of roads in mita districts of 20. The discontinuity is plainly

apparent in Figure 3, panel G. Column (3) breaks down the result by looking only at

the two highest quality types of roads: paved and gravel. The mita coefficient is large

and negative in both panels, though not statistically significant in Panel B. It is also

of interest to note that only 18 percent of district capitals in the mita sample can be

accessed by paved roads, as compared to 40 percent in the non-mita sample (INEI,

2004).

In summary, while I find little evidence that a mita effect persists through access

to schooling, there are pronounced disparities in road networks across the mita bound-

ary. Consistent with this evidence, I hypothesize that the long-term presence of large

landowners provided a stable land tenure system that encouraged public goods provi-

sion. As discussed above, the property rights of large landowners were secure, whereas

confiscation of peasant lands and numerous responding peasant rebellions were concen-

trated in mita districts (Bustamante Otero, 1987, p. 126-130; Flores Galindo, 1987,

p. 240; Ramos Zambrano, 1985, p. 29-34). Because established landowners in non-

mita districts controlled a larger percentage of the productive factors, both labor and

land, and because their property rights were more secure, it is probable that they re-

ceived higher returns to investing in public goods than did those living inside the mita

catchment. Moreover, historical evidence indicates that these landowners possessed the

political connections required to secure public goods. The established hacienda elite

lobbied successfully for roads to be constructed to pass through as many haciendas as

27

possible (Stein, 1980, p. 59).50 These roads remain, although haciendas were dissolved

by agrarian reform in the early 1970’s.51

4.3 Proximate Determinants of Household Consumption

This section examines the mita’s long run effects on the proximate determinants of

consumption. The limited available evidence does not suggest differences in investment,

so I focus on the labor force and market participation.52 Agriculture is an important

economic activity in the region I examine, providing primary employment for around 70%

of the population. Thus, I begin in column (1) of Table 9 by looking at the percentage

of the district labor force whose primary occupation is agriculture, taken from the 1993

Population Census. The coefficient on mitad is positive in both specifications and not

quite significant at the 10% level in Panel A, providing some weak evidence for a mita

effect on employment in agriculture. Further results (not shown) do not find a mita

effect on male and female labor force participation and hours worked.

In column (2), the dependent variable, taken from the 1994 Agricultural Census,

equals one if the agricultural household sells at least part of its produce in markets,

and equals zero otherwise. Panel A suggests that the mita’s long run impact reduces

participation in agricultural markets by around 23 percentage points (s.e.=0.032), a

large and highly significant effect. The result can be seen in Figure 3, Panel H. Only

ten percent of agricultural households in the mita sample participate in markets, as

compared to 34 percent in the non-mita sample.53

50The first modern road building campaigns occurred in the 1920’s and many of the region’s roadswere constructed in the 1950’s (Stein, 1980, Capunay, 1951, p. 197-199).

51The elasticity of equivalent consumption in 2001 with respect to haciendas per capita in 1689,in non-mita districts, is 0.036 (s.e. = 0.022). This association is consistent with greater security ofproperty rights in non-mita districts having a long-run positive effect on living standards.

52Data from the 1994 Agricultural Census on utilization of fifteen types of capital goods and twelvetypes of infrastructure for agricultural production do not show differences across the mita boundary, noris the length of fallowing different. I am not aware of data on private investment outside of agriculture.

53The 1994 Agricultural Census also asked households the most important factor they consideredwhen selecting which crops to plant at the beginning of the season. I repeated the same analysis usingas the dependent variables a dummy variable equal to one if the household answered that it primarilyconsidered expected prices or pre-secured sales, and equal to zero otherwise and a dummy equal toone if the household responded “we always plant the same crops.” Results (available upon request)document that households in non-mita districts are more likely to consider market incentives whenchoosing crops, whereas households in mita districts are more likely to always plant the same thing.This pattern is consistent with Field and Field (2007), who find that households with better access tolocal and regional markets are more likely to adopt new crops in response to favorable price incentives

28

Panel B, in contrast, estimates the mita’s impact on market participation to be small

and statistically insignificant. In light of this difference, it is helpful to note that the

mita coefficients for this outcome variable are large and significant in most of the twelve

robustness specifications reported in Appendix Table A2, except for the two that likewise

interact an RD polynomial in distance to the mita boundary with the mita dummy.

To provide intuition for why the specifications in Panels A and B produce different

results, Panel C estimates the mita effect by limiting the sample to observations very near

the boundary and using ordinary least squares.54 I define the set of districts very near the

boundary using several different criteria, and results suggest a negative, significant mita

effect on market participation. First, Panel C, column 1 includes only those districts

that border the boundary, estimating a mita coefficient equal to -0.178 (s.e. = 0.05).

Columns (2) through (5) take a slightly different approach to defining proximity, limiting

the sample to districts whose capitals fall within 25, 15, 10, and 5 km of the mita

boundary, respectively. This measures proximity in the same way as the distance to

the boundary running variable. When the sample is limited to districts whose capitals

fall within 25, 15 or 10 km, the mita coefficient is negative and statistically significant.

However, when comparing households living in districts whose capitals are within five

kilometers of the boundary (column (5)), disparities are no longer present. Specifically,

residents of the nine non-mita districts in this sample on average participate less in

markets than residents in the 34 non-mita districts that border the boundary.55 Overall,

I interpret the evidence in Panel C as consistent with the negative and statistically

significant estimates reported in Panel A, and these estimates are also similar to those

produced by ten of the twelve specifications considered in Appendix Table A2. The

more flexible specification in Panel B is sensitive to districts whose capitals are very

near the boundary, and market participation in these districts is unrepresentative of

that in districts bordering the boundary more generally.

A mita effect on market participation, strongly supported by Panel A and columns

(1) through (4) of Panel C, is consistent with the findings on road networks, particularly

given that recent studies on Andean Peru empirically connect poor road infrastructure to

resulting from globalization.54The specification is identical to those used in Panels A and B, except that the RD polynomial is

omitted.55This may reflect that fact that the terrain in non-mita districts whose capitals are very near the

boundary is on average more rugged (with a slope of 10.7 as compared to 9.5 in districts that borderthe boundary).

29

higher transaction costs, lower market participation, and reduced household income (Es-

cobal and Ponce, 2002; Escobal, 2001; Agreda and Escobal, 1998). Simple correlations

parallel empirical evidence linking roads to markets. 33% of agricultural households in

districts with paved road density above the median participate in markets, as compared

to 13% in districts with paved road density below the median.

There may also exist other channels through which a mita effect lowers market par-

ticipation. For example, data from the 1994 Agricultural Census reveal that the median

size of household landholdings is somewhat lower inside the mita catchment (at 1.2

hectares) than outside (at 1.4 hectares). To the extent that marketing agricultural pro-

duce involves fixed costs, a broader group of small farmers outside the mita catchment

may find it profitable. It is also possible that agricultural producers in mita districts

supplement their income by working as wage laborers rather than by producing for

markets. There is little support for this hypothesis in the data. In column (3), the

dependent variable is an indicator equal to one if a member of the agricultural house-

hold participates in secondary employment outside the agricultural unit, and equal to

zero otherwise. Estimates suggest that, if anything, the mita effect on participation in

secondary employment is negative, although it is not significant in either panel. The

1994 Agricultural Census also documents that a similar percentage of households across

the mita boundary held formal titles to their land.56

Could residents in mita districts have less desire to participate in the market economy,

rather than being constrained by poor road infrastructure? While Shining Path, a

Maoist guerilla movement, gained a strong foothold in the region during the 1980’s, this

hypothesis seems unlikely.57 Shining Path’s rise to power occurred against a backdrop

56During the 1994 Agricultural Census, which preceded a major rural land titling campaign, around17% of agricultural households both inside and outside the mita catchment held registered land titles.

57Many of the factors linked to the mita (poor infrastructure, limited access to markets, poorly definedproperty rights, and poverty) are also argued by prominent scholars and the Peruvian government tohave been leading factors promoting popular support for Shining Path (CVR, 2003, vol. 1, p. 94,McClintock, 1998; Palmer, 1994). Thus, I tested whether there was a mita effect on Shining Path(results available upon request). To measure the intensity of Shining Path, I exploited a loophole in thePeruvian constitution that stipulates that when more than two thirds of votes cast are blank or null,authorities cannot be renewed (Pareja and Gatti, 1990). In an attempt to sabotage the reelection ofauthorities during the 1989 municipal elections, Shining Path operatives encouraged citizens to expresssupport through casting blank or null (secret) ballots (McClintock, 1993, p. 79). I find that a mitaeffect increased blank/null votes by 10.7 percentage points (standard error = 0.031), suggesting greatersupport for (and intimidation by) Shining Path in mita districts. Moreover, estimates show that a mitaeffect increased the probability that authorities were not renewed by a highly significant 43.5 percentagepoints. As a specification check, I also look at blank/null votes in 2002, ten years after Shining Path’sdefeat, and there is no longer a statistically significant effect. Shining Path was crushed in the region

30

of limited support for Maoist ideology, and the movement’s attempts to reduce peasant

participation in markets were particularly unpopular and unsuccessful where attempted

(McClintock, 1998; Palmer 1994).

These results suggest that poor integration into the market economy partially ex-

plains why mita districts are poorer today. Yet, if most population and economic activity

endogenously clustered along roads, the relative poverty of mita districts would not be

that surprising. The data do not support this hypothesis: the correlation between 1940

district population density and the density of paved and gravel roads, measured in 2006,

is 0.58; when looking at this correlation using 1993 population density, it remains at

0.58. While many of Peru’s roads were built or paved in the interlude between 1940 and

1993, aggregate population responses appear minimal.

In sum, the analysis above suggests that public goods and market participation are

important proximate channels through which the mita’s effects persist. This does not

imply that they are the only channels. Others that may matter, as suggested by the

discussion of the stabilizing role of haciendas, include differences in access to justice, the

rule of law, and informal risk sharing arrangements.

Qualitative evidence from the region also underscores the importance of roads and

market access for producing household income. The citizens I spoke with, while visit-

ing eight primarily mita and six primarily non-mita provinces, were acutely aware that

some areas are more prosperous than others. When discussing the factors leading to the

observed income differences, a common theme was that it is difficult to transport crops

to markets. Thus, most residents in mita districts are engaged in subsistence farming.

Edgar Gonzales Castro, an agrarian scientist who works with indigenous communities

to improve farming techniques, argues: “It’s certainly not geography, it’s government

. . . Some provinces have been favored, with the government - particularly during the

large road building campaign in the early 1950’s - choosing to construct roads in some

provinces and completely ignore others” (Dec. 14th, 2006). At the forefront of the local

government’s mission in the large (primarily mita) province of Espinar is “to advocate

effectively for a system of modern roads to regional markets” (2008). Popular demands

have also centered on roads and markets. In 2004, (the mita district) Ilave made interna-

tional headlines when demonstrations involving over 10,000 protestors culminated with

the lynching of Ilave’s mayor, whom protestors accused of failing to deliver on promises

in 1992, and it is unlikely that the contemporary effects of the mita are short-term outcomes of ShiningPath, as these same factors are widely argued to have created support for Shining Path.

31

to pave the city’s access road and build a local market (Shifter, 2004).

5 Concluding Remarks

Numerous studies find a long run impact of history on comparative development, but few

offer empirical evidence on how the effects of institutions persist. This paper documents

and exploits plausible exogenous variation in the assignment of the mita in Peru to iden-

tify channels through which the mita influences contemporary economic development.

I estimate that the mita’s long run effects lower household consumption by around one

third and increase stunting in school children by around five percentage points. I then

document land tenure and public goods as channels through which the mita’s impacts

persist. For labor competition reasons, the mita led haciendas to form primarily outside

the mita catchment, and this effect persisted into the 20th century. Evidence suggests

that the long run presence of large landowners in non-mita districts provided a stable

land tenure system that encouraged public goods provision. Mita districts historically

achieved lower levels of education, today they are less integrated into road networks,

and their residents are substantially more likely to be subsistence farmers.

The region examined provides an important example in which the historical presence

of large landowners is associated with relatively better outcomes. While production and

politics in non-mita districts were not organized for the benefit of the masses, evidence

suggests that large landowners severed as a stabilizing force, ensuring public goods that

do benefit the general populace today. This paper’s results cast doubt on conventional

wisdom that emphasizes high land inequality as the underlying cause of Latin America’s

poor long run growth performance (Engerman and Sokoloff, 1997). A land tenure system

promoting relative equality between small, enfranchised landowners may well be useful

for sustaining high rates of economic growth. However, this is not the appropriate

counterfactual to large landowners in Peru, because the institutional structures in place

did not protect the rights of peasant smallholders.

Much work remains in acquiring a general understanding of how institutions persist

and how institutional change can be promoted. The development of general models of

institutional persistence and empirical investigation of how historical institutions interact

with contemporary forces promoting change are particularly central areas for future

research.

32

References

Acemoglu, D., M. A. Bautista, P. Qeurubin, and J. A. Robinson (2008): “Economic and polıticalInequality in Development: The Case of Cundinamarca, Colombia,” NBER Working Paper 13208.

Acemoglu, D., S. Johnson, and J. A. Robinson (2001): “The Colonial Origins of Comparative Develop-ment: An Empirical Investigation,” The American Economic Review, 91, 1369–1401.

——— (2002): “Reversal of Fortune: Geography and Institutions in the Making of the Modern World IncomeDistribution,” Quarterly Journal of Economics, 117, 1231–1294.

Agreda, V. and J. Escobal (1998): “Analasıs de la comercializacion agrıcola en el Peru,” Boletın de opinion,33.

Alvarez, E. and J. M. Caballero (1980): Aspectos cuantitativos de la reforma agraria, 1969-1979, Lima:Instituto de Estudios Peruanos.

Amat y Junient, M. (1947): Memoria de Gobierno, Sevilla.

Angrist, J. and V. Lavy (1999): “Using Maimonides’ Rule to Estimate The Effect of Class Size on ScholasticAchievement*,” Quarterly Journal of Economics, 114, 533–575.

Angrist, J. and J. Pischke (2009): Mostly harmless econometrics: an empiricist’s companion, PrincetonUniv Pr.

Arzans de Orusa y Vela, B. (1965): Historia de la Villa Imperial de Potosı Edicion de Lewis Hanke yGunnar Mendoza, Providence: Brown University Press.

Avila, J. (1952): “Exposicion e causas que justifican la necesidad de la reforma agraria en el distrito deAzangaro de la provincia del mismo nombre, Puno,” Ph.D. thesis, Universidad Nacional del Cusco.

Bakewell, P. (1984): Miners of the Red Mountain. Indian Labor in Potosi, 1545-1650, Albuquerque: Uni-versity of New Mexico Press.

Banerjee, A. and L. Iyer (2005): “History, Institutions, and Economic Performance: The Legacy of ColonialLand Tenure Systems in India,” American Economic Review, 95, 1190–1213.

Bauer, B. and A. Covey (2002): “Processes of State Formation in the Inca Heartland (Cusco, Peru),”American Anthropologist, 104, 846–864.

Brisseau, J. (1981): Le Cuzco dans sa region: etude de l’aire d’influence d’une ville andine, Lima: Institutfrancais d’etudes andines.

Bustamante Otero, L. H. (1987): “Mita y realidad: Teodomerio Gutierrez Cuevas o Rumi Maqui en elmarco de la sublevacion campesina de Azangaro: 1915-1916,” Ph.D. thesis, Pontificia Universidad Catolicadel Peru.

Canete, P. V. (1794): El codigo carolino de ordenanzas reales de las minas de Potos y demas provincias delRıo de la Plata, Buenos Aires, 1973.

Capunay, M. (1951): Leguia, vida y obra del constructor del gran Peru, Lima.

Center for International Earth Science Information Network (2004): Global Rural-Urban MappingProject, Alpha Version: Population Grids, Socioeconomic Data and Applications Center (SEDAC), ColumbiaUniversity, http://sedac.ciesin.columbia.edu/gpw (May 10th, 2007).

Chay, K., P. McEwan, and M. Urquiola (2005): “The central role of noise in evaluating interventionsthat use test scores to rank schools,” American Economic Review, 95, 1237–1258.

33

Cieza de Leon, P. (1551): El Senorio de los Incas; 2a. Parte de la Cronica del Peru, Lima: Instituto deEstudios Peruanos, 1967.

——— (1959): The Incas, Norman: University of Oklahoma Press.

Coatsworth, J. (2005): “Structures, Endowments, and Institutions in the Economic History of Latin Amer-ica,” Latin American Research Review, 40, 126–144.

Cole, J. (1985): The Potosı Mita 1573-1700. Compulsory Indian Labor in the Andes, Stanford: StanfordUniversity Press.

Comision de la Verdad y Reconciliacion (2003): Informe Final, Lima.

Conley, T. (1999): “GMM Estimation with Cross Sectional Dependence,” Journal of Econometrics, 92, 1–45.

Cook, N. D. (1975): Tasa de la visita general de Francisco Toledo, Lima: Universidad Nacional de SanMarcos.

——— (1981): Demographic Collapse: Colonial Peru 1520-1620, Cambridge: Cambridge University Press.

D’Altoy, T. (2002): The Incas, Oxford: Blackwell Publishers.

Dancuart, E. and J. Rodriguez (1902): Anales de la hacienda publica del Peru. 24 vols, Lima: Imprentade La Revista y Imprenta de Guillermo Stolte.

Deaton, A. (1997): The Analysis of Household Surveys: A Microeconomic Approach to Development Policy,Baltimore: Johns Hopkins University Press.

Deaton, A. and J. Dreze (2008): “Nutrition in India: Facts and Interpretations,” .

Direccion de estadıstica del Peru (1878): Censo general de la republica del Peru, formatdo en 1876,Lima: Imp. del Teatro.

——— (1944): Censo Nacional de Poblacion y Ocupacion, 1940, Lima.

Easterly, W. and R. Levine (2003): “Tropics, germs, and crops: How endowments influence economicdevelopment,” Journal of Monetary Economics, 50, 3–39.

Engerman, S. and K. Sokoloff (1997): “Factors Endowments, Institutions, and Differential Paths ofGrowth Among New World Economies,” in How Latin American Fell Behind, ed. by S. Haber, Stanford:Stanford University Press, 260–304.

Escobal, J. (2001): “The Benefits of Roads in Rural Peru: A Transaction Costs Approach,” Grupo de Analisispara el Desarrollo (Lima) Working Paper.

Escobal, J. and C. Ponce (2002): “‘The Benefits of Rural Roads: Enhancing Income Opportunities for theRural Poor,” Grupo de Analisis para el Desarrollo (Lima) Working Paper 40.

Espinar Municipal Government (2008): Vision, www.muniespinar.gob.pe/vision.php.

Favre, H. (1967): “Evolucion y situacion de las haciendas en la region de Huancavelica, Peru,” in La Haciendaen el Peru, ed. by H. Favre, C. Collin-Delavaud, and J. M. Mar, Lima: Instituto de Estudios Peruanos, 237–257.

Field, A. and E. Field (2007): “Globalization, Crop Choice, and Property Rights in Rural Peru, 1994–2004,”WIDER Research Paper, 72.

Flores Galindo, A. (1987): Buscando un inca: identidad y utopia en los Andes, Lima: Instituto de ApoyoAgrario.

34

Garrett, D. (2005): Shadows of Empire: The Indian Nobility of Cusco, 1750-1825, Cambridge: CambridgeUniversity Press.

Glaeser, E., R. LaPorta, F. L. de Silanes, and A. Shleifer (2004): “Do Institutions Cause Growth?”Journal of Economic Growth, 9, 271–303.

Glaeser, E. and A. Shleifer (2002): “Legal Origins,” Quarterly Journal of Economics, 117, 1193–1230.

Glave, L. M. (1989): Trajinantes: Caminos indıgenas de la sociedad colonial, Lima: Instituto de ApoyoAgraria.

Glave, L. M. and M. I. Remy (1978): Estructura Agraria e historia rural cuzquena: a proposito de lashaciendas del Valle Sagrado: 1780-1930, Cuzco, Archivo Historico del Cuzco.

Golte, J. (1980): La racionalidad de la organizacion andina, Lima: Insituto de Estudios Peruanos.

Gonzales Castro, E. (2006): Personal Interview. 14 December.

Hall, R. and C. Jones (1999): “Why do some countries produce so much more output per worker thanothers?” Quarterly Journal of Economics, 114, 83–116.

Hazen, D. (1974): “The Awakening of Puno: Government Policy and the Indian Problem in Southern Peru,1900-1955,” Ph.D. thesis, New Haven: Yale University.

Hijmans, R. et al. (2005): “Very high resolution interpolated climate surfaces for global land area,” Inter-national Journal of Climatology, 25, 1965–1978.

Hyslop, J. (1984): The Inka Road System, New York: Academic Press.

Instituto Nacional de estadıstica e Informacion de Peru (1981): VIII Censo de Poblacion.

——— (1993): IX Censo de Poblacion.

——— (1994): Tercer Censo Nacional Agropecuario.

——— (2001): Encuesta Nacional de Hogares (ENAHO).

——— (2004): Registro Nacional de Municipalidades (REMANU).

Instituto Nacional de Recursos Naturales (1997): Potencial Bioclimatica, Lima.

ISRIC World Soil Information (2007): SOTER for Latin America and Caribbean, 2.

Jacobsen, N. (1993): Mirages of Transition: The Peruvian Altiplano, 1780-1930, Berkeley: University ofCalifornia Press.

Keith, R. G. (1971): “Encomienda, Hacienda and Corregimiento in Spanish America: A Structural Analysis,”Hispanic American Historical Review, 51, 431–446.

Larson, B. (1982): Explotacion Agraria y Resistencia Campesina en Cochabamba, La Paz: Centro de Estudiosde la Realidad Economico y Social.

——— (1988): Colonialism and agrarian transformation in Bolivia: Cochabamba, 1550-1900, Princeton:Princeton University Press.

Lee, D., E. Moretti, and M. Butler (2004): “Do Voters Affect Or Elect Policies? Evidence From TheUS House*,” Quarterly Journal of Economics, 119, 807–859.

Levillier, R. (1921): Gobernantes del Peru, cartas y papeles, siglo XVI: documentos del Archivo de Indias,Madrid: Sucesores de Rivadeneyra.

35

Mar, M. and M. Mejia (1980): La reforma agraria en el Peru, Lima: Instituto de Estudios Peruanos.

McClintock, C. (1998): Revolutionary movements in Latin America: El Salvador’s FMLN and Peru’sShining Path, Washington, D.C.: United States Institute of Peace Press.

Ministro de Eduacion del Peru (2005a): Censo de Talla, Lima.

——— (2005b): Indicadores de cobertura y culminacion de la educacion basica, Lima. Available atwww.minedu.gob.pe.

——— (2006): Censo Escolar, Lima. Available at www.minedu.gob.pe.

Miranda, C. (1583): Tasa de la visita general de Francisco Toledo.

Morner, M. (1978): Perfil de la sociedad rural del Cuzco a fines de la colonia, Lima: Universidad del Pacifico.

National Aeronautics and Space Administration and the National Geospatial-IntelligenceAgency (2000): Shuttle Radar Topography Mission 30 Arc Second Finished Data.

Niles, S. (1987): Callachaca: Style and Status in an Inca Community, Iowa City: University of Iowa Press.

Nunez, J. T. (1913): Memoria leida de la ceremonıa de apertura del ano judicial de 1913 por el Presidente dela Ilustrisima Corte Superior de los departamentos de Puno y Madre de Dios, Dr. J. Teofilo Nunez, Puno:Imprenta del Seminario.

Nunn, N. (2008): “The Long-Term Effects of Africa’s Slave Trades,” Quarterly Journal of Economics, 123,139–176.

Nunn, N. and D. Puga (2007): “Ruggedness: The Blessing of Bad Geography in Africa,”http://diegopuga.org/papers/rugged.pdf.

Palmer, D. S. (1994): The Shining Path of Peru, New York: St. Martin’s Press.

Pareja Pflucker, P. and A. G. Murriel (1990): Evaluacion de las elecciones municipales de 1989:impacto politico de la violencia terrorista, Lima: Instituto Nacional de Planificacion.

Peralta Ruız, V. (1991): En pos del tributo: Burocracia estatal, elite regional y comunidades indıgenas enel Cusco rural (1826-1854), Cusco: Casa Bartolome de las Casas.

Peterson, T. C. and R. S. Vose (1997): “An overview of the Global Historical Climatology Networktemperature data base,” Bulletin of the American Meteorological Society, 78, 2837–2849.

Portocarrero, F., A. Beltran, and A. Zimmerman (1988): Inversiones publicas en el Peru (1900-1968).Una aproximacion cuantitativa, Lima: CIUP.

Portocarrero, G. and P. Oliart (1989): El Peru desde la escuela, Lima: Instituto de Apoyo Agrario.

Pulgar Vidal, J. (1950): Geografıa del Peru: Las ocho regiones naturales del Peru, Lima: Editorial Universo.

Ramos Zimbrano, A. (1984): La Rebelion de Huancane: 1923-1924, Puno: Editorial Samuel FrisanchoPineda.

Real Academia Espanola (2006): Diccionario de la lengua Espanola, www.rae.es.

Roca Sanchez, P. E. (1935): Por la clase indıgena, Lima: Pedro Barrantes Castro.

Rowe, J. (1946): “Inca Culture at the Time of the Spanish Conquest,” in Handbook of South AmericanIndians, ed. by J. Steward, Washington D.C.: Bureau of American Ethnology, vol. 2, 183–330.

36

Saavedra, J. and P. Suarez (2002): El financiamiento de la educacion publica en el Peru: El rol de lasfamilias, GRADE.

Sachs, J. (2001): “Tropical Underdevelopment,” National Bureau of Economic Research Working Paper 8119.

Saignes, T. (1984): “Las etnias de Charcas frente al sistema colonial (Siglo XVII): Ausentismo y fugas en eldebate sobre la mano de obra indıgena, 1595-1665,” Jahrbuchfr Geschichte (JAS), 21, 27–75.

Sala i Vila, N. (1991): “Revueltas indıgenas en el Peru tardocolonial,” Ph.D. thesis, Barcelona: Universitatde Barcelona.

Sanchez-Albornoz, N. (1978): Indios y tributos en el Alto Peru, Lima: Instituo de Estudios Peruanos.

Shifter, M. (2004): “Breakdown in the Andes,” Foreign Affairs, September/October.

Spalding, K. (1974): De indio a campesino: Cambios en la estructura social del Peru colonial, Lima: Instituode Estudios Peruanos.

Stein, S. (1980): Populism in Peru: The emergence of the masses and the politics of social control, Madison:University of Wisconsin Press.

Tamayo Herrera, J. (1982): Historia social e indigenismo en el Altiplano, Lima: Ediciones Treintaitres.

Tandeter, E. (1993): Coercion and market: silver mining in colonial Potosı,1692-1826, Albuquerque: Uni-versity of New Mexico Press.

Van der Klaauw, W. (2002): “Estimating the effect of financial aid offers on college enrollment: A regression-discontinuity approach,” International Economic Review, 1249–1287.

Victora, C., L. Adair, C. Fall, P. Hallal, R. Martorell, L. Richter, and H. Sachdev (2008):“Maternal and child undernutrition: consequences for adult health and human capital,” The Lancet.

Villanueva Urteaga, H. (1982): Cuzco 1689: Informes de los parrocos al obispo Mollinedo, Cusco: CasaBartolome de las Casas.

Wightman, A. (1990): Indigenous Migration and Social Change: The Forasteros of Cuzco, 1570-1720,Durham: Duke University Press.

Zavala, S. (1980): El servicio personal de los indios en el Peru, Mexico City: El Colegio de Mexico.

37

Table

1:

Sum

mary

Sta

tist

ics

Sam

ple

falls

withi

n:

<10

0km

ofm

ita

boun

dary

<75

kmof

mita

boun

dary

<50

kmof

mita

boun

dary

<25

kmof

mita

boun

dary

Insi

deO

utsi

deSE

Insi

deO

utsi

deSE

Insi

deO

utsi

deSE

Insi

deO

utsi

deSE

GIS

Mea

sure

sE

leva

tion

4076

3948

[197

.95]

4128

3984

[174

.90]

4161

4045

[169

.94]

4115

4032

[142

.01]

(89.

25)

(87.

33)

(91.

02)

(115

.27)

Slop

e5.

807.

66[0

.95]

*6.

007.

72[0

.92]

*6.

097.

36[0

.95]

6.42

7.39

[0.9

8](0

.54)

***

(0.5

7)**

*(0

.62)

**(0

.86)

Obs

erva

tion

s18

386

146

8210

070

4847

%In

dige

nous

63.5

958

.84

[11.

21]

71.0

064

.55

[8.0

9]71

.01

64.5

4[8

.47]

74.4

763

.35

[10.

93]

(9.7

6)(8

.14)

(8.4

3)(1

0.52

)O

bser

vati

ons

1112

366

831

330

683

330

329

251

Log

1572

trib

ute

1.57

1.60

[0.0

4]1.

571.

60[0

.04]

1.58

1.61

[0.0

5]1.

651.

61[0

.02]

*ra

te(0

.03)

(0.0

3)(0

.04)

(0.0

3)

%15

72tr

ibut

eto

:Sp

anis

hN

obili

ty59

.80

63.8

2[1

.39]

***

59.9

863

.69

[1.5

6]**

62.0

163

.07

[1.1

2]61

.01

63.1

7[1

.58]

(1.3

6)**

*(1

.53)

**(1

.34)

(2.2

1)Sp

anis

hPri

ests

21.0

519

.10

[0.9

0]**

21.9

019

.45

[1.0

2]**

20.5

919

.93

[0.7

6]21

.45

19.9

8[1

.01]

(0.9

4)**

(1.0

2)**

(0.9

2)(1

.33)

Span

ish

Just

ices

13.3

612

.58

[0.5

3]13

.31

12.4

6[0

.65]

12.8

112

.48

[0.4

3]13

.06

12.3

7[0

.56]

(0.4

8)*

(0.6

0)(0

.55)

(0.7

9)In

dige

nous

May

ors

5.67

4.40

[0.7

8]4.

554.

29[0

.26]

4.42

4.47

[0.3

4]4.

484.

42[0

.29]

(0.8

5)(0

.29)

(0.3

3)(0

.39)

Obs

erva

tion

s63

4147

3735

3018

24T

heun

itof

obse

rvat

ion

istw

enty

kilo

met

erby

twen

tyki

lom

eter

grid

cells

for

the

geos

pati

alm

easu

res,

the

hous

ehol

dfo

r%

indi

geno

us,an

dth

edi

stri

ctfo

rth

e15

72tr

ibut

eda

ta.

Con

ley

stan

dard

erro

rsfo

rth

edi

ffere

nce

inm

eans

betw

een

mita

and

non-

mita

obse

rvat

ions

are

inbr

acke

ts.

Rob

ust

stan

dard

erro

rsfo

rth

edi

ffere

nce

inm

eans

are

inpa

rent

hese

s.Fo

r%

indi

geno

us,th

ero

bust

stan

dard

erro

rsar

eco

rrec

ted

for

clus

teri

ngat

the

dist

rict

leve

l.T

hege

ospa

tial

mea

sure

sar

eca

lcul

ated

usin

gel

evat

ion

data

at30

arc

seco

nd(o

neki

lom

eter

)re

solu

tion

(SRT

M,20

00).

The

unit

ofm

easu

refo

rel

evat

ion

is10

00m

eter

san

dfo

rsl

ope

isde

gree

s.A

hous

ehol

dis

indi

geno

usif

its

mem

bers

prim

arily

spea

kan

indi

geno

usla

ngua

gein

the

hom

e(E

NA

HO

,20

01).

The

trib

ute

data

are

take

nfr

omM

iran

da,19

75[1

583]

.In

the

first

thre

eco

lum

ns,th

esa

mpl

ein

clud

eson

lyob

serv

atio

nslo

cate

dle

ssth

an10

0ki

lom

eter

sfr

omth

em

ita

boun

dary

,an

dth

isth

resh

old

isre

duce

dto

75,50

,an

dfin

ally

25ki

lom

eter

sin

the

succ

eedi

ngco

lum

ns.

Coe

ffici

ents

that

are

sign

ifica

ntly

diffe

rent

from

zero

are

deno

ted

byth

efo

llow

ing

syst

em:

*10%

,**

5%,an

d**

*1%

.

Table

2:

Liv

ing

Sta

ndard

s

Dep

ende

ntva

riab

leis

:Lo

geq

uiva

lent

hous

ehol

dco

nsum

ptio

n(2

001)

Stun

ted

grow

th,ch

ildre

n6-

9(2

005)

<10

0km

<75

km<

50km

<25

km<

100

km<

75km

<50

km<

25km

bord

erSa

mpl

eW

ithi

n:of

boun

d.of

boun

d.of

boun

d.of

boun

d.of

boun

d.of

boun

d.of

boun

d.of

boun

d.di

stri

ct(1

)(2

)(3

)(4

)(5

)(6

)(7

)(8

)(9

)

A:Q

uadr

atic

Pol

ynom

ialin

Dis

tanc

eto

Pot

osı

Mit

a-0

.346

***

-0.2

84**

-0.3

73**

*-0

.240

0.06

8***

0.06

3***

0.07

6***

0.06

2***

0.05

1**

(0.1

04)

(0.1

07)

(0.1

28)

(0.1

72)

(0.0

24)

(0.0

24)

(0.0

25)

(0.0

22)

(0.0

24)

Ele

vati

on-0

.247

**-0

.184

-0.1

63-0

.450

0.04

2*0.

050

0.07

1*0.

053

0.26

4***

(0.1

19)

(0.1

79)

(0.1

74)

(0.2

73)

(0.0

25)

(0.0

31)

(0.0

40)

(0.0

53)

(0.0

72)

Slop

e-0

.019

-0.0

11-0

.003

-0.0

33-0

.005

-0.0

05-0

.004

-0.0

060.

006

(0.0

15)

(0.0

15)

(0.0

15)

(0.0

23)

(0.0

03)

(0.0

04)

(0.0

04)

(0.0

05)

(0.0

07)

R2

0.05

30.

045

0.05

40.

051

0.05

00.

018

0.01

40.

029

0.05

2

B:In

tera

cted

Qua

drat

icPol

ynom

ialin

Dis

tanc

eto

Mita

Bou

ndar

yM

ita

-0.3

94-0

.531

-0.3

38-1

.327

**0.

137*

*0.

126*

0.13

6*0.

128*

0.10

2**

(0.2

72)

(0.3

22)

(0.3

52)

(0.4

82)

(0.0

70)

(0.0

72)

(0.0

72)

(0.0

65)

(0.0

51)

Ele

vati

on-0

.069

-0.0

93-0

.125

-0.4

740.

067*

0.04

7*0.

058

0.09

0**

0.24

4***

(0.1

72)

(0.1

55)

(0.1

79)

(0.3

21)

(0.0

34)

(0.0

28)

(0.0

36)

(0.0

45)

(0.0

66)

Slop

e-0

.011

-0.0

08-0

.009

-0.0

240.

002

-0.0

02-0

.001

-0.0

040.

006

(0.0

19)

(0.0

15)

(0.0

18)

(0.0

26)

(0.0

04)

(0.0

04)

(0.0

04)

(0.0

05)

(0.0

07)

R2

0.05

30.

053

0.05

10.

060

0.04

10.

016

0.01

40.

029

0.05

0

Geo

.C

ontr

ols

yes

yes

yes

yes

yes

yes

yes

yes

yes

Bou

ndar

yF.E

.sye

sye

sye

sye

sye

sye

sye

sye

sye

sC

lust

ers

7160

5227

289

239

185

9363

Obs

erva

tion

s1,

478

1,16

11,

013

580

158,

848

115,

761

100,

446

53,6

9337

,421

The

unit

ofob

serv

atio

nis

the

hous

ehol

din

colu

mns

(1)

thro

ugh

(4)

and

the

indi

vidu

alin

colu

mns

(5)

thro

ugh

(9).

Rob

ust

stan

dard

erro

rs,ad

just

edfo

rcl

uste

ring

bydi

stri

ct,ar

ein

pare

nthe

ses.

The

depe

nden

tva

riab

leis

log

equi

vale

ntho

useh

old

cons

umpt

ion

(EN

AH

O,20

01)

inco

lum

ns(1

)th

roug

h(4

),an

da

dum

my

equa

lto

1if

the

child

has

stun

ted

grow

than

deq

ualto

0ot

herw

ise

inco

lum

ns(5

)th

roug

h(9

)(M

inis

try

ofE

duca

tion

,20

05).

Mit

ais

anin

dica

tor

equa

lto

one

ifth

eho

useh

old’

sdi

stri

ctco

ntri

bute

dto

the

mita

and

equa

lto

zero

othe

rwis

e(S

aign

es,19

84;A

mat

yJu

niet

,19

47,p.

249,

284)

.E

leva

tion

and

slop

ear

eth

em

ean

area

wei

ghte

del

evat

ion

and

slop

eof

the

hous

ehol

d’s

dist

rict

.T

heun

itof

mea

sure

for

elev

atio

nis

1000

met

ers

and

for

slop

eis

degr

ees.

Pan

elA

incl

udes

aqu

adra

tic

poly

nom

iali

nE

uclid

ean

dist

ance

from

the

obse

rvat

ion’

sdi

stri

ctca

pita

lto

Pot

osı,

and

Pan

elB

incl

udes

aqu

adra

tic

poly

nom

iali

nE

uclid

ean

dist

ance

toth

ene

ares

tpo

int

onth

em

ita

boun

dary

and

inte

ract

sth

ispo

lyno

mia

lw

ith

the

mita

dum

my.

All

regr

essi

ons

incl

ude

soil

type

indi

cato

rsan

dbo

unda

ryse

gmen

tfix

edeff

ects

.C

olum

ns(1

)th

roug

h(4

)in

clud

ede

mog

raph

icco

ntro

lsfo

rth

enu

mbe

rof

infa

nts,

child

ren,

and

adul

tsin

the

hous

ehol

d.In

the

first

and

fifth

colu

mns

,th

esa

mpl

ein

clud

esob

serv

atio

nsw

hose

dist

rict

capi

tals

are

loca

ted

wit

hin

100

kilo

met

ers

ofth

em

ita

boun

dary

,an

dth

isth

resh

old

isre

duce

dto

75,50

,an

dfin

ally

25ki

lom

eter

sin

the

succ

eedi

ngco

lum

ns.

Col

umn

(9)

incl

udes

only

obse

rvat

ions

who

sedi

stri

cts

bord

erth

em

ita

boun

dary

.78

%of

the

obse

rvat

ions

are

inm

ita

dist

rict

sin

colu

mn

(1),

71%

inco

lum

n(2

),68

%in

colu

mn

(3),

60%

inco

lum

n(4

),78

%in

colu

mn

(5),

71%

inco

lum

n(6

),68

%in

colu

mn

(7),

60%

inco

lum

n(8

),an

d58

%in

colu

mn

(9).

Coe

ffici

ents

that

are

sign

ifica

ntly

diffe

rent

from

zero

are

deno

ted

byth

efo

llow

ing

syst

em:

*10%

,**

5%,

and

***1

%.

Table 3: Specification Tests

Dependent variable is:Log equivalent household consumption (2001) Stunted growth, children 6-9 (2005)

<100 km <75 km <50 km <25 km <100 km <75 km <50 km <25 km borderof bound. of bound. of bound. of bound. of bound. of bound. of bound. of bound. district

(1) (2) (3) (4) (5) (6) (7) (8) (9)

Alternative functional forms for RD polynomial: Baseline I

Linear polynomial in distance to Potosı

Mita -0.327*** -0.305*** -0.266** -0.250* 0.053** 0.049** 0.054** 0.054** 0.059**(0.102) (0.103) (0.102) (0.144) (0.024) (0.023) (0.024) (0.021) (0.026)

Cubic polynomial in distance to Potosı

Mita -0.332*** -0.373*** -0.423*** -0.359* 0.079*** 0.080*** 0.084*** 0.059** 0.053**(0.124) (0.130) (0.129) (0.196) (0.025) (0.025) (0.027) (0.027) (0.024)

Quartic polynomial in distance to Potosı

Mita -0.385*** -0.369*** -0.429*** -0.256 0.078*** 0.079*** 0.083*** 0.058** 0.053**(0.119) (0.126) (0.132) (0.176) (0.025) (0.025) (0.027) (0.027) (0.024)

RD polynomial (Baseline I) interacted with Mita

Interacted linear polynomial in distance to Potosı

Mita -0.273*** -0.244** -0.231** -0.202 0.048** 0.043** 0.039* 0.046** 0.061**(0.102) (0.110) (0.105) (0.156) (0.022) (0.021) (0.022) (0.019) (0.025)

Interacted quadratic polynomial in distance to Potosı

Mita -0.298*** -0.215* -0.394** -0.131 0.022 0.017 0.032 0.030 0.033(0.108) (0.111) (0.169) (0.241) (0.024) (0.023) (0.021) (0.023) (0.022)

Alternative functional form for RD polynomial: Baseline II

Interacted linear polynomial in distance to mita boundary

Mita -0.287 -0.297 -0.373 -0.338 0.088* 0.094* 0.093 0.076 0.098**(0.174) (0.196) (0.224) (0.314) (0.045) (0.048) (0.057) (0.047) (0.040)

RD polynomial (Baseline II) not interacted with Mita

Linear polynomial in distance to mita boundary

Mita -0.333*** -0.254** -0.249** -0.277* 0.068*** 0.055** 0.057** 0.052** 0.043*(0.109) (0.102) (0.106) (0.141) (0.025) (0.023) (0.024) (0.022) (0.025)

Quadratic polynomial in distance to mita boundary

Mita -0.298*** -0.255** -0.251** -0.285* 0.068*** 0.055** 0.060** 0.052** 0.044*(0.104) (0.103) (0.105) (0.139) (0.024) (0.024) (0.024) (0.022) (0.025)

Cubic polynomial in distance to mita boundary

Mita -0.303*** -0.261** -0.249** -0.238* 0.070*** 0.056** 0.064*** 0.055** 0.042*(0.105) (0.102) (0.102) (0.135) (0.024) (0.023) (0.024) (0.022) (0.025)

RD polynomials included from both baseline specifications

Mita -0.382 -0.545* -0.365 -1.354** 0.108* 0.111* 0.125* 0.109* 0.084(0.267) (0.326) (0.328) (0.550) (0.062) (0.065) (0.068) (0.065) (0.052)

RD polynomials in latitude, longitude, and interactions

Mita -0.348** -0.399** -0.388** -0.284* 0.092*** 0.094*** 0.095** 0.065** 0.056**(0.166) (0.157) (0.158) (0.166) (0.034) (0.034) (0.037) (0.030) (0.026)

Ordinary Least Squares

Mita -0.319*** -0.310*** -0.259** -0.278* 0.057** 0.044* 0.048* 0.056** 0.041(0.105) (0.101) (0.103) (0.146) (0.027) (0.026) (0.027) (0.021) (0.026)

Geo. Controls yes yes yes yes yes yes yes yes yesBoundary F.E.s yes yes yes yes yes yes yes yes yesClusters 71 60 52 27 289 239 185 93 63Observations 1,478 1,161 1,013 580 158,848 115,761 100,446 53,693 37,421

Robust standard errors, adjusted for clustering by district, are in parentheses. All regressions include soil type indicatorsand boundary segment fixed effects. Columns (1) through (4) include demographic controls for the number of infants,children, and adults in the household. Coefficients significantly different from zero are denoted by: *10%, **5%, and ***1%.

Table

4:

Addit

ionalSpeci

fica

tion

Test

s

Log

equi

vale

ntho

useh

old

cons

umpt

ion

(200

1)St

unte

dgr

owth

,ch

ildre

n6-

9(2

005)

Exc

lude

sE

xclu

des

Exc

lude

spo

rtio

nsof

Fle

xibl

eE

xclu

des

port

ions

ofdi

stri

cts

boun

dary

esti

mat

ion

dist

rict

sbo

unda

ryC

ontr

olfo

rIn

clud

esw

ith

Inca

form

edby

ofco

nsum

p.In

clud

esw

ith

Inca

form

edby

Bas

elin

eet

hnic

ity

Cus

coes

tate

sri

vers

equi

vale

nce

Mig

rati

onB

asel

ine

Cus

coes

tate

sri

vers

Mig

rati

on(1

)(2

)(3

)(4

)(5

)(6

)(7

)(8

)(9

)(1

0)(1

1)(1

2)

A:Q

uadr

atic

Pol

ynom

ialin

Dis

tanc

eto

Pot

osı

Mit

a-0

.373

***

-0.2

70**

-0.4

91**

*-0

.369

***

-0.3

72**

*-0

.377

***

-0.2

95**

0.07

6***

0.15

5***

0.07

4***

0.08

6***

0.05

8**

(0.1

28)

(0.1

03)

(0.1

27)

(0.1

15)

(0.1

28)

(0.1

33)

(0.1

27)

(0.0

25)

(0.0

31)

(0.0

26)

(0.0

25)

(0.0

25)

Ele

vati

on-0

.163

-0.2

48*

-0.4

56**

*-0

.067

-0.1

63-0

.154

-0.1

480.

071*

0.16

7***

0.09

7**

0.05

80.

081*

*(0

.174

)(0

.140

)(0

.140

)(0

.229

)(0

.174

)(0

.178

)(0

.168

)(0

.040

)(0

.036

)(0

.042

)(0

.038

)(0

.039

)Sl

ope

-0.0

03-0

.014

-0.0

40**

*0.

005

-0.0

03-0

.004

-0.0

12-0

.004

0.01

6***

-0.0

02-0

.004

-0.0

04(0

.015

)(0

.013

)(0

.014

)(0

.020

)(0

.015

)(0

.016

)(0

.015

)(0

.004

)(0

.004

)(0

.004

)(0

.004

)(0

.004

)R

20.

054

0.14

30.

090

0.05

50.

053

0.28

00.

047

0.01

40.

039

0.01

60.

016

0.01

4

B:In

tera

cted

Qua

drat

icPol

ynom

ialin

Dis

tanc

eto

Mita

Bou

ndar

yM

ita

-0.3

38-0

.329

-0.3

03-1

.142

**-0

.338

-0.3

47-0

.288

0.13

6*0.

175*

*0.

162*

*0.

126*

0.11

8(0

.352

)(0

.250

)(0

.385

)(0

.461

)(0

.352

)(0

.367

)(0

.338

)(0

.072

)(0

.073

)(0

.075

)(0

.071

)(0

.073

)E

leva

tion

-0.1

25-0

.188

-0.4

03**

*-0

.047

-0.1

25-0

.120

-0.0

990.

058

0.17

2***

0.07

4**

0.04

70.

066*

(0.1

79)

(0.1

39)

(0.1

42)

(0.2

02)

(0.1

79)

(0.1

82)

(0.1

76)

(0.0

36)

(0.0

32)

(0.0

36)

(0.0

36)

(0.0

35)

Slop

e-0

.009

-0.0

16-0

.044

***

-0.0

14-0

.009

-0.0

10-0

.018

-0.0

010.

015*

**0.

002

-0.0

02-0

.000

(0.0

18)

(0.0

14)

(0.0

14)

(0.0

24)

(0.0

18)

(0.0

18)

(0.0

17)

(0.0

04)

(0.0

04)

(0.0

04)

(0.0

04)

(0.0

04)

R2

0.05

10.

146

0.09

20.

055

0.05

10.

278

0.04

80.

014

0.04

00.

015

0.01

40.

013

Geo

.C

ontr

ols

yes

yes

yes

yes

yes

yes

yes

yes

yes

yes

yes

yes

Bou

nd.

F.E

.sye

sye

sye

sye

sye

sye

sye

sye

sye

sye

sye

sye

sC

lust

ers

5252

5747

5152

5218

519

518

018

318

5O

bs.

1013

1013

1173

930

992

1013

997

100,

446

127,

259

96,4

4099

,940

98,9

22R

obus

tst

anda

rder

rors

,ad

just

edfo

rcl

uste

ring

bydi

stri

ct,ar

ein

pare

nthe

ses.

All

regr

essi

ons

incl

ude

soil

type

indi

cato

rsan

dbo

unda

ryse

gmen

tfix

edeff

ects

.C

olum

ns(1

)th

roug

h(5

)an

d(7

)in

clud

ede

mog

raph

icco

ntro

lsfo

rth

enu

mbe

rof

infa

nts,

child

ren,

and

adul

tsin

the

hous

ehol

d.C

olum

n(6

)in

clud

esco

ntro

lsfo

rth

elo

gof

hous

ehol

dsi

zean

dth

era

tio

ofch

ildre

nto

hous

ehol

dm

embe

rs,us

ing

the

log

ofho

useh

old

cons

umpt

ion

asth

ede

pend

ent

vari

able

.T

hesa

mpl

esin

clud

eob

serv

atio

nsw

hose

dist

rict

capi

tals

are

less

than

50ki

lom

eter

sfr

omth

em

ita

boun

dary

.C

oeffi

cien

tsth

atar

esi

gnifi

cant

lydi

ffere

ntfr

omze

roar

ede

note

dby

the

follo

win

gsy

stem

:*1

0%,**

5%,an

d**

*1%

.

Table 5: 1572 Tribute and Population

Dependent variable is:

Log Share of Tribute Revenues toMean Spanish Spanish Spanish Indig. Percent

Tribute Nobility Priests Justices Mayors Men Boys Females

(1) (2) (3) (4) (5) (6) (7) (8)

A: Quadratic Polynomial in Distance to PotosıMita 0.038 -0.019 0.008 0.011 -0.000 -0.010 0.000 -0.006

(0.027) (0.026) (0.016) (0.009) (0.004) (0.009) (0.009) (0.011)Elevation -0.030 -0.023 0.018 0.003 0.002 0.006 -0.001 0.012

(0.039) (0.031) (0.022) (0.011) (0.007) (0.021) (0.015) (0.031)Slope 0.003 -0.003 0.002 -0.000 0.000 -0.002 -0.000 0.003

(0.006) (0.003) (0.003) (0.002) (0.001) (0.002) (0.002) (0.002)R2 0.650 0.062 0.085 0.170 0.144 0.284 0.193 0.458

B: Interacted Quadratic Polynomial in Distance to Mita BoundaryMita -0.010 0.021 -0.012 -0.001 -0.006 -0.010 0.009 -0.006

(0.047) (0.027) (0.020) (0.009) (0.006) (0.015) (0.021) (0.024)Elevation -0.012 -0.019 0.023 -0.006 0.002 0.000 0.003 0.017

(0.051) (0.027) (0.020) (0.011) (0.007) (0.021) (0.016) (0.033)Slope -0.000 -0.003 0.002 0.000 0.001 -0.001 -0.000 0.002

(0.005) (0.003) (0.002) (0.001) (0.001) (0.002) (0.002) (0.002)R2 0.622 0.099 0.113 0.149 0.202 0.348 0.207 0.448

Geo. Controls yes yes yes yes yes yes yes yesBoundary F.E.s yes yes yes yes yes yes yes yesMean Dep. Var. 1.591 0.625 0.203 0.127 0.044 0.193 0.204 0.544Observations 65 65 65 65 65 65 65 65

The dependent variable in column (1) is the log of the district’s mean 1572 tribute rate (Miranda, 1975 [1583]). Incolumns (2) through (5), it is the share of tribute revenue allocated to Spanish nobility (encomenderos), Spanishpriests, Spanish justices, and indigenous mayors (caciques), respectively. In Columns (6) through (8), it is the shareof 1572 district population composed of males (aged 18 to 50), boys, and females (of all ages), respectively. Panel Aincludes a quadratic polynomial in Euclidean distance from the observation’s district capital to Potosı, and Panel Bincludes a quadratic polynomial in Euclidean distance to the nearest point on the mita boundary and interacts thispolynomial with the mita dummy. All regressions include soil type indicators and boundary segment fixed effects.The samples include districts whose capitals are less than 50 kilometers from the mita boundary. Column (1)weights by the square root of the district’s tributary population and columns (6) through (8) weight by the squareroot of the district’s total population. 66% of the observations are from mita districts. Coefficients that aresignificantly different from zero are denoted by the following system: *10%, **5%, and ***1%.

Table 6: Land Tenure and Labor Systems

Dependent variable is:

Percent of ruralHaciendas per tributary Percent of rural1000 district population in population in

Haciendas per residents in haciendas in haciendas in Land gini indistrict in 1689 1689 c. 1845 1940 1994

(1) (2) (3) (4) (5)

A: Quadratic Polynomial in Distance to PotosıMita -13.647*** -8.515*** -0.228** -0.138*** 0.107***

(2.470) (1.702) (0.086) (0.048) (0.037)Elevation -2.290 -0.393 0.076 -0.039 -0.065

(4.063) (3.755) (0.106) (0.120) (0.059)Slope 0.395 0.202 0.011 0.024*** -0.014**

(0.583) (0.483) (0.014) (0.009) (0.007)R2 0.537 0.509 0.320 0.377 0.189

B: Interacted Quadratic Polynomial in Distance to Mita BoundaryMita -7.479* -4.530 -0.194* -0.091 0.256***

(3.814) (3.366) (0.113) (0.111) (0.079)Elevation -0.362 -1.482 0.082 0.007 -0.085

(4.874) (4.316) (0.116) (0.125) (0.052)Slope 0.352 0.276 0.009 0.024*** -0.013**

(0.563) (0.451) (0.012) (0.009) (0.006)R2 0.556 0.537 0.328 0.377 0.225

Geo. Controls yes yes yes yes yesBoundary F.E.s yes yes yes yes yesMean Dep. Var. 6.500 5.336 0.135 0.263 0.783Observations 74 74 81 119 181

The unit of observation is the district. Robust standard errors are in parentheses. The dependent variable incolumn (1) is haciendas per district in 1689 and in column (2) is haciendas per 1000 district residents in 1689(Villanueva Urteaga, 1982). In column (3) it is the percentage of the district’s tributary population residing inhaciendas c. 1845 (Peralta Ruiz, 1991), in column (4) it is the percentage of the district’s rural populationresiding in haciendas in 1940 (Censo Nacional de Poblacion y Ocupacion, 1944), and in column (5) it is thedistrict land gini (Tercer Censo Nacional Agropecuario, 1994). Panel A includes a quadratic polynomial inEuclidean distance from the observation’s district capital to Potosı, and Panel B includes a quadraticpolynomial in Euclidean distance to the nearest point on the mita boundary and interacts this polynomialwith the mita dummy. All regressions include soil type indicators and boundary segment fixed effects. Thesamples include districts whose capitals are less than 50 kilometers from the mita boundary. Column (3) isweighted by the square root of the district’s rural tributary population and column (4) is weighted by thesquare root of the district’s rural population. 58% of the observations are in mita districts in columns (1) and(2), 59% in column (3), 62% in column (4), and 66% in column (5). Coefficients that are significantly differentfrom zero are denoted by the following system: *10%, **5%, and ***1%.

Table 7: Education

Dependent variable is:

Mean years Mean yearsLiteracy of schooling of schooling

1876 1940 2001(1) (2) (3)

A: Quadratic Polynomial in Distance to PotosıMita -0.024*** -0.219*** -0.806

(0.006) (0.081) (0.486)Elevation 0.015 0.047 0.156

(0.014) (0.134) (0.855)Slope 0.002* 0.020* 0.053

(0.001) (0.011) (0.084)R2 0.393 0.313 0.018

B: Inter. Quad. Polynomial in Dist. to Mita Bound.Mita 0.006 -0.075 -0.241

(0.019) (0.172) (1.331)Elevation 0.010 0.040 0.199

(0.012) (0.102) (0.632)Slope 0.002* 0.018 0.023

(0.001) (0.011) (0.078)R2 0.403 0.381 0.022

Geo. Controls yes yes yesBoundary F.E.s yes yes yesMean Dep. Var. 0.036 0.470 4.457Clusters 95 118 52Observations 95 118 4,038

The unit of observation is the district in columns (1) and (2) and the individual in column (3). Robuststandard errors, adjusted for clustering by district, are in parentheses. The dependent variable is meanliteracy in 1876 in column (1) (Censo general de la republica del Peru, 1878 ), mean years of schooling in 1940in column (2) (Censo Nacional de Poblacion y Ocupacion, 1944 ), and individual years of schooling in 2001 incolumn (3) (ENAHO, 2001). Panel A includes a quadratic polynomial in Euclidean distance from theobservation’s district capital to Potosı, and Panel B includes a quadratic polynomial in Euclidean distance tothe nearest point on the mita boundary and interacts this polynomial with the mita dummy. All regressionsinclude soil type indicators and boundary segment fixed effects. The samples include districts whose capitalsare less than 50 kilometers from the mita boundary. Columns (1) and (2) are weighted by the square root ofthe district’s population. 64% of the observations are in mita districts in column (1), 63% in column (2), and67% in column (3). Coefficients that are significantly different from zero are denoted by the following system:*10%, **5%, and ***1%.

Table 8: Roads

Dependent variable is:

Density ofDensity of Density of paved/gravellocal road regional road regionalnetworks networks roads

(1) (2) (3)

A: Quadratic Polynomial in Distance to PotosıMita 2.224 -40.587*** -35.666***

(13.577) (10.192) (9.066)Elevation -83.651*** -19.971 4.887

(22.725) (13.918) (11.719)Slope -7.965** -4.467** -2.433

(3.308) (1.810) (1.653)R2 0.221 0.268 0.262

B: Inter. Quad. Polynomial in Dist. to Mita Bound.Mita -15.409 -52.476** -26.075

(28.763) (24.119) (20.226)Elevation -80.645*** -27.615* -7.921

(21.331) (14.059) (11.657)Slope -6.969** -4.851** -2.494

(3.274) (1.867) (1.718)R2 0.228 0.260 0.234

Geo. Controls yes yes yesBoundary F.E.s yes yes yesMean Dep. Var. 85.34 33.55 22.51Observations 185 185 185

The unit of observation is the district. Robust standard errors are in parentheses. The road densities aredefined as total length in meters of the respective road type in each district divided by the district’s surfacearea, in km2. They are calculated using a GIS map of Peru’s road networks (Ministro de Transporte, 2006).Panel A includes a quadratic polynomial in Euclidean distance from the observation’s district capital toPotosı, and Panel B includes a quadratic polynomial in Euclidean distance to the nearest point on the mitaboundary and interacts this polynomial with the mita dummy. All regressions include soil type indicators andboundary segment fixed effects. The samples include districts whose capitals are less than 50 kilometers fromthe mita boundary. 66% of the observations are in mita districts. Coefficients that are significantly differentfrom zero are denoted by the following system: *10%, **5%, and ***1%.

Table 9: Consumption Channels

Dependent variable is:

HouseholdPercent of district Agricultural member

labor force in household sells employed outsideagriculture - part of produce in the agricultural

1993 markets - 1994 unit - 1994(1) (2) (3)

A: Quadratic Polynomial in Distance to PotosıMita 0.091 -0.226*** -0.015

(0.055) (0.032) (0.018)Elevation -0.015 -0.026 0.067**

(0.091) (0.037) (0.031)Slope -0.003 -0.004 -0.002

(0.009) (0.006) (0.003)R2 0.175 0.140 0.013

B: Interacted Quadratic Polynomial in Distance to Mita BoundaryMita 0.196 0.023 -0.039

(0.150) (0.055) (0.066)Elevation -0.023 -0.124*** 0.042

(0.078) (0.040) (0.026)Slope 0.005 0.005 -0.002

(0.011) (0.006) (0.004)R2 0.194 0.154 0.011

Geo. Controls yes yes yesBoundary F.E.s yes yes yesMean Dep. Var. 0.697 0.173 0.245Clusters 179 178 182Observations 179 160,990 183,596

Sample Falls Within:

border <25 km <15 km <10 km <5 kmdistrict of bound. of bound. of bound. of bound.

C: Ag. Mkts. - OLS (1) (2) (3) (4) (5)

Mita -0.178*** -0.158*** -0.112*** -0.070* -0.012(0.050) (0.042) (0.037) (0.037) (0.045)

Geo. Controls yes yes yes yes yesBoundary F.E. yes yes yes yes yesMean Dep. Var. 0.14 0.17 0.15 0.13 0.07Clusters 60 89 55 38 17Observations 55,838 86,988 46,375 31,175 14,534

Robust standard errors, adjusted for clustering by district in columns (2) and (3) of Panels A and B and allcolumns of Panel C, are in parentheses. The dependent variable in column (1) of Panels A and B is thepercentage of the district’s labor force engaged in agriculture as a primary occupation (IX Censo dePoblacion, 1993). In column (2) of Panels A and B and all columns of Panel C, the dependent variable is anindicator equal to one if the agricultural unit sells at least part of its produce in markets, and in column (3) ofPanels A and B, it is an indicator equal to one if at least one member of the household pursues secondaryemployment outside the agricultural unit (Tercer Censo Nacional Agropecuario, 1994). Panel A includes aquadratic polynomial in Euclidean distance from the observation’s district capital to Potosı, and Panel Bincludes a quadratic polynomial in Euclidean distance to the nearest point on the mita boundary and interactsthis polynomial with the mita dummy. All regressions include soil type indicators and boundary segment fixedeffects. Column (1) of Panels A and B is weighted by the square root of the district’s population. In Panels Aand B, 66% of the observations in column (1) are in mita districts, 68% in column (2), and 69% in column (3).In Panel C, 63% of the observations in column (1), 63% of the observations in column (2), 60% of theobservations in column (3), and 59% of the observations in column (4) are in mita districts. Coefficients thatare significantly different from zero are denoted by the following system: *10%, **5%, and ***1%.

Figure 1

!

!

Study Boundary

Mita Boundary

5000 m

0 m

Potosi

Huancavelica

UyuniSalt Flat

The mita boundary is in black and the study boundary in light gray. Districts falling insidethe contiguous area formed by the mita boundary contributed to the mita. Elevation is shownin the background.

47

Figure 2

!

!

!! !!!

! !!

!

!

!

! !

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!!

!

!

!

!

!!

!

!

!

!

!

!

!

!

!

!

!

!!

!

!

!!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

! !

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!

!!

!

!

!

!

!

!

!

!

!

!

!

!

!

!!

!

!!

!

!

!

!

!

!

!

!

!!

!

!

!

!

!

!

!

!

!!

!

!

!

!

!

! District Capitals

Study Boundary

District Boundaries

Grid Cells

>5000 m

0 m

Mita districts fall between the two thick lines. The circles show district capitals within 50kilometers of the mita boundary. The boundaries for the 20 x 20 km grid cells - used in Table1 - are in light gray. District boundaries are in black, and elevation is shown in thebackground.

48

Figure 3

56

78

9lo

g e

quiv

. consum

ption

−50 −25 0 25 50Distance to mita boundary (km)

A. Equivalent Consumption (2001)

0.2

5.5

.75

perc

ent

stu

nte

d

−50 −25 0 25 50Distance to mita boundary (km)

B. Childhood Stunting (2005)

015

30

45

60

no.

of

hacie

ndas

−50 −25 0 25 50Distance to mita boundary (km)

C. Haciendas (1689)

0.3

.6.9

% in h

acie

ndas

−50 −25 0 25 50Distance to mita boundary (km)

D. Haciendas (1845)

49

Figure 3 (cont.)0

.25

.5.7

51

% in

hac

iend

as

−50 −25 0 25 50Distance to mita boundary (km)

E. Haciendas (1940)

24

68

10ye

ars

of e

duc.

−50 −25 0 25 50Distance to mita boundary (km)

F. Schooling (2001)

010

020

030

0m

roa

ds/k

m2

−50 −25 0 25 50Distance to mita boundary (km)

G. Density of Roads (2006)

0.2

5.5

.75

sell

prod

uce

in m

kt.

−50 −25 0 25 50Distance to mita boundary (km)

H. Partic. in Ag. Markets (1994)

This figure plots distance to the mita boundary against various outcomes. Circles to the leftof the vertical line fall outside the mita catchment and circles to the right fall inside. Thethick lines give predicted outcomes from a regression that includes a second order polynomialin distance to Potosı and the mita dummy, and the thin lines are 95% confidence bands.

50

6. Detailed Data Description

Mita assignment. A list of colonial districts that contributed to the mita is taken

from Saignes (1984) (Potosı) and Amat y Junient (1947, p. 249, 264) (Huancavelica).1

Given that mita assignments are at the level of the colonial district and living standards

data are at the level of the contemporary district, I used the following procedure to

determine the colonial district to which every contemporary district pertained:

1. A colonial district consisted of a principal population center and smaller population

centers (anexos) located in the surrounding countryside. Many of the anexos later

became separate districts, that today still bear the same names. Thus, to facilitate

matching, I first obtained a detailed list of Peruvian colonial population centers,

by district, from Geografıa del Peru Virreinal (Bueno, 1951 [1764-1778]). For each

of these population centers, we know from Saignes and Amat y Junient whether

it contributed to the mita, which varied at the level of the colonial district, and

Geografıa also lists its colonial province.

2. I next compiled a list of all contemporary districts, by province, in southern and

central Peru. Note that a contemporary district in most cases consists of a principal

population center and its surrounding countryside.

3. Finally, I assigned every contemporary district from the list compiled in step 2) a

mita status as follows:

(a) I began by using names to match, province-by-province, the districts compiled

in step 2) with the colonial population centers compiled in step 1).2 Colo-

nial provinces correspond closely with modern provinces, making matching

province-by-province feasible. In no case is there more than one contempo-

rary district or more than one colonial population center with the same name

1The list of subjected districts remained the same throughout most of the colonial period, and Iuse the original mita assignments. Specifically, the first mita repartimiento (list of subjected districts)was drawn up in 1573. In 1578, eighteen districts subjected in the previous 1575 list do not reappear.They were primarily districts with small populations that do not appear in any later colonial censusesor documents, which suggests that they were incorporated into nearby districts (Bakewell, 1984, p. 83).Moreover, several districts in Condesuyos (now part of Arequipa) were briefly required to contribute in1578, but were subsequently re-exempt. These districts are coded as non-mita, with results robust toexcluding the small portion of the boundary along which they fall.

2There are three administrative levels in Peru. The largest is the department, of which there arefive in the region that I examine. Below the department is the province and below the province thedistrict, alternatively termed a municipality or a canton in other Latin American countries.

A–1

in a single province. 86 % of contemporary non-mita districts and 72% of

contemporary mita districts were assigned mita status using this procedure.

(b) For an additional 7 contemporary mita districts (3.3 %), I located a district

of the same name in the same province in Bachmann’s Historia de la demar-

cacion polıtica del Peru (1869), a detailed account of historical demarcation

in Peru that lists districts formed during the early post-Independence period

and the colonial unit from which each split. These districts could thus be

matched with their colonial district using this source.

(c) For the remaining 14 % of contemporary non-mita districts and 25% of con-

temporary mita districts, it was not possible to document their corresponding

colonial district using existing sources on historical political demarcation. All

but eight of these districts fell within the interior of the mita catchment, as

constructed using the districts matched in the previous steps. The eight dis-

tricts (4 non-mita and 4 mita) that fell along the boundary (2.6% of contem-

porary districts being matched) were all located in contemporary provinces

that consist entirely of mita or non-mita districts. Since it is highly probable

that each of the eight districts originally belonged to another district in its

province, I assigned them the mita status of the remaining districts in the

province.

This procedure is summarized in Appendix Table A3. It is also of interest to calculate

how many contemporary districts existed as juridical entities historically, where I de-

fine a contemporary district as existing historically if it’s district capital also served as

the capital of a colonial district. As documented in column (2) of Table A3, 70% of

contemporary non-mita districts existed as juridical entities during the colonial period,

as compared to 51% of contemporary mita districts (Rodriguez Gutierrez, 2000; Bach-

mann, 1869). This implies that mita districts have been somewhat more likely to split

into multiple districts during the post-Independence period than non-mita districts.3

Living Standards. Household level data on consumption and ethnicity are from

the National Household Survey (ENAHO), which the Peruvian Institute of Statistics

and Information (INEI) collected in the fourth quarter of 2001. ENAHO is similar to

3Districts typically split when a previously smaller hamlet in a district reaches a pre-specified size(this cutoff has varied across time - see Bachmann (1869) and Gutierrez (2000)). The more pronouncedcolonial demographic collapse in mita districts (Wightman, 1990, p. 72) may have offered more scopefor later population recovery, leading more new districts to form.

A–2

the World Bank Living Standards Measurement Survey, but offers a substantially larger

sample and more extensive geographic coverage. Consumption is measured in 2000 soles.

I subtract total transfers from total consumption, and normalize to Lima metropolitan

prices using the local deflation factors provided in ENAHO (2001).

Individual level data on heights are taken from a census collected by the Ministry of

Education that records the heights of six to nine year old school children in the region. I

use the complete micro dataset, which lists each child’s age in months, gender, height in

centimeters, and whether or not the child is stunted. Following international standards,

children whose heights are more than two standard deviations below their age-specific

median are classified as stunted, with the medians and standard deviations calculated

by the World Health Organization from an international reference population.

Geographic controls. I obtain the coordinates of district capitals from departmen-

tal statistical reports published by INEI (2001). A GIS map with district administrative

boundaries was also produced by INEI. I first code each district as inside or outside the

mita catchment using the mita assignment data described above. Then, I use geospatial

software to calculate the Euclidean distance of each district capital to Potosı and to the

nearest point on the mita boundary, as well as the location of the point.4

Elevation data are from the Shuttle Radar Topography Mission (SRTM), organized

by the U.S. National Aeronautics and Space Agency (2000). The data are at 30 arc

second resolution, which corresponds to a cell size of around one square kilometer. I

use the SRTM data to obtain both the area-weighted elevation and slope within each

district.5

I utilize soil type data, at a scale of one to five million, produced by the Soil Terrain

Database for Latin America (SOTERLAC). SOTERLAC employs the standard FAO

soil type categorization. I construct a series of soil type dummies equal to one for the

soil type(s) which predominate over the greatest percentage of the district’s landmass

area, and equal to zero otherwise.

Data used for additional robustness checks. The variable indigenous is taken

from ENAHO (2001), which asks the household head and spouse the primary language

they speak at home. This indicator is coded as one if the household head primarily speaks

4An equidistant cylindrical projection centered in Peru is used to ensure that distances are minimallydistorted when projecting the earth’s surface to a flat plane.

5For these calculations, I use the UTM WGS1984 - Zone 18S projection, which produces very littledistortion when calculating surface areas for the region examined in this paper.

A–3

an indigenous language (in most cases Quechua) and is coded as zero otherwise. The

locations of rivers are found using a GIS dataset of world rivers prepared by the Earth

Science Research Institute (2004). The locations of Inca royal estates are obtained from

D’Altoy (2002). D’Altoy provides a comprehensive list of estates that are mentioned

in Inca histories written during the colonial era or that have been located by modern

archaeological digs. Data on migration are from the 1993 Population Census.

Pre-mita outcomes. I obtain data on local per capita tribute contributions, on the

allocation of tributes revenues to various groups, and on local demographics just prior

to the mita’s enactment from Viceroy Francisco Toledo’s Tasa de la Visita General

(Tribute Assessment, General Visit). Toledo blamed demographic collapse on excessive,

unregulated rates of tribute extraction by local Spanish elites. Thus, he coordinated an

in depth inspection of modern Peru, Bolivia, and Ecuador in the early 1570’s to evaluate

the maximum tribute that could be demanded from local groups without threatening

subsistence. In order to assess ability to pay, colonial authorities ordered teams of

surveyors to list the ages and occupations of residents; inspect the communities’ grain

storage facilities; uncover the tribute that residents provided in the past; investigate a

series of geographic and economic questions relating to natural resources and agricultural

production; record the tribute, labor services, and land received by indigenous leaders

and Spanish administrators; and investigate a variety of other questions. Based on

these assessments of ability to pay, authorities assigned varying tribute obligations at

the level of the district - socioeconomic group, with districts containing either one or

two socioeconomic groups. (Districts with two socioeconomic groups usually consisted

of a wealthier group engaged in farming and a poorer group engaged in fishing.) These

per capita contributions reflect Spanish authorities’ best estimates of local economic

prosperity, with more prosperous groups paying more in tribute.

Moreover, the assessment mandates how the tribute revenues were to be divided

between rents for Spanish nobility (encomenderos), salaries for Spanish priests, salaries

for local Spanish authorities (justicias), and salaries for indigenous mayors (caciques).

These data are informative about the financing of local government, about the extent

to which the Spanish nobility were permitted to extract local revenues, and about the

relative power of competing local administrators to obtain tribute revenues. Finally, the

tribute assessment also records the number of tribute-paying males (those aged 18 to

50), boys, old men, and women (of all ages) in each district. These data record the best

demographic picture available to colonial administrators just prior to the mita.

A–4

It is estimated that the original documents from the visita comprised 6,000 to 12,000

folios (Cook, 1975). Cristobal de Miranda produced an unabridged copy of the tribute

assessment portion of these documents in 1583. This copy has been preserved for the

entire study region, where data were collected primarily in 1572 and 1573 (Miranda,

1975 [1583]).

I aggregate the 1572 data - which is at the level of the encomienda - to modern dis-

tricts. Recall that an encomienda is a contiguous piece of territory in which appointed

Spaniards collected tribute and labor services. As part of his book on the Peruvian

encomienda, historian Jose Puente Brunke (1992) created detailed maps showing the

estimated geographic center of each encomienda. I aggregate 210 encomiendas to con-

temporary districts by overlaying the contemporary boundaries on the historical maps.

While the precise encomienda boundaries are not known, because encomiendas were typ-

ically small, this process is likely to be highly accurate. Moreover, many encomiendas

could also be matched using names, and both procedures yield the same results.6

Haciendas. Data on the concentration of haciendas in 1689 are contained in de-

tailed parish reports commissioned by Bishop Manuel de Mollinedo and submitted by

all parishes in the bishopric of Cusco. The bishopric included present day Cusco and

Apurimac departments, as well as portions of modern Puno and Arequipa departments,

thus providing coverage for most districts within 100 kilometers of the mita boundary.

The reports, submitted by 134 parishes, range from one to thirty-nine pages. All list the

number of haciendas in the parish’s jurisdiction. The data are at the level of the parish

subdivision. The reports were published by Horacio Villanueva Urteaga (1982).

I also utilize district level data on haciendas from the 19th century. These data,

collected by the republican government and preserved in the Treasury Section of Cusco’s

Municipal Archives, give the percentage of the rural tributary population (males between

the ages of 18 and 50) residing in haciendas, for districts in the present-day departments

of Cusco and Apurimac (two of the five departments in the region examined). Data

from 1845, 1846, and 1850 are combined to form the c. 1845 dataset. For some districts,

data are available for more than one year within this period. The numbers provided

change very little, and the earliest observation is used. The data are contained in Victor

6Nine 1572 encomiendas cannot be matched with current districts because their exact locations areunknown. Most had very low populations and likely disappeared soon after 1572 due to populationcollapse (Cook, 1982). The contemporary provinces in which these encomiendas were located areknown, and so I match each with the district containing it’s province’s capital. If I instead drop theseobservations, results are unchanged.

A–5

Peralta Ruiz’s 1991 compilation of Cusco tribute records.

Finally, data from the 1940 Peruvian Population Census on the number of inhabitants

in over 23,000 population centers (where anything from a small rural hut to a large city

is classified as a population center) are aggregated to the district level to calculate

the percentage of the rural population residing in haciendas. The census specifically

uses the category hacienda in classifying population centers. Other rural categories are

recognized and unrecognized indigenous communities and peasant landholdings of family

or sub-family size (estancias).

The 1689 hacienda data are at the level of the colonial parish. While parishes

are religious administrative divisions, in practice they corresponded closely with the

secular colonial administrative districts. Thus, I aggregate the 1689 data to the level

of the modern district using the information in Table A3. If a single colonial district

corresponds to more than one modern district, it is assigned to the contemporary district

that contains the colonial district capital. Few districts have merged, implying that the

unit of observation is similar to what it would be if I used the 1689 parish as the unit of

observation. I follow the same procedure to aggregate the 1845 and 1940 hacienda data

to the level of the modern district.

Education. The 1876 Population Census provides district level data on literacy.

For each district, it lists how many individuals are able to read, to write, or neither. A

literate individual is defined as one who can read, write, or both. The 1940 Population

Census provides information on mean years of schooling in each district. Individual level

data on years of schooling are drawn from ENAHO 2001.

Road networks. I calculate the densities of local and regional road networks using

a GIS road network map of Peru, produced by the Ministry of Transportation (2006).

Roads are classified as paved, gravel, non-gravel, and trocha carrozable. The total length

of the respective type of road within each district, accounting for changes in elevation, is

divided by the surface area of the district to obtain a road network density. Data on the

type of road providing access to district capitals (paved, dirt, horse track, or footpath)

are from the 2004 Peruvian Municipal Register, a census of district capitals collected by

INEI.

Shining Path. Data on the percent of votes cast blank or null in the 1989 municipal

elections come from Pareja and Gatti (1990), as do data on whether provincial and

district authorities were renewed. Data on blank/null votes in 2002 are from the National

A–6

Elections Board (Oficina Nacional de Procesos Electorales).

Consumption Channels. District level data on the percentage of the labor force

whose primary occupation is agriculture are obtained from the 1993 Peruvian Population

Census, collected by INEI. An individual is categorized in agriculture if he or she is

an agricultural wage laborer or primarily engaged in agricultural production. I create

this variable using the finest occupational categories available, as the more aggregated

groupings produced by INEI place unskilled agricultural workers in an “other” category

rather than classifying them as agricultural producers. The 1994 Peruvian Agricultural

Census is used to investigate market participation and supplementary employment. An

agricultural household is defined as participating in markets if it sold at least part of its

produce from one of its plots produced during the most recent harvest in markets.

A–7

References

Bachmann, C. J. (1869): Historia de la demarcacion polıtica del Peru, Lima: Impr.

G. Clauss.

Bueno, C. (1951): Geografıa del Peru Virreinal, Lima: Daniel Valcarcel.

Puente Brunke, J. (1992): Encomienda y encomenderos en el Peru: Estudio social

y politico de una institucion colonial, Sevilla: Diputacion Provincial.

Rodriguez Gutierrez, W. (2000): El Peru en casa: demarcacion territorial, Lima:

Servicio de Investigaciones Parlamentarias.

A–8

Table A1: 1572 Tribute Details

Percent of Mean per Standarddistricts capita deviation of Percent of

Tribute type contributing contribution contribution total tribute

Precious metals 100 4.151 0.591 0.816Grains 89.6 0.666 0.451 0.120Textiles 53.7 0.348 0.227 0.037Animals 92.5 0.146 0.093 0.027Total 100 5.070 0.388 1.00

Source: Miranda (1975 [1583]). Values are in 1572 pesos. The sample is limited to fall within 50 km ofthe mita boundary.

A–9

Table

A2:

Speci

fica

tion

Check

s

equiv

.st

unte

dtr

ibute

hac.

hac.

hac.

lit.

educ

educ

reg.

%in

mark

etco

ns.

gro

wth

1572

1689

1845

1940

1876

1940

2001

roads

ag.

part

ic.

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

(9)

(10)

(11)

(12)

Base

line

I:Q

uadrati

cpoly

nom

ialin

dis

tance

toPoto

sıM

ita

-0.3

73***

0.0

76***

0.0

38

-13.6

47***

-0.2

28**

-0.1

38***

-0.0

24***

-0.2

19***

-0.8

06

-40.5

87***

0.0

91

-0.2

26***

(0.1

28)

(0.0

25)

(0.0

27)

(2.4

70)

(0.0

86)

(0.0

48)

(0.0

06)

(0.0

81)

(0.4

86)

(10.1

92)

(0.0

55)

(0.0

32)

Base

line

II:In

teracte

dquadrati

cpoly

nom

ialin

dis

tance

tom

ita

boundary

Mit

a-0

.338

0.1

36*

-0.0

10

-7.4

79*

-0.1

94*

-0.0

91

0.0

06

-0.0

75

-0.2

41

-52.4

76**

0.1

96

0.0

23

(0.3

52)

(0.0

72)

(0.0

47)

(3.8

14)

(0.1

13)

(0.1

11)

(0.0

19)

(0.1

72)

(1.3

31)

(24.1

19)

(0.1

50)

(0.0

55)

Alt

ernati

ve

functi

onalfo

rm

sfo

rR

Dpoly

nom

ial:

Base

line

ILin

ear

poly

nom

ialin

dis

tance

toPoto

sıM

ita

-0.2

66**

0.0

54**

0.0

34

-13.4

74***

-0.2

22***

-0.1

11**

-0.0

23***

-0.2

00***

-0.4

21

-42.5

45***

0.0

80

-0.2

23***

(0.1

02)

(0.0

24)

(0.0

24)

(2.3

85)

(0.0

84)

(0.0

50)

(0.0

06)

(0.0

75)

(0.4

65)

(9.9

95)

(0.0

53)

(0.0

32)

Cubi

cpo

lynom

ialin

dis

tance

toPoto

sıM

ita

-0.4

23***

0.0

84***

0.0

43

-13.4

09***

-0.2

28**

-0.1

27**

-0.0

22***

-0.2

15**

-0.9

45*

-36.3

17***

0.1

24**

-0.2

03***

(0.1

29)

(0.0

27)

(0.0

27)

(2.4

68)

(0.0

99)

(0.0

54)

(0.0

07)

(0.0

84)

(0.5

12)

(10.0

40)

(0.0

62)

(0.0

31)

Quart

icpo

lynom

ialin

dis

tance

toPoto

sıM

ita

-0.4

29***

0.0

70***

0.0

44

-13.2

77***

-0.2

30**

-0.1

58***

-0.0

23***

-0.1

96**

-0.9

86**

-36.3

00***

0.0

98*

-0.2

11***

(0.1

32)

(0.0

22)

(0.0

27)

(2.4

66)

(0.1

00)

(0.0

47)

(0.0

07)

(0.0

78)

(0.4

88)

(10.0

74)

(0.0

51)

(0.0

29)

RD

poly

nom

ial(B

ase

line

I)in

teracte

dw

ith

Mit

aIn

tera

cted

linea

rpo

lynom

ialin

dis

tance

toPoto

sıM

ita

-0.2

31**

0.0

39*

0.0

44

-13.2

14***

-0.2

28***

-0.1

16**

-0.0

18***

-0.0

71

-0.1

30

-42.6

32***

-0.0

08

-0.1

38***

(0.1

05)

(0.0

22)

(0.0

30)

(3.0

95)

(0.0

78)

(0.0

57)

(0.0

06)

(0.0

69)

(0.4

47)

(10.1

02)

(0.0

47)

(0.0

32)

Inte

ract

edqu

adra

tic

poly

nom

ialin

dis

tance

toPoto

sıM

ita

-0.3

47*

0.0

32

0.0

46

-12.4

36***

-0.2

34***

-0.2

00***

-0.0

10

0.0

26

-0.7

67

-46.7

09***

-0.0

83

-0.1

43***

(0.1

76)

(0.0

22)

(0.0

33)

(2.8

98)

(0.0

72)

(0.0

70)

(0.0

08)

(0.0

79)

(0.6

24)

(12.6

60)

(0.0

51)

(0.0

35)

Alt

ernati

ve

functi

onalfo

rm

for

RD

poly

nom

ial:

Base

line

IIIn

tera

cted

linea

rpo

lynom

ialin

dis

tance

tom

ita

boundary

Mit

a-0

.373

0.0

93

0.0

24

-12.1

31***

-0.1

70*

-0.0

86

-0.0

08

-0.1

97

-1.7

57*

-35.8

09*

0.2

10

-0.0

53

(0.2

24)

(0.0

57)

(0.0

33)

(3.0

77)

(0.0

89)

(0.0

85)

(0.0

13)

(0.1

70)

(0.8

78)

(18.2

12)

(0.1

28)

(0.0

49)

RD

poly

nom

ial(B

ase

line

II)

not

inte

racte

dw

ith

Mit

aLin

ear

poly

nom

ialin

dis

tance

tom

ita

boundary

Mit

a-0

.249**

0.0

57**

0.0

46**

-12.8

99***

-0.2

22***

-0.1

11**

-0.0

24***

-0.2

15***

-0.4

13

-41.4

26***

0.0

83

-0.2

36***

(0.1

06)

(0.0

24)

(0.0

22)

(2.2

19)

(0.0

62)

(0.0

51)

(0.0

06)

(0.0

73)

(0.4

68)

(10.3

42)

(0.0

51)

(0.0

35)

Quadra

tic

poly

nom

ialin

dis

tance

tom

ita

boundary

Mit

a-0

.251**

0.0

60**

0.0

47**

-12.8

59***

-0.2

17***

-0.1

17**

-0.0

24***

-0.2

22***

-0.4

11

-41.9

14***

0.0

95**

-0.2

28***

(0.1

05)

(0.0

24)

(0.0

21)

(2.2

17)

(0.0

62)

(0.0

49)

(0.0

06)

(0.0

73)

(0.4

75)

(10.3

10)

(0.0

48)

(0.0

33)

Cubi

cpo

lynom

ialin

dis

tance

tom

ita

boundary

Mit

a-0

.249**

0.0

64***

0.0

49**

-13.1

94***

-0.2

19***

-0.1

11**

-0.0

26***

-0.2

33***

-0.4

11

-42.0

40***

0.1

02**

-0.2

30***

(0.1

02)

(0.0

24)

(0.0

22)

(2.1

97)

(0.0

64)

(0.0

48)

(0.0

07)

(0.0

76)

(0.4

60)

(10.3

38)

(0.0

48)

(0.0

32)

RD

poly

nom

ials

inclu

ded

from

both

base

line

specifi

cati

ons

Mit

a-0

.365

0.1

25*

-0.0

32

-6.1

63

-0.1

84

-0.0

88

0.0

07

-0.0

79

-0.3

30

-58.0

35**

0.1

94

0.0

13

(0.3

28)

(0.0

68)

(0.0

49)

(4.1

73)

(0.1

19)

(0.1

17)

(0.0

19)

(0.1

72)

(1.2

40)

(24.4

04)

(0.1

50)

(0.0

55)

RD

poly

nom

ials

inla

titu

de,lo

ngit

ude,and

inte

racti

ons

Mit

a-0

.388**

0.0

95**

0.0

16

-12.8

44***

-0.0

42

-0.1

14*

-0.0

16*

-0.2

41**

-1.1

04**

-29.3

58**

0.1

39*

-0.1

86***

(0.1

58)

(0.0

37)

(0.0

22)

(2.7

09)

(0.0

72)

(0.0

58)

(0.0

09)

(0.1

20)

(0.5

49)

(11.9

73)

(0.0

84)

(0.0

39)

Ordin

ary

Least

Squares

Mit

a-0

.259**

0.0

48*

0.0

47**

-12.9

00***

-0.2

24***

-0.1

13**

-0.0

23***

-0.1

96**

-0.3

79

-42.5

54***

0.0

74

-0.2

30***

(0.1

03)

(0.0

27)

(0.0

22)

(2.2

07)

(0.0

62)

(0.0

49)

(0.0

06)

(0.0

77)

(0.4

54)

(10.1

50)

(0.0

56)

(0.0

36)

Geo

.C

ontr

ols

yes

yes

yes

yes

yes

yes

yes

yes

yes

yes

yes

yes

Boundary

F.E

.syes

yes

yes

yes

yes

yes

yes

yes

yes

yes

yes

yes

Clu

ster

s52

185

65

74

81

119

95

118

52

185

179

178

Obse

rvati

ons

1013

100,4

46

65

74

81

119

95

118

4038

185

179

160,9

90

Table A3: Assignment Non-Mita Districts

District Colonial Cusco Historical MatchingName District Metro Province Source

Abancay yes no Abancay G.V.Accha yes no Chilques y Masques G.V.Achoma yes no Arequipa G.V.Alca yes no Condesuyos de Cusco G.V.Ancahuasi no (1986) no Abancay MapAndagua yes no Arequipa G.V.Andaray yes no Arequipa G.V.Anta yes no Abancay G.V.Ayo yes no Arequipa G.V.Cabanaconde yes no Arequipa MapCachimayo no no Abancay MapCahuacho no no Arequipa MapCaicay yes no Chilques y Masques G.V.Calca yes no Calca y Lares G.V.Callalli yes no Arequipa G.V.Cayarani yes no Arequipa G.V.Caylloma yes no Arequipa G.V.Ccapi yes no Chilques y Masques G.V.Ccorca no yes Cusco G.V.Chachas yes no Arequipa G.V.Challabamba yes no Chilques y Masques G.V.Charcana yes no Condesuyos de Cusco G.V.Chichas yes no Arequipa G.V.Chilcaymarca no no Arequipa G.V.Chinchaypujio no no Abancay MapChinchero yes no Urubabma G.V.Chivay yes no Arequipa MapChoco yes no Arequipa G.V.Chuquibamba yes no Arequipa G.V.Colcha yes no Chilques y Masques G.V.Colquepata yes no Chilques y Masques G.V.Coporaque (Caylloma) yes no Arequipa G.V.Cotahuasi yes no Condesuyos de Cusco G.V.Coya yes no Calca y Lares G.V.Curahuasi yes no Abancay G.V.Cusco yes yes Cusco G.V.Huambo no (1875) no Arequipa G.V.Huancarani no (1987) no Chilques y Masques G.V.Huanipaca no (1893) no Abancay G.V.Huanoquite no no Chilques y Masques G.V.Huarocondo no (1896) no Abancay G.V.Huayllabamba yes no Urubabma G.V.Huaynacotas yes no Condesuyos de Cusco G.V.Ichupampa yes no Arequipa G.V.Iray no no Arequipa MapLamay yes no Calca y Lares G.V.Lari yes no Arequipa G.V.Limatambo yes no Abancay G.V.

A–11

Table A3: Assignment Non-Mita Districts

District Colonial Cusco Historical MatchingName District Metro Province Source

Maca yes no Arequipa G.V.Machaguay no (1889) no Arequipa G.V.Madrigal yes no Arequipa G.V.Maras yes no Urubabma G.V.Mollepata no no Abancay G.V.Ollantaytambo yes no Urubabma G.V.Omacha yes no Chilques y Masques G.V.Orcopampa yes no Arequipa G.V.Paccaritambo yes no Chilques y Masques G.V.Pampacolca yes no Arequipa G.V.Pampamarca (La Union) yes no Condesuyos de Cusco G.V.Paruro yes no Chilques y Masques G.V.Pillpinto no no Chilques y Masques G.V.Pisac yes no Calca y Lares G.V.Poroy no yes Cusco G.V.Pucyura yes no Abancay MapPuyca no (1891) no Condesuyos de Cusco G.V.Quechualla yes no Arequipa G.V.Salamanca yes no Arequipa G.V.San Antonio De Chuca no no Arequipa MapSan Jeronimo yes yes Cusco G.V.San Pedro De Cachora no no Abancay G.V.San Salvador yes no Calca y Lares G.V.San Sebastian yes yes Cusco G.V.Santiago no yes Cusco G.V.Sayla yes no Arequipa MapSaylla no yes Cusco G.V.Sibayo no no Arequipa G.V.Tamburco no no Abancay MapTapay yes no Arequipa G.V.Taray yes no Calca y Lares G.V.Tauria no no Arequipa MapTipan no no Arequipa G.V.Tisco yes no Arequipa G.V.Tomepampa yes no Condesuyos de Cusco G.V.Toro yes no Condesuyos de Cusco G.V.Tuti yes no Arequipa G.V.Unon no no Arequipa MapUrubamba yes no Urubabma G.V.Viraco yes no Arequipa G.V.Wanchaq no (1987) yes Cusco G.V.Yanaquihua yes no Arequipa G.V.Yanque yes no Arequipa G.V.Yaurisque yes no Chilques y Masques G.V.Yucay yes no Urubabma G.V.Zurite yes no Abancay G.V.

A–12

Table A3: Assignment Mita Districts

District Colonial Cusco Historical MatchingName District Metro Province Source

Accomarca no no Vilcas Huaman G.V.Achaya yes no Azangaro G.V.Acomayo yes no Quispicanchis G.V.Acopia no no Quispicanchis G.V.Acos yes no Quispicanchis G.V.Alto Pichigua no (1994) no Canas y Canchis MapAnco-Huallo no no Andahuaylas G.V.Andahuaylas yes no Andahuaylas G.V.Andahuaylillas yes no Quispicanchis G.V.Andarapa no no Andahuaylas G.V.Antabamba yes no Aymaraes G.V.Apongo no no Vilcas Huaman G.V.Asillo yes no Azangaro G.V.Asquipata no (1986) no Vilcas Huaman MapAtuncolla yes no Lampa G.V.Ayaviri yes no Lampa G.V.Belen no no Lucanas G.V.Cabana yes no Lampa G.V.Cabanilla yes no Lampa G.V.Cabanillas no no Lampa G.V.Calapuja yes no Lampa G.V.Caminaca yes no Azangaro G.V.Canaria yes no Vilcas Huaman G.V.Capacmarca yes no Chumbivilcas G.V.Capaya no no Aymaraes G.V.Caracoto yes no Lampa G.V.Caraybamba no no Aymaraes G.V.Carhuanca yes no Vilcas Huaman G.V.Cayara no no Vilcas Huaman G.V.Ccarhuayo no no Quispicanchis MapCcatca yes no Quispicanchis G.V.Chacoche no no Aymaraes G.V.Chalcos no no Lucanas G.V.Chalhuanca yes no Aymaraes G.V.Challhuahuacho no (1994) no Cotabambas MapChamaca yes no Chumbivilcas G.V.Chapimarca yes no Aymaraes G.V.Chavina no no Lucanas G.V.Checacupe yes no Canas y Canchis G.V.Checca yes no Canas y Canchis G.V.Chiara yes no Andahuaylas G.V.Chilcayoc no no Lucanas G.V.Chincheros yes no Andahuaylas G.V.Chipao yes no Lucanas G.V.Chumpi yes no Lucanas BachmannChuquibambilla yes no Cotabambas BachmannCirca yes no Aymaraes G.V.Cocharcas yes no Andahuaylas G.V.Colca yes no Vilcas Huaman G.V.Colcabamba yes no Aymaraes G.V.Colquemarca yes no Chumbivilcas G.V.Colta yes no Parinacochas G.V.Combapata yes no Canas y Canchis G.V.Concepcion no no Vilcas Huaman G.V.

A–13

Table A3: Assignment Mita Districts

District Colonial Cusco Historical MatchingName District Metro Province Source

Condoroma yes no Canas y Canchis G.V.Coporaque (Espinar) yes no Canas y Canchis G.V.Coracora yes no Parinacochas G.V.Corculla yes no Parinacochas G.V.Coronel Castaneda no no Lucanas MapCotabambas yes no Cotabambas G.V.Cotaruse no no Aymaraes G.V.Coyllurqui no no Cotabambas G.V.Cupi yes no Lampa G.V.Curasco no (1993) no Cotabambas G.V.Curpahuasi no no Cotabambas MapCusipata no no Quispicanchis MapEl Oro no no Aymaraes G.V.Espinar yes no Canas y Canchis G.V.Gamarra no no Cotabambas G.V.Haquira yes no Cotabambas G.V.Huacana yes no Lucanas G.V.Huaccana no (1985) no Andahuaylas MapHuambalpa yes no Vilcas Huaman G.V.Huancapi no no Vilcas Huaman G.V.Huancarama yes no Andahuaylas G.V.Huancaray no no Andahuaylas G.V.Huaquirca no no Aymaraes G.V.Huaro no no Quispicanchis MapHuaya yes no Vilcas Huaman MapHuayana no (1984) no Andahuaylas G.V.Huayllati yes no Cotabambas G.V.Huayllo no no Aymaraes MapIndependencia no (1986) no Vilcas Huaman MapJose Domingo Choquehuanca no no Azangaro MapJuan Espinoza Medrano no no Aymaraes G.V.Juliaca yes no Lampa G.V.Justo Apu Sahuaraura no (1984) no Aymaraes MapKaquiabamba no (1995) no Andahuaylas MapKishuara no no Andahuaylas MapKunturkanki no no Canas y Canchis MapLambrama yes no Aymaraes G.V.Lampa (Lampa) yes no Lampa G.V.Lampa (Paucar del Sara Sara) yes no Parinacochas G.V.Langui yes no Canas y Canchis G.V.Layo yes no Canas y Canchis G.V.Livitaca yes no Chumbivilcas G.V.Llalli yes no Lampa G.V.Llusco yes no Chumbivilcas G.V.Lucre (Aymaraes) no no Aymaraes G.V.Lucre (Quispicanchi) no yes Quispicanchis G.V.Macari yes no Lampa G.V.Mamara yes no Cotabambas G.V.Manazo no no Lampa G.V.Manazo no no Lampa G.V.Mara yes no Cotabambas G.V.Marangani yes no Canas y Canchis G.V.Marcabamba no no Parinacochas MapMicaela Bastidas no no Cotabambas Map

A–14

Table A3: Assignment Mita Districts

District Colonial Cusco Historical MatchingName District Metro Province Source

Morcolla no no Lucanas MapMosoc Llacta no no Quispicanchis MapNicasio yes no Lampa G.V.Nunoa no no Lampa G.V.Ocobamba yes no Andahuaylas G.V.Ocongate yes no Quispicanchis G.V.Ocoruro no no Canas y Canchis MapOcros yes no Vilcas Huaman G.V.Ocuviri yes no Lampa G.V.Ongoy yes no Andahuaylas G.V.Oropesa (Antabamba) yes no Aymaraes G.V.Oropesa (Quispicanchi) yes yes Quispicanchis G.V.Orurillo yes no Lampa G.V.Oyolo yes no Parinacochas G.V.Pacapausa yes no Lucanas BachmannPachaconas no (1872) no Aymaraes G.V.Pacobamba no no Andahuaylas MapPacucha no no Andahuaylas MapPaico yes no Lucanas G.V.Palca no (1901) no Lampa BachmannPallpata no no Canas y Canchis MapPampachiri yes no Andahuaylas G.V.Pampamarca (Canas) yes no Canas y Canchis G.V.Pararca yes no Parinacochas BachmannParatia no no Lampa MapPataypampa no no Cotabambas MapPaucarcolla yes no Paucarcolla G.V.Pausa yes no Parinacochas G.V.Pichigua yes no Canas y Canchis G.V.Pichirhua yes no Aymaraes G.V.Pitumarca no no Canas y Canchis G.V.Pocohuanca no no Aymaraes MapPomacanchi yes no Quispicanchis G.V.Pomacocha no no Andahuaylas G.V.Progreso no no Cotabambas MapPucara yes no Lampa G.V.Puno yes no Paucarcolla G.V.Puquio yes no Lucanas G.V.Puyusca no no Lucanas MapQuehue no no Canas y Canchis G.V.Querobamba yes no Lucanas G.V.Quinota no no Chumbivilcas G.V.Quiquijana yes no Quispicanchis G.V.Ranracancha no (1993) no Andahuaylas MapRondocan yes no Quispicanchis G.V.Sabaino no (1872) no Aymaraes MapSan Antonio (Grau) no no Cotabambas MapSan Antonio (Puno) yes no Paucarcolla BachmannSan Antonio De Cachi no no Andahuaylas MapSan Cristobal no (1986) no Lucanas G.V.San Francisco De Ravacayco no no Lucanas MapSan Javier De Alpabamba no no Parinacochas G.V.

A–15

Table A3: Assignment Mita Districts

District Colonial Cusco Historical MatchingName District Metro Province Source

San Jeronimo yes no Andahuaylas G.V.San Jose De Ushua no no Parinacochas MapSan Juan De Chacna no no Aymaraes MapSan Miguel De Chaccrampa no (1990) no Andahuaylas MapSan Pablo yes no Canas y Canchis G.V.San Pedro (Canchis) no no Canas y Canchis G.V.San Pedro (Lucanas) no no Lucanas MapSan Pedro De Larcay no no Lucanas G.V.San Salvador De Quije no no Lucanas MapSanayca no no Aymaraes G.V.Sancos yes no Lucanas BachmannSangarara no (1861) no Quispicanchis G.V.Santa Ana De Huaycahuacho no no Lucanas G.V.Santa Lucia no no Lampa MapSanta Maria De Chicmo no no Andahuaylas MapSanta Rosa no (1990) no Cotabambas MapSanta Rosa (Melgar) yes no Lampa G.V.Santiago De Paucaray no no Lucanas MapSantiago De Pupuja yes no Azangaro G.V.Santo Tomas yes no Chumbivilcas G.V.Sara Sara no (1985) no Parinacochas MapSaurama no (1986) no Vilcas Huaman MapSicuani yes no Canas y Canchis G.V.Soras yes no Lucanas G.V.Soraya yes no Aymaraes G.V.Suyckutambo no no Canas y Canchis MapTalavera yes no Andahuaylas G.V.Tambobamba yes no Cotabambas G.V.Tapairihua no no Aymaraes G.V.Tinta yes no Canas y Canchis G.V.Tintay no no Aymaraes G.V.Tiquillaca yes no Paucarcolla G.V.Tirapata no no Azangaro MapToraya no no Aymaraes G.V.Tumay Huaraca no no Andahuaylas MapTupac Amaru no no Canas y Canchis G.V.Turpay no no Cotabambas G.V.Turpo no no Andahuaylas G.V.Umachiri no (1982) no Lampa G.V.Upahuacho no no Lucanas MapUranmarca no (1985) no Andahuaylas MapUrcos yes no Quispicanchis G.V.Velille yes no Chumbivilcas G.V.Vilavila yes no Lampa G.V.Vilcabamba no no Cotabambas G.V.Vilcas Huaman no no Vilcas Huaman G.V.Vilque yes no Paucarcolla G.V.Virundo no (1985) no Cotabambas MapVischongo yes no Vilcas Huaman G.V.Yanaca no no Aymaraes G.V.Yanaoca yes no Canas y Canchis G.V.

G.V. = Geografıa del Peru Virreinal (Bueno, 1951 [1764-1778]), Bachmann = Historia de lademarcacion polıtica del Peru (Bachmann, 1869).

A–16