is centralization a solution to the soft budget constraint problem

8/22/2019 Is Centralization a Solution to the Soft Budget Constraint Problem

1/39

Is Centralization a Solution to the Soft Budget Constraint Problem?

Avi Ben-Bassat1Hebrew University and IDI

Momi Dahan2

Hebrew University and IDI

Esteban F. Klor3

Hebrew University, IDI and CEPR

June 2011

1Department of Economics, The Hebrew University of Jerusalem, Mount Scopus, Jerusalem


2/39

Is Centralization a Solution to the Soft Budget Constraint Problem?

Abstract: to be added.

Keywords: Soft Budget Constraints, Centralization, Local Government, Bailout Program,

Summoned Committee.


3/39

1.Introduction

Whereas the last decades of the 20th

century brought about a large wave of fiscal

decentralization in countries around the world, we observe in the last few years a move in the

opposite direction. Currently, central governments tend to intervene and constraint the behavior

of local governments in developed and developing countries as well as countries in transition.

The book edited by Rodden et al. (2003), for example, shows that the central government

increasingly expanded its oversight of local governments in a large number of federations,

ranging from developed economies as the U.S. (Inman, 2003) and Canada (Bird and Tazzonyi,

2003), developing countries like Argentina (Webb, 2003) and India (McCarten, 2003), and

economies in transition like Hungary (Wetzel and Papp, 2003).

The original drive for decentralization posited that lower level governments tend to make

more efficient decisions over local public revenues and expenditures for a number of reasons.

First, due to inter-jurisdictional competition, local politicians have a strong incentive to provide

better services to attract more residents (Tiebout, 1956). In addition, local politicians are more

attune with the preferences of the local population and better informed of their needs and


4/39

behavior, whereby they shift part of the financial burden associated with local provision of

services to higher levels. That is, when the central government tends to bailout localities in

financial distress, localities that repeatedly deviate from a balanced budget constraints are able to

extract transfers from others, more financially responsible localities. Hence, a common-pool

practice behind the financing of local expenditures creates strong incentives for local officials to

overspend. This phenomenon is known in the literature as the soft-budget constraint problem

following the seminal work of Kornai (1979).1

The central government has a direct way of dealing with soft-budget constraint problems.

This includes developing an institutional setting that imposes the right incentives on local

governments, and includes a credible no-bailout policy. In most countries of the world, however,

central governments are unable to commit to a no-bailout policy for political reasons. In these

cases, central governments resort to hierarchical oversight and regulation to reign on localities

soft-budget constraints. These types of interventions, also called administrative subordination,

often consist on centrally appointing local politicians rather than electing them.

This paper studies whether administrative subordination is an effective tool to solve soft-


5/39

the central government. In particular, they consist of (i) the appointment of an accountant that

has to approve every expenditure and reports directly to the central government; (ii) the

imposition of a recovery program specifically tailored for each locality that specifies, among

other things, tax collections targets and number of employees that need to be fired; and (iii) the

dismantling of the mayor and members of the local council that are replaced by a summoned

board as the only authority of the locality. Since their inception, these programs were instituted

in more than half of Israels localities in order to curtail their deficits and expenditures.

Despite the widespread use of hierarchical oversight and administrative subordination in

Israel and elsewhere, it is not clear that these policies are effective in imposing hard budget

constraints on local governments. The case studies mentioned above report mixed results of

hierarchical oversight. These studies, however, do not present any systematic empirical evidence

on the effects of centralization on the fiscal performance of local governments. In this paper, we

focus on the widespread use of hierarchical oversight and administrative subordination in Israel

since 2004 to analyze their effects on localities fiscal performance and their provision of public

goods.

Our results suggest that the administrative subordination of local politicians has had only


6/39

characteristics, we observe that central intervention is only effective in Arab localities by

lowering their expenses and slightly increasing their tax collection.

We decompose the effects of intervention between the three different types of

administrative subordination enacted by the central government. The analysis shows that

whereas the appointment of an accountant that reports directly to the central government has a

significant effect on localities expenditures and tax collection, the effects of a recovery program

are nil. Hence, our analysis concludes that central intervention has a limited effect on localities

fiscal performance, and that limited intervention is desirable over more centralization of local

governments.

2. Institutional and Fiscal Background

Over the last two decades, the fiscal performance of Israels localities has been

characterized by repeated crises, which resulted in severe service interruptions both at the local

and national level. Usually, the central government intervened and solved the conflict by bailing-

out localities in financial difficulties. For example, the central government covered the localities


7/39

faced and tended to reward at the polls mayors that increased local expenditures and deficits

before the elections (Rosenberg, 1992).2

After the repeated bailouts and transfers during the 1990s, the central government reached

the conclusion that one of the main causes for the constant fiscal failure of localities was the

central governments behavior and the pernicious incentives that behavior created. In part to

solve the problem created by the soft budget constraints, and in part because of the economic

impact of the global economic crisis and the Second Palestinian uprising, the central government

substantially lowered its transfers to localities between 2001 and 2003. As a consequence, a large

number of localities suffered a severe economic crisis, whereby their debts and deficits ballooned

and the quality of services they provided deteriorated (Ben-Bassat and Dahan, 2009).

Unwilling to increase its transfers to localities and bail them out, the Israeli central

government decided in 2003 to implement several new measures regarding its dealings with local

governments. The main objective of these measures was to restore the budget discipline of local

governments. These measures consisted on different types of central intervention on local

governments to ensure that localities do not deviate from pre-determined fiscal objectives. Since


8/39

(iii) the dismantling of the mayor and members of the local council while imposing a summoned

board as the main authority of the locality.3

The level of intrusiveness of the central government, and the corresponding damage on

local democratic institutions, gradually increases as we move from an intervention of type (i) to

an intervention of type (iii). The first type of intervention, the appointment of an accountant that

reports directly to the Interior Ministry, is meant to provide the central government with direct

control over a localitys expenditures. The appointed accountant has to approve in advance every

expense of the locality, including the hiring of new employees. In some localities, the appointed

accountant (or a different functionary under the accountants responsibility) is also in charge of

collecting local tax revenues.

The second type of intervention consists of the formulation of a recovery program for the

locality under intervention. While in the past localities that received general grants from the

central government to cover their deficits were said to be in a recovery program, there were no

clear guidelines or objectives to be fulfilled under this program. This situation drastically

changed with the new program that the government approved in 2003 and started in 2004. In


9/39

locality, the Interior Ministry rewards localities that meet the recovery programs targets with

transfers that equal half of their short-term debt (50% as a grant and the remainder 50% as a

loan). Hence, the main change that occurred in 2004 is that the governments grants to localities

were now conditioned on the locality achieving pre-determined fiscal goals.

The final type of intervention is the imposition of a summoned board. This is the most

severe measure of administrative subordination taken to deal with the fiscal crisis of some

localities. While using it, the central government basically replaces by decree the localities

elected officials (the mayor and the members of the local council) by a summoned committee.

Once in place, all roles and responsibilities are transferred from the elected mayor to the

appointed committee. The committee then governs the locality up until the next local election,

which is usually three years after the appointment of the committee. In some cases, under the

discretion of the central government, a committee may be appointed to two terms of three years

each.

Arguably, the most important change introduced by the new plan of 2003 (and

implemented in 2004) was the introduction of clear rules determining the criteria used to


10/39

Accordingly, the Interior Ministry may appoint an accountant to localities with a current

deficit out of total revenue above 10 percent and a short-term debt above 15 percent of total

revenue. In addition, for the central government to be able to appoint an accountant it has to be

the case that the localities underperform with respect to their local tax collections. The exact

thresholds for underperforming regarding tax collections vary according the localities socio-

economic index.4

The fiscal situation of a locality has to be even worst for the Interior Ministry to be able to

impose on the locality a specially tailored recovery program. In particular, a program is

constructed for localities whose short-term debt is above 17.5 percent of its total revenue.

Recovery programs were also restricted to localities that were entitled to (unconditional)

equalization grants and with more than 4,000 residents.5

In addition to a recovery program, the

Interior Ministry may appoint an accountant to some localities if it considers that these localities

are not going to properly execute the requirements outlined in their recovery program. Indeed,

4Every locality in Israel is ranked by its socio-economic status according to a scale from 1 (the lowest

status) to 10 (the highest status). The rank is based on the localities residents standard of living (income


11/39

during the period at issue many localities are under both types of intervention: they have an

accountant appointed by the Interior Ministry and are under a recovery program.

A locality has to be even under a worst fiscal situation for a summoned board to be

imposed. Accordingly, summoned boards are imposed on localities whose current deficit is over

15 percent of their total revenue, whose short-term debt is over 30 percent of their total revenue,

and whose tax collection rates are below the thresholds specified in Table 1. In addition, the

ministry of interior must dismiss a localitys elected bodies and impose a summoned board when

its council has not approved the localitys budget within three months after the beginning of the

fiscal year.

3. Data

For the purposes of our analysis we use data on central intervention and on economic and

demographic characteristics of localities in Israel between 2000 and 2008. The data on central

interventions come directly from the Interior Ministry. While the data on economic and

demographic characteristics of the localities come from the Israeli Central Bureau of Statistics


12/39

until 2008.6

The table shows that the move from a regime with a lot of discretion to one with

clear and objective rules resulted in a substantial increase on interventions, especially on Arab

localities. For example, we see that in 2004, the first year of the new program, 60 out of 73

localities with an Arab majority were under intervention. Moreover, the Interior Ministry

appointed in 2004 an accountant in 23 out of 59 Arab localities under the recovery program. This

shows the commitment of the central government to implementing the reform intended to

generate a hard budget constraint on the localities, mostly at the cost of the administrative

subordination of the localities fiscal powers to the central government. It also highlights the lack

of confidence the central government had on the localities ability (or willingness) to reach the

goals of the recovery program by themselves, without a centrally appointed accountant

monitoring the localities expenditures.

The picture is somewhat similar in localities with a Jewish majority. As in Arab localities,

we see a widespread use of recovery programs, with over 40% of Jewish localities under this type

of intervention. Unlike Arab localities, however, the central government appointed an accountant

in addition to a recovery program only in five localities, thus showing more faith on the ability of


13/39

used. That is, almost all Arab localities intervened in 2004 continue to be intervened between

2005 and 2008, with 41 out of 73 Arab localities under intervention since 2004. We observe

more decay in the number of Jewish localities under intervention over time, with only 19

localities under intervention every year since the beginning of the new centralization program.

Importantly, however, the severity of the type of intervention increases over time, with a

substantial increase on the use of summoned boards since 2007, especially in Arab localities.

The number of summoned boards appointed in 2007-2008 is unprecedented, to the point that it is

larger than the number of appointed summoned boards in the previous five decades combined.

B. Localities Fiscal and Demographic Data

We study the effect of intervention on all available variables measuring the fiscal situation

of localities, their expenditures and their ability to collect revenues. These include different

measures of the localities debt, deficit and expenditures on services, as well as their efficiency in

terms of collecting local property taxes and water bills.7 By looking at all these measures we are

able to analyze the effectiveness of the intervention program in terms of the changes it causes on


14/39

Table 3 presents summary statistics for the main variables of interest. The table tabulates

the localities according to ethnicity, and to whether or not a locality is ever under intervention

between 2004 and 2008. The table shows separate summary statistics for the different four

groups of localities for the period before the beginning of the intervention program (2000 to

2003) and the period after the beginning of the intervention program (2004 to 2008). As a first

pass, the table plainly shows some of the differences between Jewish localities and Arab

localities. While we focus on the period before the implementation of the intervention program,

we observe that Arab localities revenue is significantly lower than that of Jewish localities, and

therefore their debt as a share of their revenue is substantially higher even if the average deficit

per capita is similar across Arab and Jewish localities. As a consequence, expenditures tend to be

lower in Arab localities compared to those of Jewish localities, even though Arab localities

receive more transfers from the central government. Finally, the data show that Arab localities

are less populated, have a lower socio-economic status and a lower percent of students that

complete their matriculation compared to Jewish localities.

Let us now focus on the comparison of localities within their ethnic group before the


15/39

environment, as shown by their lower share of the largest party in the local council, in

comparison to the rest of the localities.8

We now switch our attention to the years 2004 to 2008, the period after the

implementation of the intervention program. The table shows that Jewish and Arab localities

under intervention still have higher debts and deficit than localities not under intervention. We

also observe, however, a substantial change in localities expenditures. Accordingly, localities

under intervention now have significantly lower expenditures compared to the rest of the

localities. In addition, Arab localities under intervention collect higher rates of local taxes

compared to the rest of the Arab localities.

To obtain a better measure of the effects of intervention on Jewish and Arab localities we

need to look at the difference-in-differences of the average variables of interest. That is, we need

to calculate the differences for each outcome of interest between the averages before and after the

program of the differences between localities under intervention and the rest of the localities.

Such a calculation confirms our previous discussion. That is, we observe that (i) the deficit of

Jewish localities under intervention relative to the rest of the localities decreased by NIS 65 per


16/39

there is a substantial increase in transfers from the central government of NIS 66.5 per capita in

Arab localities.9

There are several caveats regarding the preliminary conclusions based on Table 3. First of

all, they only show unconditional changes on the outcomes of interest. We develop below a

multivariate regression approach to address this issue. Secondly, Table 3 makes evident that

treated and untreated localities are substantially different in terms of their observables before the

treatment. Therefore, it seems safe to assume that there are significant differences between

treated and untreated localities also in terms of their unobservable characteristics. This creates an

inference problem since central intervention is definitely correlated with the localities

characteristics.

To concentrate on a more homogenous group of localities, we restrict our sample to those

localities whose short-term debt as a share of their total revenue was in 2003 between 10% and

25%. That is, given that localities with short-term debt as a share of total revenue above 17.5%

are included in the recovery program, our range concentrates on localities within 7.5 percentage

points of this threshold, and compares those slightly below the threshold (which were not treated)


17/39

Table 4 replicates the analysis of Table 3 but focusing exclusively on this subsample of

localities. Table 4 leads us to slightly different conclusions than those reached based on Table 3.

While we still observe that debt and deficit of intervened localities increase (relative to the rest of

the localities) and their expenditures decrease (but only for Jewish localities), the magnitude of

the effect is substantially lower than that obtained in Table 3. It is plausible, therefore, that the

effects observed for the entire sample are spurious. Another possibility is that the threat of

intervention affects the behavior of localities not intervened but slightly above the intervention

threshold. Finally, it may be that comparing the before and after means of the variables of

interest masks important gradual but slow changes caused by the program. Figures 1 and 2

address the last possibility.

These figures depict, for every outcome of interest, the difference in means between those

localities intervened in 2004 2008 and the remainder localities. Figure 1 focuses on the

localities debt, deficit and expenditures. The figure clearly depicts the gradual worsening fiscal

situation of the treated localities during the years 2000-2003, before the beginning of the

program. The figure also shows the gradual but slow relative improvement of the fiscal situation


18/39

untreated localities. Our empirical framework below is especially designed to pay particular

attention to the gradual effect of the intervention program on the outcomes of interest.

4. Empirical Framework

Our empirical specification allows us to examine how central intervention affects the

performance of localities. The following linear regression model presents our preferred

specification to estimate relationship between central intervention and the localities

performance:

(1)

where Yit is the outcome variable of interest, a list of which appears in Tables 3 and 4;

(Intervention)i,t-1 is a dummy variable that equals one if locality i is under central intervention in

year t-1; (Intervention)i,t-1 * (Jewish Locality)i is a dummy variable that equals one for Jewish

localities under intervention in year t-1. Thus, our analysis allows for a different effect of the

Yi,t =1(Intervention)i,t1+2(Intervention)i,t1(Jewish Locality)i +

1(Years Intervention)i,t1 +2(Years Intervention)i,t1(Jewish Locality)i +Xi,t1+t+c +uit,


19/39

locality fixed effect; and thes, s and the vector are unknown parameters that need to be

estimated. Unobserved determinants of the localities performances are captured by the error

term, uijt.11

The goal of the proposed econometric specification on equation (1) is to identify the joint

impacts of the s and s. In particular, we are interested on the immediate effect of

intervention as well as on the protracted effect of intervention over time. Based on equation (1),

the immediate effect of intervention is equal to (1 + 1) for Arab localities, and to (1 + 2 + 1

+ 2) for Jewish localities. The gradual effect of intervention over time equals 1 for Arab

localities and (1 + 2) for Jewish localities.

5. Results

Table 5 presents the results of the estimations of equation (1), which show the effect of

central intervention on the localities main outcomes of interest. Before discussing the results,

we need to keep in mind that they are based on the entire sample of localities, and that the

estimation does not differentiate between different types of administrative subordination


20/39

capita, with an additional decrease of over NIS 100 per any additional year of intervention.

Intervention has also a significant effect on total expenditures and total salary payments, both for

Jewish and Arab localities. According to the estimates, central intervention causes a decrease of

over NIS 150 in a localitys total expenditure and of over NIS 100 in a localitys total salary

payments per capita. Moreover, every additional year under intervention brings about a decrease

of expenditures of slightly over NIS 100 per capita for an Arab locality and of NIS 84 per capita

for Jewish localities. In Arab localities, a third of the yearly expenditures decrease are due to a

decrease in total salary payments, whereas for Jewish localities the yearly decrease on salary

payments accounts for almost half of their decrease on total expenditures.12

In addition, we

observe that localities improve their collection of local property taxes. This is especially

pronounced in Arab localities. According to the estimates, an Arab locality increases its

collection of property taxes by 7.4 percent due to intervention, and every additional year of

intervention translates into an additional 2.8 percent increase in property taxes collected.13

Finally, we also observe that treated localities receive significantly more transfers from the

central government, which especially increase among Arab localities. Arab localities under


21/39

Summarizing, the results suggest that the administrative subordination of localities is

associated with a decrease in their expenditures, an increase in local tax collection, and an

increase in transfers from the central government. All of the above bring about a decrease of the

localities deficit. The apparent improvement on fiscal outcomes, however, does not immediately

translate on better public good provision. The percentage of students that finish their

matriculation is not affected by intervention, nor is the intervened localities net migration (see

Appendix Table 1).

While giving us a first view at the effects of central intervention, the estimates presented

in Table 5 may be spurious for the reasons already mentioned above. The sub-sections below

address these identification concerns.

5.1 Accounting for Localities Different Unobservable Characteristics

An important concern related to the estimation of the effect of central intervention on

localities performances is that treated and untreated localities are substantially different in terms

of their unobservable characteristics. The evidence presented in Table 3, which shows that


22/39

leading us to believe that they are also similar in terms of their unobservable characteristics that

vary over time.

The message that emerges from Table 6 is very different from the one obtained from

Table 5. Accordingly, once we compare the effects of intervention across similar localities, the

analysis suggests that the effects of administrative subordination are more limited than those

observed in Table 5. In particular, the results confirm only effects of intervention on Arab

localities. We observe that administrative subordination causes a reduction of total expenditures

and salary payments, and an increase on central transfers in Arab localities only. The decrease of

total expenditures due to intervention equals NIS 146 per capita, and almost all of it (NIS 124 per

capita) is caused by decreases in total salary payments. The increase of central transfers due to

intervention equals NIS 183 per capita. These effects are of similar magnitudes to those found in

Table 5. In addition, we observe that continued central intervention causes a significant increase

in Arab localities collection of property taxes and water bills (Appendix Table 2).

The results of Table 6 clearly establish that the actual central intervention of a locality

does not have a statistical significant effect on Jewish localities and has a limited effect on Arab


23/39

seem to significantly affect a localitys fiscal performance, it is possible that adopting a program

that includes a stick (in addition to a carrot) has a significant effect on localities that are not under

intervention. Unfortunately, the available data and institutional details of the intervention

program do not allow us to empirically test the validity of the possibility that the threat of

intervention affects a localitys performance.

5.2 Differentiating by Type of Intervention

The previous sub-section showed that central intervention has a significant effect on the

fiscal behavior of Arab localities. The analysis in the previous section, however, does not

distinguish between the three different forms of administrative subordination implemented by the

central government. This is the focus of the current subsection.

As already discussed in section 2, the Israeli government implemented three different

types of central intervention: (i) the appointment of an accountant that approves every

expenditure and reports directly to the central government; (ii) the imposition of a recovery

program specifically tailored for each locality; and (iii) the dismantling of the mayor and


24/39

interventions in equation (1).14

Every cell of Table 7 reports the total immediate effect of each

type of intervention or the total effect of each additional year of each type of intervention

separately for Jewish and Arab localities. Numbers in parentheses are thep-values from anF-test

on the significance of each type of intervention (or years of intervention).

The results in the table corroborate the conclusions obtained in Table 6; that is, central

intervention affects mostly total expenditures and total salary payments of Arab localities. Table

7 also shows that most of the effect is caused by the appointment of an accountant, which is the

mildest of the three types of interventions. Interestingly, the imposition of a recovery program

does not seem to have a significant effect on any of the outcomes of interest either for Jewish or

Arab localities.

Summarizing, this table shows that different types of administrative subordination have

different efficacy on localities fiscal performances. The imposition of a recovery program,

which includes a carrot and a stick and is supposed to incentivize local politicians into sound

fiscal policies, does not have a strong effect on localities performances. On the contrary, a less

intrusive program that consist of the appointment of an accountant that reports directly to the


25/39

6. Conclusions

To be added.

7. References

Ben-Bassat, Avi and Momi Dahan (2011). Social identity and voting behavior, Public Choice,

forthcoming.

Ben-Bassat, Avi and Momi Dahan (2009). The crises of localities in Israel: Efficiency versus

representation, in Ben-Bassat, A., and M. Dahan (eds.), The Political Economics of the

Municipalities in Israel, Jerusalem: The Israel Democracy Institute (in Hebrew), 15-91.

Bertrand, Marianne, Esther Duflo, and Sendhil Mullainathan (2004). How much should we trust

differences-in-differences estimates? Quarterly Journal of Economics, 119(1): 249-275.

Bird, Richard M. and Almos Tassonyi (2003). Constraining subnational fiscal behavior in

Canada: Different approaches, similar results? in Rodden, J., Eskeland, G.S., and Litvack,

J. (eds.), Fiscal Decentralization and the Challenge of Hard Budget Constraints,

Cambridge, Massachusetts: The MIT Press, 85-132.


26/39

Inman, Robert P. (2003). Transfers and bailouts: Enforcing local fiscal discipline with lessons

from the U.S. federalism, in Rodden, J., Eskeland, G.S., and Litvack, J. (eds.), Fiscal

Decentralization and the Challenge of Hard Budget Constraints, Cambridge,

Massachussetts: The MIT Press, 35-83.

Kornai, Janos (1979). Resource-constrained versus demand-constrained systems,

Econometrica, 47(4), 801-819.

Kornai, Janos, Eric Maskin, and Gerard Roland (2003). Understanding the soft budget

constraint,Journal of Economic Literature, 41(4): 1095-1136.

McCarten, William J. (2003). The challenge of fiscal discipline in the Indian States, in Rodden,

J., Eskeland, G.S., and Litvack, J. (eds.),Fiscal Decentralization and the Challenge of HardBudget Constraints, Cambridge, Massachusetts: The MIT Press, 249-286.

Pettersson-Lidbom, Per (2010). Dynamic commitment and the soft budget constraint: An

empirical test,American Economic Journal: Economic Policy, 2(3): 154-179.

Oates, Wallace E. (1972). Fiscal Federalism, New York: Harcourt.

Reingewertz, Yaniv (2010a). Do municipal amalgamations work? Evidence from


27/39

Sadan-Samet, A., 2009. The actual budget process in municipalities, in Ben-Bassat, A., and M.

Dahan (eds.), The Political Economics of the Municipalities in Israel, Jerusalem: The Israel

Democracy Institute (in Hebrew).

Tiebout, Charles (1956). A pure theory of local expenditures, The Journal of Political

Economy, 64(5), 416-424.

Webb, Stephen (2003). Argentina: Hardening the provincial budget constraint, in Rodden, J.,

Eskeland, G.S., and Litvack, J. (eds.), Fiscal Decentralization and the Challenge of Hard

Budget Constraints, Cambridge, Massachusetts: The MIT Press, 189-211.

Wetzel, Deborah and Anita Papp (2003). Strengthening hard budget constraints in Hungary, in

Rodden, J., Eskeland, G.S., and Litvack, J. (eds.), Fiscal Decentralization and the

Challenge of Hard Budget Constraints, Cambridge, Massachusetts: The MIT Press, 393-

427.


28/39

Figure 1: Fiscal Outcomes


29/39

Figure 2: Other Outcomes of Interest


30/39

Current

Deficit

Short Term

Debt(as a share of

Total

Revenue)

(as a share of

Total

Revenue)

1 - 4 5 - 7 8 - 10

Type of Intervention:

- Accountant Over 10% Over 15% 50% 55% 60%

- Recovery Program

Between

17.5% and

50%

- - -

- Summoned Board Over 15% Over 30% 40% 50% 60%

- The rates of tax collection specified in the table regarding the imposition of a summoned board refer to local property taxes. In

addition, a summoned board is imposed if the collection of water bills is below 70%.

- The central government can impose a summoned board also if the local budget has not been approved at the local council within

three months after the beginning of the fiscal year

(by Socio-Economic Index)

Local Tax Collection

TABLE 1

Criteria for Inclusion of Locality on Intervention Program

Notes: Information obtained from the Israeli Ministry of Interior.

- The rates of tax collection specified in the table regarding the imposition of an accountant refer to collection of water and local

property taxes.


31/39

Jewish Arab Jewish Arab Jewish Arab Jewish Arab Jewish Arab

Accountant 6 24 9 29 13 38 22 53 21 53

Recovery Program 54 59 54 53 39 31 32 41 17 33

Summoned Boards 0 0 1 3 2 7 8 17 8 19

Localities under intervention 55 60 54 55 42 48 37 62 28 57

Duration of Intervention:

- 1 year 55 60 7 9 6 6 6 2 7 1

- 2 years 51 53 17 17 15 7 13 4

- 3 years 37 41 12 15 10 8

- 4 years 29 41 12 11

- 5 years 21 41

2004 2005

Note: During the years 2004 - 2008 there were in Israel 120 Jewish localities and 73 Arab localities. Notice that some localities may have more than one type of intervention at any

given year.

2006 2007

TABLE 2

Number of Local Governments under Central Intervention (by Type of Intervention), 2004-2008

2008


32/39

Intervened

during

2004-2008

Never

Intervened

Intervened

during

2004-2008

Never

Intervened

Intervened

during

2004-2008

Never

Intervened

Intervened

during

2004-2008

Never

Intervened

Outcomes of Interest

- Deficit (NIS per capita) 587 215 563 230 381 75 461 62

(635) (532) (528) (402) (611) (296) (713) (168)

- Short Term Debt as a share of Total Revenue 0.259 0.144 0.322 0.217 0.280 0.105 0.470 0.124

(0.175) (0.202) (0.287) (0.181) (0.201) (0.098) (0.347) (0.097)

- Total Economic Expenditures (NIS per capita) 5,069 4,917 3,729 3,927 4,920 5,122 3,609 3,904

(1,759) (1,221) (819) (895) (1,406) (1,433) (610) (823)

- Total Salary Payments (NIS per capita) 1,979 1,975 1,766 1,760 1,796 1,906 1,594 1,634

(706) (613) (460) (511) (644) (587) (400) (441)

- Transfers from Central Government (NIS per capita) 1,343.0 910.4 1,434.4 1,613.0 1,084.0 538.4 1,331.3 1,264.8

(1,195) (779) (611) (692) (1,029) (581) (576) (631)

- Share of Local Property Taxes Collected 0.579 0.694 0.236 0.228 0.531 0.646 0.224 0.217

(0.178) (0.165) (0.158) (0.210) (0.176) (0.175) (0.133) (0.254)

- Percent of students that completed their matriculation 59.5 64.8 51.9 52.2 55.2 62.5 47.1 45.3

(10.9) (9.8) (11.4) (9.7) (15.4) (11.6) (11.4) (17.0)

Control Variables of Interest

- Population Size (in thousands) 35.121 44.204 11.040 9.115 38.926 53.072 12.895 11.543

(87.085) (69.818) (9.459) (8.567) (94.515) (75.990) (10.835) (12.473)

- Socio-Economic Status 5.504 6.554 2.762 2.400 5.497 6.432 2.825 2.029

(1.864) (1.989) (1.023) (1.446) (1.811) (1.936) (0.977) (1.445)

- Share of Largest Party in Municipal Council 0.321 0.415 0.263 0.405 0.317 0.337 0.254 0.446

(0.148) (0.412) (0.142) (0.268) (0.149) (0.140) (0.130) (0.256)

Notes: Data from ICBS.

TABLE 3

Summary Statistics on Localities' Outcomes of Interest - Entire Sample

Jewish Localities Arab Localities

2000 - 2003


2004 - 2008


33/39

Arab Localities

Intervened

during

2004-2008

Never

Intervened

Intervened

during

2004-2008

Never

Intervened

Intervened

during

2004-2008

Never

Intervened

Intervened

during

2004-2008

Never

Intervened

Outcomes of Interest

- Deficit (NIS per capita) 290 225 274 77 276 101 220 156

(294) (287) (248) (118) (412) (298) (360) (187)

- Short Term Debt as a share of Total Revenue 0.128 0.123 0.125 0.104 0.176 0.126 0.224 0.185

(0.067) (0.058) (0.075) (0.056) (0.112) (0.064) (0.121) (0.108)

- Total Expenditures (NIS per capita) 4,858 4,544 3,422 3,055 4,774 4,780 3,446 3,015

(1,463) (648) (731) (210) (1,212) (899) (510) (165)

- Total Salary Payments (NIS per capita) 1,752 1,769 1,633 1,452 1,648 1,796 1,482 1,205

(501) (385) (397) (168) (497) (487) (354) (184)

- Transfers from Central Government 1,170.7 795.5 1,400.2 1,150.1 867.2 522.8 1,214.9 807.4

(1,122) (587) (593) (383) (863) (525) (475) (194)

- Share of Local Property Taxes Collected 0.630 0.654 0.293 0.128 0.587 0.589 0.252 0.141

(0.200) (0.167) (0.234) (0.117) (0.202) (0.180) (0.205) (0.119)

- Percent of students that completed their matriculation 62.8 61.0 52.5 50.9 59.749 59.6 47.2 49.1

(11.8) (7.6) (11.3) (9.4) (14.871) (7.0) (12.5) (13.0)

Control Variables of Interest

- Population Size (in thousands) 50.405 41.692 10.857 19.550 56.161 50.323 12.226 24.570

(141.191) (47.971) (8.683) (14.722) (153.211) (51.288) (9.727) (16.993)

- Socio-Economic Status 5.619 6.233 2.536 1.000 5.771 6.108 2.643 1.000

(2.184) (1.511) (1.250) (0.000) (2.067) (1.226) (1.240) (0.000)

- Share of Largest Party in Municipal Council 0.312 0.483 0.311 0.399 0.318 0.301 0.298 0.300

(0.120) (0.722) (0.179) (0.167) (0.110) (0.064) (0.164) (0.035)

Notes: Data from ICBS.

2004 - 2008

TABLE 4

Summary Statistics on Localities' Outcomes of Interest - Only Localities within 7.5 percentage poins of Intervention Threshold


2000 - 2003

Jewish Localities


34/39

Deficit per

Capita

- Intervention -27.922 -0.0188 -0.0173 -72.565 *** 0.0451 0.0939 * 0.0739

- Intervention in Jewish Localities -13.6075 0.0223 0.0043 18.3008 -0.0291 -0.0069 -1.0618

- Years of Intervention -38.135 0.0175 -0.0275 *** -34.066 *** 0.0285 ** 0.2172 *** -1.8300 ***

- Years of Intervention in Jewish Localities -65.964 -0.0521 *** 0.0107 -6.5217 -0.0127 -0.0758 ** 1.9840 ***

0.2239 0.9735 0.0004 0.0000 0.0162 0.0000 0.0866

0.0019 0.1310 0.0130 0.0033 0.1020 0.0001 0.3445

0.0022 0.0038 0.0205 0.0049 0.1223 0.0141 0.7277

[87.81]

[116.8]

[49.28]

[58.93]

1,387

Short Term

Debt as Share

of TotalRevenue

Total

Expenditures

per Capita (inln)

Total Salary

Payments per

Capita

Local

Property

Taxes

Collected per

Capita (in ln)

[0.034] [1.386]

[0.054] [0.015] [24.15] [0.034] [1.216]

1,344 1,398 1,3101,396

[0.018] [33.87]

[0.013] [0.516][0.050]

[0.052]

Transfers from

central

government

per capita (in

ln)

[0.052]

Percent of

Students that

finished theirmatriculation

[0.067]

P-value on Effect of Years of Intervention on Jewish Localities

P-value on Effect of Intervention on Arab Localities

P-value on Effect of Intervention on Jewish Localities

Each column in each panel presents the results of a separate panel regression model. All regressions control for each locality number of residents (in logs), average salary for employees (from

previous year), socio-economic status, percentage of seats of larger party and number of parties (divided by number of legislators), as well as years and localities fixed effects. * indicatesstatistically significant at 10% level, ** indicates statistically significant at 5% level; *** indicates statistically significant at 1% level.

[0.034]

1,289

[0.018] [0.007] [14.35]

TABLE 5

The Effect of Central Intervention on Localities Outcomes of Interest - Entire Sample

Number of observations 1,258

[12.69]

[0.011] [0.528]

[0.018] [0.007]


35/39

- Intervention -83.678 0.0302 -0.0201 -80.235 ** -0.0234 0.0701 2.7475

- Intervention in Jewish Localities -95.99 -0.0299 0.0135 71.533 0.0334 0.0083 -5.1228

- Years of Intervention 26.281 -0.0030 -0.0231 -44.352 0.0610 * 0.0761 -2.2464 **

- Years of Intervention in Jewish Localities 39.817 -0.0024 0.0154 51.875 * -0.0749 *** -0.0668 3.2642 ***

0.4254 0.2324 0.0774 0.0071 0.4487 0.0237 0.8354

0.2236 0.8083 0.2269 0.9729 0.9297 0.4043 0.2755

0.3069 0.5995 0.4349 0.7675 0.6137 0.9267 0.2958

Deficit per

Capita

Short Term

Debt as Share

of TotalRevenue

Total

Expenditures

per Capita (inln)

338

[27.91]

[30.53]

[99]

[116]

[45.01]

[67.00]

Transfers from

central

government

per Capita (in

ln)

[0.084]

[0.036]

[0.031]

[39.90] [0.049]

Total Salary

Payments per

Capita

Local

Property

Taxes

Collected per

Capita (in ln)

Percent of

Students that

finished theirmatriculation

[0.035] [47.11]

[0.025] [0.031]

[0.044]

[3.015]

[0.012] [0.018] [1.125]

[0.024] [3.180]

[0.015] [0.017] [1.199]

347 347




Number of observations 331 345 340

[0.120]

[0.075]

[0.062]

325

Each column in each panel presents the results of a separate panel regression model. All regressions control for each locality number of residents (in logs), average salary for employees (from

previous year), socio-economic status, percentage of seats of larger party and number of parties (divided by number of legislators), as well as years and localities fixed effects. * indicates

statistically significant at 10% level, ** indicates statistically significant at 5% level; *** indicates statistically significant at 1% level.

TABLE 6

The Effect of Central Intervention on Localities Outcomes of Interest - Including only localities within 7.5 percentage points of Intervention Threshold


36/39

-99.292 0.0248 -0.0666 ** -182.5 ** 0.0989 ** -3.7921 0.0540

-2.9449 -0.0041 -0.0379 *** -101.34 *** 0.0646 5.1074 *** -0.0479

687.24 0.0401 -0.0170 587.1 -0.0588 -3.9212 0.4968 **

-544.429 -0.0139 -0.1419 *** -760.2 * 0.0681 3.5280 -0.3295

-17.3344 -0.0136 -0.0021 -14.5534 -0.0148 -1.1727 0.1244 **

45.133 0.0023 0.0119 29.655 0.0243 -4.1630 *** 0.1110

-68.74 -0.0054 -0.0008 12.5506 -0.0245 -2.1459 0.0580

33.036 -0.0024 -0.0029 -10.3234 -0.0192 1.2529 -0.0194

TABLE 7

(0.683)(0.792)(0.814)(0.465)

(0.581) (0.103) (0.606)

(0.855)(0.222)(0.562)

(0.424)(0.564)(0.899)(0.283)

(0.317) (0.784) (0.958) (0.614)

(0.746) (0.639) (0.0317)

(0.134)(0.001)(0.625)

(0.149) (0.023)

(0.41)(0.232)(0.527)(0.082)

(0.001)(0.001)(0.842)(0.946)

(0.13) (0.181) (0.475) (0.235)

(0.03) (0.224)

347 340 325

(0.185)(0.002)(0.286)

(0.473)

Number of observations 331 345 347

(0.769) (0.621) (0.933) (0.681)

Effect of Recovery Program on Arab Localities

Effect of Years of Recovery Program on Arab Localities

Effect of Recovery Program on Jewish Localities

Effect of Years of Recovery Program on Jewish Localities

(0.0004)(0.721)(0.151)

Effect of Accountant Intervention on Arab Localities

Effect of Years of Accountant Intervention on Arab Localities

Effect of Accountant Intervention on Jewish Localities

Effect of Years of Accountant Intervention on Jewish Localities

(0.286) (0.421) (0.011) (0.296)(0.038)

Transfers from

central

government

(in ln)

Deficit per

Capita

Short Term Debt

as Share of Total

Revenue

Total

Expenditures

(in ln)

Total Salary

Payments

Local

Property

Taxes

Collected (inln)

Percent of

Students with

Bagruyot

338

The Marginal Effect of Each Type of Intervention on Localities Outcomes of Interest - Including only localities within 7.5 percentage points of Intervention Threshold

Each column presents theresults of a separate panel regression model.All regressions control for each locality number of residents(in logs),averagesalary foremployees(fromprevious year), socio-

economic status, percentage of seats of larger party and number of parties (divided by number of legislators), as well as years and localities fixed effects. P-Value of F-Test of joint effect of interest

appears in parentheses below the respective coefficient. * indicates statistically significant at 10% level, ** indicates statistically significant at 5% level; *** indicates statistically significant at 1%

level.


37/39

- Intervention -0.0316 0.0419 -0.0111 -0.0301 -0.0525 ** -0.0335 0.0956 ** -0.0001

- Intervention in Jewish Localities 0.0432 -0.0086 -0.0114 0.0284 -0.0080 0.0245 -0.0581 0.0217

- Years of Intervention 0.0069 -0.0042 -0.0188 * -0.0003 -0.0107 -0.0053 0.0388 0.0155

- Years of Intervention in Jewish Localities -0.0611 *** -0.0683 *** -0.0144 -0.0132 -0.0332 * -0.0118 -0.1116 * -0.0198

0.5631 0.4083 0.0980 0.0843 0.0130 0.0867 0.0088 0.4930

0.0484 0.2339 0.0090 0.3938 0.0021 0.1995 0.7525 0.5983

0.0001 0.0251 0.0016 0.3444 0.0185 0.2369 0.2754 0.8787

Each column presents the results of a separate panel regression model. All regressions control for each locality number of residents (in logs), average salary for employees (from previous year), socio-economic

status, percentage of seats of larger party and number of parties (divided by number of legislators), as well as years and localities fixed effects. * indicates statistically significant at 10% level, ** indicates

statistically significant at 5% level; *** indicates statistically significant at 1% level.

Appendix Table 1

The Effect of Central Intervention on Localities Outcomes of Interest - Entire Sample

1,561 1,425 1,545

[0.024][0.011] [0.012] [0.017]

Number of observations 1,402 1,403 1,566 1,566 1,566




[0.016]

[0.021] [0.021] [0.014] [0.057]

[0.036]

[0.021] [0.032] [0.011] [0.014] [0.015] [0.015] [0.029]

[0.025]

[0.057] [0.046] [0.025] [0.023] [0.032] [0.032] [0.129]

[0.057] [0.049] [0.019] [0.022] [0.025] [0.028] [0.046]

Migration (in

ln)

Total Debt as

a Share of

Total Revenue

Debt per

Capita (in ln)

Expenditures

on Education

(in ln)

Expenditures

on Welfare (in

ln)

Salary

Payments on

Education (in

ln)

Salary

Payments on

Welfare

Income from

Water Bills (in

ln)


38/39

- Intervention 0.0219 0.1316 ** -0.0124 -0.0429 -0.0116 -0.0358 0.0826 -0.0732

- Intervention in Jewish Localities -0.0489 -0.1783 *** -0.0070 0.0752 0.0380 0.0514 0.0689 -0.0154

- Years of Intervention 0.0220 0.0191 0.0087 -0.0105 -0.0149 -0.0049 0.1300 * 0.0157

- Years of Intervention in Jewish Localities -0.0287 -0.0261 -0.0129 0.0199 0.0162 0.0051 -0.2052 * -0.0313

0.2359 0.0682 0.9126 0.0536 0.6341 0.3193 0.0438 0.2659

0.3551 0.2976 0.3611 0.3783 0.6010 0.5733 0.4598 0.1061

0.6710 0.8550 0.8090 0.7409 0.9609 0.9896 0.5510 0.7567

The Effect of Central Intervention on Localities Outcomes of Interest - Selected Localities

Appendix Table 2

Each column presents the results of a separate panel regression model.All regressions control for each locality number of residents (in logs), average salary for employees (from previous year), socio-economic

status, percentage of seats of larger party and number of parties (divided by number of legislators), as well as years and localities fixed effects. * indicates statistically significant at 10% level, ** indicates

statistically significant at 5% level; *** indicates statistically significant at 1% level.

347 313 339

[0.037][0.030] [0.031] [0.036]

Number of observations 329 329 347 347 347




[0.031]

[0.021] [0.037] [0.029] [0.107]

[0.054]

[0.022] [0.043] [0.033] [0.024] [0.039] [0.030] [0.069]

[0.045]

[0.035] [0.069] [0.047] [0.048] [0.067] [0.068] [0.151]

[0.031] [0.060] [0.038] [0.036] [0.056] [0.060] [0.074]

Migration (in

ln)

Total Debt as

a Share of

Total Revenue

Debt per

Capita (in ln)

Expenditures

on Education

(in ln)

Expenditures

on Welfare (in

ln)

Salary

Payments on

Education (in

ln)

Salary

Payments on

Welfare

Income from

Water Bills (in

ln)


39/39

0.04 147.593 -54.8015 -44.6 *** -108.3904 * -19.1144 34.4766 *** -0.0721 *

0.0453 79.5313 -67.4260 ** -6.33 -68.3997 *** -2.4729 -5.1873 -0.0087

-0.05 -2.36 438.9054 -289.1 77.5993 -21.5120 18.1188 0.1659

0.15 93.698 -657.6481 ** 319.1 ** -183.0100 ** 25.2186 * -356.3456 *** -0.6857 ***

0.0136 68.4621 56.0545 2.2226 53.3983**

1.1353 10.2065-0.0339

0.001 -14.685 76.3128 2.052 53.2268 4.0658 6.7767 -0.0074

-0.0125 6.3704 -7.9333 18.9570 2.6434 1.0418 7.6311 -0.1140

-0.0037 4.373 -13.3337 10.8219 1.7555 -1.6506 0.2213 -0.0157

The Marginal Effect of Each Type of Intervention on Localities Outcomes of Interest - Including only localities within 7.5 percentage points of Intervention Threshold

Each column presents theresults of a separate panel regression model. All regressions control for each locality number of residents (in logs), average salary for employees (from previous year), socio-economic status,

percentage of seats of larger party and number of parties (divided by number of legislators), as well as years and localities fixed effects. P-Value of F-Test of joint effect of interest appears in parentheses below the respectivecoefficient. * indicates statistically significant at 10% level, ** indicates statistically significant at 5% level; *** indicates statistically significant at 1% level.

313 348Number of observations 329 329 347 347 347 347

.(0.989) (0.836)

(0.82) (0.547) (0.107)

Effect of Years of Recovery Program on Jewish Localities

(0.82) (0.96) (0.622) (0.55) (0.922)

(0.325) (0.588) (0.902)

Effect of Recovery Program on Jewish Localities

(0.652) (0.965) (0.777) (0.517) (0.905)

(0.827) (0.4681) (0.554)

Effect of Years of Recovery Program on Arab Localities

(0.963) (0.888) (0.134) (0.839) (0.1)

(0.065) (0.001) (0.001)

Effect of Recovery Program on Arab Localities(0.698) (0.615) (0.189) (0.843) (0.048)

(0.102) (0.631) (0.374)

Effect of Years of Accountant Intervention on Jewish Localities

(0.108) (0.763) (0.0329) (0.036) (0.014)

(0.654) (0.582) (0.644)

Effect of Accountant Intervention on Jewish Localities

(0.463) (0.991) (0.178) (0.124) (0.164)

(0.118) (0.001) (0.064)

Effect of Years of Accountant Intervention on Arab Localities

(0.137) (0.401) (0.011) (0.683) (0.)

Effect of Accountant Intervention on Arab Localities

(0.34) (0.237) (0.352) (0.008) (0.074)

Migration (in

ln)

Total Debt as a

Share of Total

Revenue

Debt per Capita

(in ln)

Expenditures

on Education

(in ln)

Expenditures

on Welfare (in

ln)

Salary

Payments on

Education (in

ln)

Salary

Payments on

Welfare

Income from

Water Bills (in

ln)

TABLE A3

is centralization a solution to the soft budget constraint problem

Documents