religious discrimination among childcare workers ......(chandigarh and new delhi) among 183...

1

ReligiousDiscriminationamongChildcareWorkers:ExperimentalEvidencefromIndia

Utteeyo Dasgupta1, Subha Mani2, Prakarsh Singh3

February 2017

Abstract: Although many studies have documented the presence of discrimination, there is very little research identifying the sources of discrimination. To better understand the sources of discrimination, we introduce a novel allocation game that aims to identify two possible sources of religious discrimination: taste-based and statistical. We implement lab-in-the-field experiments in two cities (Chandigarh and New Delhi) among 183 childcare workers (known as Anganwadi workers) in India. We do not find evidence of either taste-based discrimination or statistical discrimination among these caregivers, and our results indicate that workers in our sample are primarily motivated by welfare based criteria in making their allocation decisions; religious identities do not influence choices of our subjects.

Keywords: lab-in-the-field experiment, discrimination, health, allocation game, India, religion JEL classification: C9, D3, I1, O1 Acknowledgements: We are grateful to Fordham University, Dean of the Faculty at Amherst College, and UNU-WIDER for funding support. Prakarsh Singh also thanks Bill & Melinda Gates Foundation for supporting travel to India, and Alka Yadav and Konso Mbakire for research assistance. We would also like to thank participants at the UNU-WIDER Discrimination Workshop. This project has been approved by the Institutional Review Boards at Fordham University and Amherst College. The views expressed in this paper do not reflect those of the funding agencies. All errors are our own. 1 Department of Economics, Wagner College and Center for International Policy Studies, Fordham University, New York. 2 Department of Economics and Center for International Policy Studies, Fordham University, New York; Population Studies Center, University of Pennsylvania, Philadelphia; IZA, Bonn. 3 Department of Economics, Amherst College, Amherst; Center for Advanced Study of India, University of Pennsylvania, Philadelphia; IZA, Bonn.

1

1. Introduction

Discrimination has a long-standing history in society and has manifested itself through a

myriad of identities such as religion, race, caste, gender, and wealth.1 Social identities can

influence choices and economic activities crucially (Akerlof and Kranton 2010; Gangadharan

et al. forthcoming), and sustained discrimination based on identity often leads to violent

conflicts and imposes huge economic costs on society. Recent civil wars such as those in

Syria, Iraq, Central African Republic, and Sri Lanka were all fought along sectarian lines.

Although there exist theoretical foundations characterizing the sources of

discrimination—taste-based and statistical—there is relatively little empirical work that

focuses on identifying these different sources of discrimination. Taste-based discrimination is

a result of intrinsic preferences for a particular group or observable characteristic such as

religion, caste, color, or gender, and typically practiced towards the minority group (Becker

1971). Statistical discrimination on the other hand occurs where identities such as race, caste,

or gender serve as a proxy for other less easily observable characteristics such as ability,

productivity, and delinquency (see Arrow 1972; Phelps 1972).

Traditionally, cross-sectional regressions have been used to identify discriminatory

behavior. These regressions control for observable characteristics such as education, parental

income, age, location, and other socioeconomic factors, and additionally include an indicator

for identity such as being Black in the USA, or being a woman, or belonging to a lower caste

in India (see Altonji and Blank 1999 for a review); a significant coefficient on the identity

variable is used to indicate the presence of discrimination. The drawback of such a regression

is that identity could also serve as a proxy for unobserved characteristics (such as risk

preference, competition, social preference, productivity, ability) and consequently suffers 1 See, for example, Hoff and Pandey (2004), List (2004), Hanna and Linden (2012), Bertrand and Mullainathan (2004), and Banerjee et al.

(2009).

2

from omitted variables bias. These regression-based approaches neither measure market

discrimination nor disentangle the relative importance of animus versus statistical

discrimination (Guryan and Charles, 2013). The resumé audit methodology, often used in

studying discrimination in labor and housing markets (Anderson et al. 2012; Banerjee et al.

2009; Bertrand and Mullainathan 2004; Siddique 2009) is an improvement over controlling

for observables. Here, fictitious resumés are sent out for the same job with identical

qualifications but differing in characteristics such as gender, ethnicity, and race. These

studies capture how the response rate from the supply-side depends upon applicant

characteristics and can often unearth evidence of taste-based discrimination. More recently,

the resume audit methodology has also been used to comment on statistical discrimination

(see Kaas and Manger 2012).

Since laboratory experiments facilitate greater control over more difficult to observe

information, psychologists, and more recently, economists, have used them to elicit

discriminatory behavior. Laboratory experiments in psychology, which are almost

exclusively based on hypothetical incentives for choices, typically find a substantial in-group

bias and distinct favoritism towards groups designated as superior, or groups sharing the

same attitude/preferences as the decision-maker (Vaughn, Tajfel and Williams (1981), Billig

and Tajfel (1973) , Diehl (1988), Devine (1989), Turner and Brown (1978), Klein and Azzazi

(2001). This experimental literature also indicates that subtle cues can influence how people

process information where implicit associations seem to be prevalent and powerful. In

contrast to the psychology experiments, economics experiments on discrimination primarily

have monetary incentives for choice consequences. Some of the early economic experiments

find that subjects with randomly assigned and laboratory created status groups behave

differently compared to those without such status (Ball, Eckel, Grossman & Zame (2001),

Ball and Eckel (1996), Fershtman and Gneezy 2001). However, Lane (2015) in a recent

3

review and meta analysis comprising 77 experimental studies, points out that despite an

increase in the number of studies that examine discrimination using lab experiments, results

indicate only a moderate tendency towards out-group discrimination. In fact he notes that the

majority of results are found to be null results across the literature. Lane further clarifies that

the strength of discrimination against out-group members, if present, depends crucially on the

group identity being used in the experiments – it is stronger when identity is artificially

induced in the laboratory as compared to cases where subjects are divided by existing

ethnicity or nationality dimensions. Additionally, Berge et al. (2015), and Habyarimana et al.

(2007) suggest that in their lab-in-the-field settings, a lack of discriminatory behavior along

ethnic lines can stem from mechanisms different from that due to commonly attributed social

biases.

However, in all these experiments there are very few experiments that attempt to

distinguish between statistical and taste-based sources of discrimination by design (Lane

2015). Our focus is to contribute to this growing literature by designing an experiment that

attempts to separate out statistical discrimination from taste based discrimination in the

context of religious majority-minority groups in India. Islam is the second-largest religion in

Hindu-majority India, making up 14.9% of the country's population with about 180 million

adherents (2011 census). Majority and minority religious groups have been shown to play

important roles in economic choices in the Indian sub-continent (Gupta et al. 2016) and hence

studying discrimination along religious lines remains important for India. In this paper we

specifically test for discrimination against Muslim children by Hindu childcare workers (also

known as Anganwadi workers). Anganwadi workers in India are in an important position to

influence health and learning outcomes for children under six years of age. Hence evaluating

the presence, and sources of discrimination has important policy implications for over 3

million pre-school age children in India. In this context, we seek to separate out

4

discrimination that arises from preferences for religion only (taste-based discrimination) from

discrimination that comes from preferences for an unobserved characteristic for which

religion only serves as a proxy (statistical discrimination). We design a novel allocation game

where each Anganwadi worker decides how to allocate their endowments between mothers

of two randomly chosen children under three different information treatments. Findings from

choices under the three treatments provide no evidence of taste-based discrimination or

statistical discrimination among the child caregivers. We follow up with a complementary set

of experimental games to find that Anganwadi workers’ sharing decisions are instead

influenced by the poverty level and welfare of the recipients.

Our paper contributes to the discrimination literature in important ways. First, we

address Lane’s (2015) concern about the absence of experiments distinguishing the two types

of discrimination. We specifically design an experiment to identify and distinguish between

taste-based and statistical sources of religious discrimination. Second, we implement our

experiment among actual caregivers in the public health sector in India allowing us to

provide greater external validity, in contrast to typical framed experiments run among student

subjects.2 Third, we introduce a new allocation game to study discrimination where the cost

to the allocator is fixed, allowing us to completely control over selfish payoff maximizing

choices (in contrast to existing allocation games used by Fershtman and Gneezy 2001; List

2004). The game allows us to focus exclusively on altruistic preferences (or the lack of them)

towards different groups where the allocations provide a direct measure of any existing

preference bias. Fourth, common sense view suggests that childcare workers are not solely

motivated by financial incentives. We examine that by comparing their decisions in

incentivized and non-incentivized contexts. One of the limitations of the experiment is that it

2 Framed allocation games involving student subjects have allowed researchers to elicit choices and decisions under very different

environments such as health care, organ donation, charity, etc. (Kass et al. 2013; Leonard et al. 2013).

5

cannot distinguish between the various sources of statistical-based discrimination, which can

result over time from taste-based discrimination initially (Neilson and Ying, 2016).

Section 2 provides an overview of the context. Section 3 describes the experiment and

provides an analytical sketch of worker behavior. Section 4 discusses the results and

concluding remarks follow in Section 5.

2. Scope of Religious Discrimination

Health and nutrition outcomes matter not only for lowering child mortality but it is well

documented that poor nutrition under 5 years is strongly related to adult height and affects

completed grades of schooling, grade progression, test scores, and adult income (Behrman et

al. 2009; Maluccio et al. 2009; Mani 2012; Stein et al. 2003, 2008; Victora et al. 2008). ICDS

is the largest Early Childhood Development (ECD) program in the world, encompassing 1.3

million day care centers where child caregivers (Anganwadi workers) are responsible for

improving children’s nutritional status as well as pre-school readiness in India. Consequently,

Anganwadi workers in India are in an important position to influence health and learning

outcomes for children under six years of age.

These workers are responsible for distributing the government-funded mid-day meals aimed

towards improving nutrition among children. The workers can improve child health through

two main channels: first, through the distribution of mid-day meals and second, through

nutrition counselling to mothers. Both these tasks involve time and effort on which the

Anganwadi worker is constrained. A World Bank report by Gragnolati et al. (2005) found

both to be limited, with children not getting adequate nutrition in centers and almost no

effective communication between workers and mothers. Singh (2015) provides further

evidence on the leakages and inefficiencies in the ICDS program in India. In this paper, we

focus on constraints on monetary resources, and in particular evaluate how workers allocate

6

their fixed monetary resources between children of two different religions (Hindu and

Muslim).

In a seminal paper, Bhalotra et al. (2010) document that Muslim children in India face

significantly lower mortality risks in India as compared to Hindu children, despite being from

more socio-economically disadvantaged backgrounds and receiving less antenatal care.

However, Brainerd and Menon (2015) find that a corresponding Muslim advantage in health

outcomes exists only during infancy. Children from Hindu households become significantly

taller and heavier than Muslim children between the ages of 1 and 5 years across the Indian

subcontinent. The reversal of health fortunes that the authors find does not go away despite

controlling for maternal health, education, age at marriage, disease environment, birth order,

sanitation, prenatal care, breastfeeding, or household assets. Thus, even though demand-side

factors cannot explain the Hindu advantage post infancy, supply-side factors such as the

quantity and quality of child care centers, and the effort of the child caregivers might be

responsible for the reversal of health between Muslim and Hindu children.

This may especially be possible as the responsibility for child care and nutrition

intake shifts (during the day) from the mother to the Anganwadi worker. In fact when we

used health data collected in this setting in 2010, to run a simple regression of malnutrition

status on child observables (age, gender) and mother’s characteristics (education, income,

assets, occupation), we are only able to explain 2 percent of the variation in malnutrition.

Adding Anganwadi fixed effects increases the R-squared to 10 percent, suggesting that the

day care environment may be a significant correlate in determining the health status of a

child.3 This suggests the importance of Anganwadi workers in influencing the health status,

and in particular the potential detrimental effect on health if they are not conscientious

3 These results are available from the authors upon request.

7

enough or strategically choose whom to help and whom to ignore. For example, Anganwadi

workers might not always allocate greater effort towards those who are more malnourished,

but instead exert greater effort towards supporting religious in-group members opening up

the possibility of religious discrimination (Singh and Mitra forthcoming).4

Our interest here is to evaluate and elicit whether workers have preferences for

members of the same religious group. If they do, it might provide a supply side explanation

to unequal health outcomes among children of the two religious groups. Moreover, religious

background could also serve as a proxy for other unobservables such as child’s nutritional

status or socioeconomic background that may induce workers to change their own behavior

towards the two groups inducing statistical discrimination. In the paper, we distinguish

between the two sources of discrimination and the games are designed to elicit both. We

focus exclusively on Anganwadi workers who are Hindu, especially since almost all (97.8%)

of the health care workers in our sample are Hindus. We are not in a position to evaluate

discriminatory attitudes of Muslim health care workers towards Hindu children due to the

lack of observations on Muslim workers in our sample.

3. Experiment

3.1 The Discrimination Game

We designed an allocation game (the Discrimination Game) where each health care

worker5 (Anganwadi worker) allocated a fixed sum of money between mothers of two

4 In Chandigarh, 80.8 per cent of the population is Hindu and 4.9% is Muslim. In Delhi, 81.2% of the population is Hindu and 12.9% is

Muslim (Census of India 2011).

5 We use the terms Anganwadi worker, child caregiver, and health worker interchangeably throughout the paper.

8

randomly chosen pre-school age children.6 Each worker’s allocation decision was observed

under the following three treatments: (1) information and no-identity (INFO-NOID), (2)

information and identity (INFO-ID), (3) no-information and identity (NOINFO-ID).7 In

each treatment, the worker received Rs. 120, to divide between two mothers.8 She had to

choose from one of the four allocation options listed in Table 1. Since we wanted to elicit

workers’ distributional preferences for children with different religious identities, we

purposely kept the amount the worker can allocate to herself fixed in all choices to control for

selfish payoff-maximizing preferences across the treatments and focus exclusively on

distribution attitudes towards others.9 Subject instructions are provided in Appendix A2.

In the INFO-NOID treatment, the worker had no information on the child’s religious

identity. She was only told that both children suffer from grade 2 malnutrition or moderate

levels of undernutrition as measured by weight-for-age z-scores (see instructions for Task 1,

Appendix A2).10 Observed choices in this treatment allowed us to rule out any irrational

behavior among our subjects when they are not provided any information on the recipients’

religious identities. We should expect the modal choice to be the equal split.

In the INFO-ID treatment, the worker was provided with the names of the two

children where child 1 had a Muslim name and child 2 had a Hindu name. Further, she was

provided the information that both children suffer from moderate levels of undernutrition as

6 Lane (2015) in his meta analysis notes that the extent of discrimination against the out-group is strongest when the decision making

subject is asked to allocate payoffs between inactive players belonging to the in-group and out-group, it is stronger than in any other

decision-making context.

7 Fershtman and Gneezy (2001) and List (2004) both use a dictator game where the amount transferred by player A (dictator) to player B

(receiver), if affected by player B’s background characteristic, is a strong indication of taste-based discrimination.

8 Rs. 120 is equivalent to 2 USD at the prevailing exchange rate at the time.

9 This is in contrast to other papers that also allow the dictator to keep the money for themselves, allowing them to exploit their dictator

position (Fershtman and Gneezy 2001; List 2004).

10 In the context of the weight-for-age z-score measure used by Anganwadis, moderate malnutrition is defined as the child being between 2

and 3 standard deviations below the WHO specified mean for the reference population of the same age and sex. In Chandigarh, about 41 per

cent of the children enrolled in these centers were malnourished in July 2014 (Singh and Masters 2016).

9

measured by weight-for-age z-scores (see instructions for Task 2, Appendix A2). Since both

children suffer from equal degrees of malnutrition, unequal allocations in favor of the Hindu

(or the Muslim) child would indicate taste-based discrimination. 11 It is important to note here

that to minimize experimenter demand effects, we avoided providing explicit information on

religious identity. So instead of stating “mother of Hindu child” or “mother of Muslim child”,

we provided the child’s name. Here we made use of the fact that Muslim and Hindu children

generally have easily distinguishable names (see for example Banerjee et al. (2009) who

suggest that, “Muslim names were universally recognized as such”). Additionally, we

provided the decision maker with other information such as the health status of the child to

obfuscate experiment demand effects further (see discussion in Berge et al. 2015). In March,

2017 we followed up with a phone survey to verify whether our participants can actually

identify the religion and the gender of the children based on the names we had used. We

found that 100% of our sample that could be surveyed on the phone could correctly identify

the religion and the gender correctly.12 Finally, note that the main results reported in Table 4

continue to hold for the sample of women who participated in our phone survey (see

Appendix Table A2).

In the NOINFO-ID treatment, there was no information on the child’s nutritional

status. The workers only received information about the child’s name. In this treatment,

unequal allocation in favor of the Hindu (or the Muslim) child would indicate the presence of

11 In the lab-in-the-field experiments conducted in Chandigarh, workers were making choices over actual Hindu and Muslim children living

in another part of the city.

12 We thank Daniel Houser for this useful suggestion. We targeted all 75 Anganwadi workers who participated in our incentivized

experiments in Chandigarh via phone and were able to reach out 90.67% of these women. In New Delhi, we targeted 99 of the 108 workers

for whom contact numbers were available and were able to reach out 92% of these women. We find the response rates to be similar across

treatments in both Chandigarh and New Delhi. In Chandigarh, 89% of the workers in the INFOID treatment and 92% of the workers in the

NOINFO-ID and INFO-NOID treatments participated in the phone surveys. This difference is not significantly different (p-value = 0.72). In

New Delhi, 93% of the workers in the INFOID treatment and 91% of the workers in the NOINFO-ID and INFO-NOID treatments

participated in the phone surveys. This difference is also not significantly different (p-value = 0.71).

10

discrimination that can accrue to taste or statistical sources (see instructions for Task 3,

Appendix A2). We subtract the differences in allocations between the recipient mothers of

the two children in INFO-ID treatment from the differences in allocations between the two

children in the NOINFO-ID treatment to disentangle the presence (if any) of statistical

discrimination from taste-based discrimination.

Table 1: Choices in the Discrimination Game

OPTIONS A I want to keep Rs. 40 for myself

I want to give Rs. 40 to the mother of child 1 I want to give Rs. 40 to the mother of child 2

B I want to keep Rs. 40 for myself I want to give Rs. 0 to the mother of child 1 I want to give Rs. 80 to the mother of child 2

C I want to keep Rs. 40 for myself I want to give Rs. 80 to the mother of child 1 I want to give Rs. 0 to the mother of child 2

D I want to keep Rs. 40 for myself I want to give Rs. ……...(write amount here) to the mother of child 1 I want to give Rs. ……..(write amount here) to the mother of child 2 [Note: the three amounts must add up to Rs. 120]

3.2 Religious preference and behavior

We provide below a simple model to explain religious preference and plausible

behavior in our experiment design where we modify the basic setup in Berge et al. (2015) and

Cappalen et al. (2007) for our specific purposes. The allocator i, has to divide a fixed sum X

between an in-group member and a member of the out-group. Suppose he proposes an

amount xIN for the in-group member and xOUT =X – xIN for the out-group member. In our

specific context in-group members share the same religion, while out-group members do not.

We assume that the allocator i has a simple utility function of the following

form:

11

!!(!!") = !! . !!" − !! .(!!" −!!)!

2!

where !! ≥ 0 is the parameter of religious bias, !! > 0 is the parameter for fairness, and m is

the amount the allocator believes to be a norm dependent fair share in the absence of

information on religious identity. Note that our utility function differs from a typical utility

function in that there is no own payoff maximizing option. This is in line with our experiment

design where the allocator’s own payoffs are always fixed. Hence the utility comes from in-

group bias and preference for norm-dependent fairness. The norm dependent fairness can

depend on the context. For example the allocator might believe that a poor should get more

than the rich, or females should get more than males, etc. Notice then for a given value of !,

the allocator’s utility increases with an increase in the proposed share !!", and additionally,

deviations from m affect his/her utility negatively.

Given an interior solution, the optimal proposal for xIN would be

!!"∗ = !! + !! .!!

Hence the optimal proposed share depends on the norm dependent fair share along with the

sensitivity towards sharing with in-group members. Further, when the allocator has no

preference for in-group members, i.e., ! = 0, the allocator allocates based on his/her norm

dependent fair share independent of any religious favoritism. Similarly, if !! → ∞ ,

!! . !! → 0, and hence preference for socially desirably fairness norms become more

important than any in-group bias in deciding allocation amounts. In that case, even though

religious discrimination tendencies are absent, allocation outcomes can be different than an

equal split allocation based on underlying social attitudes and norms towards other social

parameters.

12

3.3 Design and protocol

Our experiment sessions were conducted among Anganwadi workers in Chandigarh and New

Delhi in India. In the baseline INFO-NOID treatment, we randomly chose two mothers of

pre-school age children (from Anganwadis in another geographical block) to receive the

payouts chosen by the worker in the experiment. In INFO-ID treatment and NOINFO-ID

treatments, the religious identities of the recipient mothers were disclosed by providing the

names of their children. In all treatments we chose names of males to control for any gender

preference. The recipient mothers in all treatments were chosen from a different geographical

zone to minimize the possibility of strategic interactions and/or familiarity bias.

We chose to randomly assign subjects either to a single treatment session consisting

of only treatment 2 (the information and identity treatment), or a session consisting of

treatments 1 and 3 (the information and no-identity treatment and no-information and identity

treatment).13 While we would have ideally liked to randomize subjects into one and only one

of the three treatments, we were constrained by sample size. We made the choice of assigning

treatment 1 with treatment 3 because treatment 1 was really a game to check for rational

choice since it did not include information on our main variable of interest, the religious

identity. We further reasoned that putting treatment 1 and 2 together might make subjects

more conscious of our ultimate interest as they participate in the two treatments sequentially.

This was not the case if we put treatments 1 and 3 together since they appear as eliciting

13 The power of an experimental design gives, for a given effect size and statistical significance level, the probability that we will be able to

reject the hypothesis of no treatment effects when it is false. In Figure A1 of the Appendix we show the minimum sample size required to

detect a medium effect size of 0.45 standard deviation at the 5 percent significance level. Figure A1 in the Appendix shows that we need a

total of approximately 150 workers (approximately 75 workers in each treatment) to secure an 80 percent chance of rejecting the null of zero

treatment effects when it is false. Note that our power calculations are based on pooling the two samples from Chandigarh and New Delhi.

Also note that the main results in the paper hold for the pooled sample and are available from the authors.

13

distribution choices under two different environments, one conditional on health status and

the other conditional on religious identities.

Chandigarh Sessions: The Anganwadi workers were paid a show-up fee of Rs. 50 for

participating in the experiment. The invitations were sent to all possible Anganwadi workers

employed in the ICDS’ Block 1 slum areas in Chandigarh. In total, over 90 per cent of the

invited workers participated. If a subject participated in the multiple treatment session we

randomly chose one of the treatments for additional payments to avoid wealth effects.

Average payout per subject in Chandigarh is Rs 120.

New Delhi Sessions: These were non-incentivized sessions. A comparison of results from the

incentivized and non-incentivized sessions allows us to comment on the substitutability of the

two design choices (Coppola 2014; Dohmen et al. 2011; Gneezy and Rustichini 2000).

Additionally, it provides insights into the role of monetary incentives among health care

workers. All session-related information is reported in Table 2.

Table 2: Experiment sessions in Chandigarh and New Delhi

Chandigarh New Delhi

Number of Anganwadi workers 75 108 Sample size in treatment 1 and 3 37 47 Sample size in treatment 2 38 61 Number of sessions 2 2 Show-up fee 50 None Discrimination Game Yes Yes Average payout per subject 120 0 Month of experiment May, 2015 August, 2015

4. Results

4.1 Main findings

14

Subject randomization into different treatments was successful. We find no significant

difference in the family background characteristics between Anganwadi workers assigned to

each of the treatments in Chandigarh as well as in New Delhi, verifying balance among

subject characteristics in each of our treatments (see Column 3, Table 3).

We categorize subjects’ behavior in the experiment into three groups: (1) equal split: subject

chooses to give equally to both mothers, (2) pro-Muslim: when the subject chooses to give

more to the mother of the Muslim child, and (3) pro-Hindu: when the subject chooses to

give more to the mother of the Hindu child. Note that by design, religious identity was not

available to the subjects for our baseline treatment (INFO-NOID). They were asked to choose

between giving more to child 1 or child 2, or an equal split. Figure 1 summarizes the average

choices made in the experiment. More than 80 per cent of the subjects chose to allocate

equally to both mothers in all three treatments.

Table 3: Differences in baseline characteristics by treatment and location

Mean [s.d.]

Treatment 2 (1)

Mean [s.d.]

Treatments 1 and 3

(2)

P-value from t-test on

difference between

Column 1 & 2 (3)

Panel A: Chandigarh sample Age in years 41.09

[9.09] 40.60 [7.59]

0.81

Years of experience 11.88 [8.82]

10.43 [9.13]

0.50

Education (= 1 have 12 or more grades of schooling, 0 otherwise)

0.54 [0.50]

0.62 [0.49]

0.52

Panel B: New Delhi sample Age in years 42.86

[8.35] 45.15 [8.60]

0.17

Education (in years) 9.75 [3.73]

10.51 [2.53]

0.23

Husband’s education (in years) 10.80 [4.55]

11.42 [4.32]

0.47

Household size 5.34 [1.96]

4.81 [1.86]

0.15

Number of children 2.34 [0.89]

2.55 [1.23]

0.31

15

Notes: The same subject makes decisions in treatments 1 and 3.

In the INFO-NOID treatment, where there was no information on religious identities of the

recipients, we expect the workers to allocate their resources equally between the two mothers

(barring any irrational choices or order effects). In the INFO-ID treatment, when religious

identity was provided and the two children were described to be moderately malnourished,

the presence of taste-based discrimination would suggest a choice of unequal allocation

among the two mothers. In the NOINFO-ID treatment, where only information on religious

identity was provided, but information on the children’s nutritional status was not revealed,

taste-based as well as statistical discrimination would lead to unequal allocation for the two

mothers. We set up the three hypotheses accordingly on rationality, taste-based

discrimination, and statistical discrimination in Table 4. Our findings in Table 4 establish the

following. First, with the information on the health status of the children and no other

information, the workers equally divided resources between both mothers - H1 cannot be

rejected. Second, in the presence of information on religious identity, subjects did not

discriminate on the basis of religious identity - H2 cannot be rejected. Third, there was no

evidence of statistical discrimination - H3 cannot be rejected.

Table 4: Tests of Taste and Statistical Discrimination

Hypotheses

Chandigarh sample

Decision*

New Delhi sample

Decision*

Differences in observed outcomes in

the New Delhi

sample and the

Chandigarh sample

Decision*

Choices are Rational without information on

H1: Proportion of subjects allocating more to Child 2 = Proportion of subjects allocating more to Child 1 in INFO-NOID treatment.

Fail to reject (0.18)



Household monthly income in Rupees 20606.56 [14726.71]

18276.6 [11357.25]

0.37

16

religion.

There is no evidence of taste-based discrimination.

H2: Proportion of subjects allocating more to Hindu child= Proportion of subjects allocating more to Muslim child in INFO-ID treatment.




There is no evidence of statistical discrimination.

H3: The difference in proportion of subjects allocating more to Hindu child and proportion of subjects allocating more to Muslim child in NOINFO-ID = The difference in proportion of subjects allocating more to Hindu child and proportion of subjects allocating more to Muslim child in INFO-ID




Notes: p-values are reported within brackets.

17

Figure 1: Distribution of Choices in the Three Treatments: Chandigarh and New Delhi

86.49

10.81

2.70

020

4060

80Pe

rcen

tage

of A

ngan

wadi

wor

kers

INFO-NOID: Chandigarh

Equal split Preference for child 2Preference for child 1

89.47

2.637.89

020

4060

8010

0Pe

rcen

tage

of A

ngan

wadi

wor

kers

INFO-ID: Chandigarh

Equal split Pro-HinduPro-Muslim

81.08

5.41

13.51

020

4060

80Pe

rcen

tage

of A

ngan

wadi

wor

kers

NOINFO-ID: Chandigarh


89.36

8.512.13

020

4060

8010

0Pe

rcen

tage

of A

ngan

wadi

wor

kers

INFO-NOID: New Delhi

Equal split Preference for child 2Preference for child 1

88.33

6.67 5.00

020

4060

80Pe

rcen

tage

of A

ngan

wadi

wor

kers

INFO-ID: New Delhi


80.85

8.51 10.64

020

4060

80Pe

rcen

tage

of A

ngan

wadi

wor

kers

NOINFO-ID: New Delhi


18

Results from our experiment suggest that there is no evidence for taste-based or statistical-based

discrimination in our sample. Additionally, when we compare observed outcomes from the incentivized

experiments in Chandigarh with the non-incentivized experiments in New Delhi, we find no difference in

our conclusions for each of the hypotheses (See p-value of differences in Column 3, Table 4). Weak

evidence or absence of it for discriminatory preferences however might arise from three different sources

in our experiment. First, it might be due to low statistical power. Second, there might be a lack of saliency

across treatment conditions leading subjects to choose the same in all treatments. Third, experimenter

demand/scrutiny effects might attenuate underlying discriminatory tendencies. We try to mitigate some of

these concerns next.

4.2 Statistical power

As described in footnote 12, our study is already designed to detect medium effect sizes. However, to rule

out additional sample size related concerns, we report results from the bootstrap technique outlined in

Moffat (2016).14 We compute bootstrap standard errors using 999 replications for a two-tailed test at the

5 percent significance level. In Table 5, we replicate our main findings with bootstrap p-values. The p-

values reported in Table 5 are consistent with our main findings of no religious discrimination (either

taste-based or statistical).

Table 5: Taste vs. statistical discrimination (bootstrap standard errors)

Hypothesis

Decision

Chandigarh

Decision

New Delhi

H1: Choices are Rational without information on religion Fail to reject

(p = 0.15)

Fail to reject

(p = 0.12)

14 Adèr et al. (2008) points out: “When the sample size is insufficient for straightforward statistical inference, if the underlying distribution is

well-known, bootstrapping provides a way to account for the distortions caused by the specific sample that may not be fully representative of the

population.”

19

H2: There is no evidence of taste-based discrimination Fail to reject

(p = 0.29)

Fail to reject

(p = 0.71)

H3: There is no evidence of statistical discrimination Fail to reject

(p = 0.74)

Fail to reject

(p = 0.62)

4.3 Treatment saliency

Since we find no evidence of religious discrimination (taste or statistical) in our sample, it is

possible that workers are indifferent to the information provided in the three treatments and

consequently choose the equal split rational outcome in each treatment. However, this would

suggest that we should observe similar equal split choices in other informational conditions as

well. To rule out saliency concerns, we ran an additional experiment with a different sample of

Anganwadi workers in Chandigarh in January, 2016.

Here we test whether workers remained indifferent to other types of information. As before,

workers were asked to allocate a fixed amount between two mothers with different

characteristics: (a) mother of a moderately malnourished and mother of a healthy child, (b) a

poor and a rich mother, (c) an illiterate and a literate mother, (d) a non-working and a working

mother, (e) a mother with low household assets and one with high assets, (f) a mother with a

physically challenged child and a mother with a child without any disability, and finally (g) a

mother of a boy and a mother of a girl. Forty-nine workers participated in the experiment. Only

situation (a) was incentivized and the remainder were thought experiments along the lines of the

earlier Delhi experiment.

20

In Figure 2, we find that workers make allocation decisions based on a number of

characteristics. In particular, they provide significantly greater amounts to (a) the mother of a

moderately malnourished child, (b) the poor mother, (c) the illiterate mother, (d) the non-

working mother, (e) the mother with low assets, and (f) the mother with a physically challenged

child. Moreover, the proportions are much large for (b), (e), and (f) compared to (a) and (c),

suggesting that some social norms and attitudes towards social justice might influence the

workers more strongly. For example, workers might prefer to give more to those they think are

less able to fend for themselves and their children. For (g), we do not find significant differential

allocation by child’s gender and the sign predicts favorable treatment towards girls (See

Appendix Table A1 for related hypotheses and results). The results suggest that although

religious preferences do not affect subject choices, welfare-based and desert-based distributive

principles seem to influence workers’ allocation decisions (Feinberg 1970, Rescher 1966). Also

ruling out small sample size driving the main null findings.

21

Figure 2: Tests for treatment saliency in Chandigarh

26.53

69.39

4.08

020

4060

80Pe

rcen

tage

of A

ngan

wad

i wor

kers

Equal split Moderately malnourished childNormal weight

10.20

83.67

6.12

020

4060

80Pe

rcen

tage

of A

ngan

wad

i wor

kers

Equal split Household income is Rs 2000Household income is Rs 8000

46.9448.98

4.08

010

2030

4050

Perc

enta

ge o

f Ang

anw

adi w

orke

rs

Equal split Mother is illiterateMother is literate

30.61

59.18

10.20

020

4060

Perc

enta

ge o

f Ang

anw

adi w

orke

rs

Equal split Mother is not workingMother is employed

12.24

83.67

4.08

020

4060

80Pe

rcen

tage

of A

ngan

wad

i wor

kers

Equal split No fridge and water filterFridge and water filter

6.12

87.76

6.12

020

4060

80Pe

rcen

tage

of A

ngan

wad

i wor

kers

Equal split PolioNormal

63.27

10.20

26.53

020

4060

Perc

enta

ge o

f Ang

anw

adi w

orke

rs

Equal split SonDaughter

22

4.4 Experimenter demand and scrutiny effects

Finally, as indicated earlier, our experiment is designed to minimize the experiment

demand effects in two important ways. First, instead of directly assigning religious identities

(Hindu or Muslim) we make use of names that reveal religious identity (e.g: Amit or Akbar).

Second, additional information on the children’s health status along with mothers being the

payoff recipients can help minimize the focus on religion for the decision maker.

5. Discussion

In economics generally, and development economics in particular, there has been an

intentional shift towards more evidence-based policies. Policies that address discrimination

are no exception. However, the nature of policies is likely to be very different depending

upon the sources of discrimination. For example, policies targeting intrinsic hatred towards

another group could involve either a “soft” solution like changing perceptions or a “hard”

solution like punishment for racism. On the other hand, statistical discrimination could be

attenuated by providing more information about factors for which religion is being viewed as

a proxy.

In this paper, we design an experiment and introduce a novel allocation game to

identify statistical and taste based discrimination among government-employed pre-school

caregivers in India. Our results suggest that among childcare workers in Chandigarh and New

Delhi there is no evidence of either taste-based or statistical religious discrimination.

Anganwadi workers in our sample do not seem to discriminate along religious lines; instead

they seem to exhibit a social-justice-based attitude towards the poor and the undernourished.

We also find that incentivizing decisions does not change their allocation behavior. It is

plausible that these findings may be specific to public health workers who may be

23

intrinsically motivated and are often identified as different from professionals in other

sectors.

Our elicited preferences thorough experimental tasks also seem to support findings

from primary data collected on health status of children in these geographical areas. Singh

and Masters (2016) collected data on weights of children present in the universe of 245

Anganwadis located in urban slums of Chandigarh to find that the weight-for-age

malnutrition rates are nearly identical for both Hindu and Muslim children in Chandigarh, at

37.5% and 38.2% respectively. While this is encouraging from a policy perspective, the lack

of evidence on religious discrimination does not imply that such practices are absent in other

areas of public health, or among Anganwadi workers in other states in India. Further work to

evaluate such evidence in an experimental framework can give us clues about the sources of

discrimination based on religion, skin color, ethnicity, caste or gender.

24

References

Adèr, H. J., Mellenbergh G. J., & Hand, D. J. (2008). Advising on research methods: A consultant's companion. Huizen, The Netherlands: Johannes van Kessel Publishing.

Agan, Amanda Y. and Starr, Sonja B. (2016). ”Ban the Box, Criminal Records, and Statistical Discrimination: A Field Experiment”. U of Michigan Law & Econ Research Paper No. 16-012.

Akerlof, G. A. & Kranton, R. E. (2010). “Identity Economics: How Our Identities Shape Our Work, Wages, and Well-Being”. Princeton Univ. Press.

Altonji, J.G. and R. Blank (1999). “Race and Gender in the Labor Market.” In O. Ashenfelter and D. Card (eds) Handbook of Labor Economics, vol. 3C. Amsterdam: Elsevier Science.

Anderson, L., N. Jakobsson, and A. Kotsadam (2012). “A Field Experiment of Discrimination in the Norwegian Housing Market: Gender, Class, and Ethnicity.” Land Economics, 88(2): 233–40.

Arrow, K. (1972). “Models of Job Discrimination.” In A.H. Pascal (ed.) Racial Discrimination in Economic Life. Lexington, MA: D.C. Heath.

Banerjee, A., S. Datta, and S. Mullainathan. (2009). “Labor Market Discrimination in Delhi: Evidence from a Field Experiment.” Journal of Comparative Economics, 37(1): 14–27.

Ball, S.B. & Eckel, C.C. (1996). Buying status: Experimental evidence on status in negotiation. Psychology and Marketing, 13(4), pp.381–405.

Ball, S. et al. (2001). Status in Markets. The Quarterly Journal of Economics, 116(1), pp.161–188.

Baltussen, G., G. Thierry Post, M.J. van den Assem, and P.P. Wakker. (2012). “Random Incentive Systems in a Dynamic Choice Experiment.” Experimental Economics, 15: 418–43.

Becker, G.S. (1971). The Economics of Discrimination, 2nd edn, 6th impr. Chicago: University of Chicago Press.

Behrman, J.R., M.C. Calderon, S. Preston, J. Hoddinott, R. Martorell, and A.D. Stein. (2009). “Nutritional Supplementation of Girls Influences the Growth of their Children: Prospective Study in Guatemala.” American Journal of Clinical Nutrition 90(November): 1372–79.

Berge, L. I. O., Bjorvatn, K., Galle, S., Miguel, E., Posner, D. N., Tungodden, B., & Zhang, K. (2015). How Strong are Ethnic Preferences? NBER Working Paper 21715, (April), 0–50. http://doi.org/10.1007/s13398-014-0173-7.2

Bertrand, M., and S. Mullainathan. (2004). “Are Emily and Greg More Employable than Lakisha and Jamal? A Field Experiment on Labor Market Discrimination.” American Economic Review, 94(4): 991–1013.

Bettinger, E., and R. Slonim. (2007). “Patience among Children.” Journal of Public Economics, 91: 343–63.

Bhalotra, S., C. Valente, and A. van Soest. (2010). “The Puzzle of Muslim Advantage in Child Survival in India.” Journal of Health Economics, 29: 191-204.

Billig, M. & Tajfel, H. (1973). Social categorization and similarity in intergroup behaviour. European Journal of Social Psychology, 3(1), pp.27–52.

Bolle, F. (1990). “High Reward Experiments without High Expenditure for the Experimenter.” Journal of Economic Psychology, 11: 157–67.

Brainerd, E. and Menon, N. (2015). ”Religion and Health in Early Childhood: Evidence from South Asia.” Population and Development Review, 41: 439–463. doi:10.1111/j.1728-4457.2015.00067.x

25

Cappelen, AW, J Konow, EØ Sørensen, B Tungodden. (2013). “Just Luck : An Experimental Study of Risk Taking and Fairness.” American Economic Review 103 (4): 1398–1413. doi:10.1257/aer.103.4.1398. Carlson,

Census of India. (2011). “Population by Religious Community—2011.” Office of the Registrar General & Census Commissioner.

Coppola, M. (2014). “Eliciting Risk-preferences in Socio-economic Surveys: How Do Different Measures Perform?” Journal of Socio-Economics, 48: 1–10.

Devine, P.G. (1989). Stereotypes and prejudice: Their automatic and controlled components. Journal of Personality and Social Psychology, 56(1), pp.5–18.

Diehl, M. (1988). Social identity and minimal groups: The effects of interpersonal and intergroup attitudinal similarity on intergroup discrimination. British Journal of Social Psychology, 27(4), pp.289–300.

Dohmen, T., A. Falk, D. Huffman, U. Sunde, J. Schupp, and G. G. Wagner. (2011). “Individual Risk Attitudes: Measurement, Determinants, and Behavioral Consequences.” Journal of the European Economic Association, 9(3): 522–50.

Feinberg, J. (1970). “Justice and Personal Desert,” Doing and Deserving, Princeton, NJ: Princeton University Press, 55–94.

Fershtman, C., and U. Gneezy. (2001). “The Discrimination in a Segmented Society: An Experimental Approach.” Quarterly Journal of Economics, 116: 351–77.

Gangadharan, L, T. Jain, P. Maitra, and J. Vecci, (forthcoming), Social Norms and Governance: The Behavioral Response to Female Leadership, European Economic Review. Gneezy, U., and A. Rustichini. (2000). “Pay Enough or Don’t Pay at All.” Quarterly Journal of

Economics, 115(3): 791–810.

Guillot, M., and Keera Allendorf. (2010). "Hindu-Muslim Differentials in Child Mortality in India." Genus, 66(2): 43-68.

Gupta G., M. Mahmud, P. Mitra, S. Mitra and A. Neelim. (2016). Religion, Minority Status and Trust: Evidence from a Field Experiment. Working Paper.

Guryan, J. and Charles, K. K. (2013), Taste-based or Statistical Discrimination: The Economics of Discrimination Returns to its Roots. The Economic Journal, 123: F417–F432. doi:10.1111/ecoj.12080.

Habyarimana, J, M Humphreys, DN Posner, JM Weinstein. (2007). “Why Does Ethnic Diversity Undermine Public Goods Provision?” American Political Science Review 101 (4): 709–25. doi:10.1017/S0003055407070499.

Hanna, R.N., and L.L. Linden. (2012). “Discrimination in Grading.” American Economic Journal: Economic Policy, 4(4): 146–68.

Hoff, K., and P. Pandey. (2004). “Belief Systems and Durable Inequalities: An Experimental Investigation of Indian Caste”, World Bank Working Paper Series. Washington, DC: World Bank.

Hungama Report. (2011). Hungama: Fighting Hunger & Malnutrtion. Naandi Foundation.

Kass, M., E. Fatas, C. Eckel, and D. Arce. (2013). “The UN in the Lab.” Social Choice and Welfare, 45(3): 625–51.

Kaas, L. and C. Manger. (2011). Ethnic Discrimination in Germany’s Labour Market: A Field Experiment. German Economic Review, 13 (1): 1-20.

26

Klein, O. & Azzi, A., 2001. Do High Status Groups Discriminate More? Differentiating between Social Identity and Equity Concerns. Social Behavior and Personality: an international journal, 29(3), pp.209–221.

Lane, T. (2015). “Discrimination in the Laboratory: A Meta-analysis of Economics Experiments.” European Economic Review, Available at: http://dx.doi.org/10.1016/j.euroecorev.2015.11.011 (accessed December 29, 2015).

Leonard, T., K. Shuval, A. de Oliveira, C.S. Skinner, C. Eckel, and J.C. Murdoch (2013). “Health Behavior and Behavioral Economics: Economic Preferences and Physical Activity Stages of Change in a Low-income African American Community.” American Journal of Health Promotion 27(4): 211–21.

List, J. (2004). “The Nature and Extent of Discrimination in the Marketplace: Evidence from the Field.” Quarterly Journal of Economics, 119(1): 49–89.

Maluccio, J.A., J. Hoddinott, J.R. Behrman, A. Quisumbing, R. Martorell, and A.D. Stein (2009). “The Impact of Nutrition During Early Childhood on Education among Guatemalan Adults.” Economic Journal 119(April): 734–63.

Mani, S. (2012). “Is There Complete, Partial, or No Recovery from Childhood Malnutrition—Empirical Evidence from Indonesia.” Oxford Bulletin of Economics and Statistics, 74(5): 691–715.

Moffatt, P.G. (2016). Experimetrics Econometrics for Experimental Economics. London: Palgrave Macmillan.

Neilson, William and Ying, Shanshan (2016). “From taste-based to statistical discrimination.” Journal of Economic Behavior & Organization, 129, issue C, p. 116-128.

Phelps, E.S. (1972). “The Statistical Theory of Racism and Sexism.” American Economic Review, 62(4): 659–61.

Rescher, N. (1966). Distributive Justice: A Constructive Critique of the Utilitarian Theory of Distribution, Indianapolis: Bobbs-Merrill.

Singh, P. (2015). “Performance Pay and Information: Reducing Child Undernutrition in India.” Journal of Economic Behavior & Organization, 112(April): 141–63.

Singh, P., and S. Mitra (forthcoming). “Incentives, information and malnutrition: Evidence from an experiment in India.” European Economic Review (in press).

Singh, P., and W. Masters (2016). “Impact of Caregiver Incentives on Child Health: Evidence from an Experiment with Anganwadi Workers in India.” IZA Discussion Paper No. 10083. Available at: http://ftp.iza.org/dp10083.pdf

Siddique, Z. (2009). “Caste Based Discrimination: Evidence and Policy.” Available at: http://papers.ssrn.com/sol3/papers.cfm?abstract_id=1550883 (accessed May 12, 2016).

Stein, A.D., H.X. Barnhart, M. Hickey, U. Ramakrishnan, D.G. Schroeder, and R. Martorell (2003). “Prospective Study of Protein-energy Supplementation Early in Life and of Growth in the Subsequent Generation in Guatemala.” American Journal of Clinical Nutrition, 78: 162–7.

Stein, A.D., M. Wang, A. Di Girolamo, R. Grajeda, U. Ramakrishnan, M. Ramirez-Zea et al. (2008). “Nutritional Supplementation in Early Childhood, Schooling and Intellectual Functioning in Adulthood: A Prospective Study in Guatemala.” Archives of Pediatric and Adolescent Medicine, 162: 7(July): 612–18.

Turner, J. C. , & Brown, R. J. Social status, cognitive alternatives, and intergroup relations. In H. Taifel , (Ed.), Differentiation between social groups: Studies in the social psychology of intergroup relations. London: Academic Press, 1978.

27

Vaughan, G.M., Tajfel, H. & Williams, J., 1981. Bias in Reward Allocation in an Intergroup and an Interpersonal Context. Social Psychology Quarterly, 44(1), p.37.

Victora, C.G. , L. Adair, C. Fall, P.C. Hallal, R. Martorell, L. Richter et al., on behalf of the Maternal and Child Undernutrition Study Group (2008). “Undernutrition 2: Maternal and Child Undernutrition: Consequences for Adult Health and Human Capital.” The Lancet, 371(9609): 340–57.

Zizzo, D.J. (2010). “Experimenter Demand Effects in Economic Experiments.” Experimental Economics, 13: 75–98.

28

Appendix: Additional Figures and Tables

Figure A1: Power calculations

Table A1: Workers Allocation Decision in Follow-up Experiment

Hypothesis

Decision Difference (p-value)

Chandigarh (1)

H1: In INFO-NOID, amount allocated to the mother of a moderately malnourished child – amount allocated to the mother of a child with normal weight = 0

Reject (0.00)

H2: In INFO-NOID, amount allocated to the mother of a child with a household income of Rs. 2000 per month – amount allocated to the mother of a child with a household income of Rs. 8000 per month = 0

Reject (0.00)

H3: In INFO-NOID, amount allocated to an illiterate mother who has a child – amount allocated to a literate mother who has child = 0

Reject (0.00)

H4: In INFO-NOID, amount allocated to a non-working mother who has a child – amount allocated to a working mother who has a child = 0

Reject (0.00)

H5: In INFO-NOID, amount allocated to the mother of a child whose house does not have fridge and water filter – amount allocated to the mother of a child who has both a fridge and water filter at home = 0

Reject (0.00)

H6: In INFO-NOID, amount allocated to a mother whose child has Polio – amount allocated to mother whose child is normal = 0

Reject (0.00)

H7: In INFO-NOID, amount allocated to a boy child’s mother – amount allocated to a girl child’s mother = 0

Do not Reject (0.18)

100

150

200

250

Tota

l sam

ple

size

(N)

.7 .75 .8 .85 .9 .95Power (1-β)

Parameters: α = .05, N2N1 = 1, δ = .45

Satterthwaite's t test assuming unequal variancesH0: µ2 = µ1 versus Ha: µ2 ≠ µ1

Estimated total sample size for a two-sample means test

29

Table A2: Tests of Taste and Statistical Discrimination among women who participated in the phone surveys

Hypotheses

Chandigarh sample

Decision*

New Delhi sample

Decision*

Choices are Rational without information on religion.

H1: Proportion of subjects allocating more to Child 2 = Proportion of subjects allocating more to Child 1 in INFO-NOID treatment.



There is no evidence of taste-based discrimination.

H2: Proportion of subjects allocating more to Hindu child= Proportion of subjects allocating more to Muslim child in INFO-ID treatment.



There is no evidence of statistical discrimination.

H3: The difference in proportion of subjects allocating more to Hindu child and proportion of subjects allocating more to Muslim child in NOINFO-ID = The difference in proportion of subjects allocating more to Hindu child and proportion of subjects allocating more to Muslim child in INFO-ID



30

Appendix A2: Subject instructions for Chandigarh

Instructions for Task 2

Thank you for your participation. You will be paid Rs. 50 for your participation.

Each of you will participate in one task in this experiment. At the end of the task, in addition to the fixed fee, 25 per cent of you will be randomly chosen for additional payments. Here is a bag that contains tokens with all subject IDs written on them. One of the workers will be asked to pick a certain number of these tokens and announce the IDs. For example, if there are 40 participants in a session, we will pick 10 IDs and pay the participants who have those IDs. These randomly chosen participants will be paid according to their decision made on this task.

It is important that you read the instructions for the task before making a decision. If you do not understand, you will not be able to participate effectively. In case you have a question while reading the instructions, please raise your hand and we will escort you to the nearby room and clarify your query.

Each page has an ID# on it. Do not show this ID# to any other participant or allow it to be visible to anyone during or after this experiment. There should be no talking or discussion of the task amongst you while waiting for the experiment to begin. We request you to remain seated, and not do anything unless instructed by the experimenter. Also do not look at others’ responses once the experiment begins. You might be disqualified otherwise.

If you are ready, then we will proceed.

Each of you will make decisions that will decide your own payouts and the payouts of mothers of two randomly chosen pre-school children. The children will be chosen from our sample through a lottery. They belong to an Anganwadi Center in an urban slum of Chandigarh that is not in your block. Note that only mothers of children identified as “moderately malnourished/suffer from grade 2 malnourishment” are eligible to receive the payouts from the experiment.

All decisions will remain private, and no one will get to know who was matched with whom (either during, or after the experiment due to the code structure of the experiment).

In this task you have inherited Rs. 120. Your task is to decide on how much of that money you want to keep for yourself, how much of it you want to give to give to the two randomly chosen mothers of grade 2 malnourished children.

Please choose only one of the following four options: A, B, C or D. Notice that in all the options, the sum of the allocations between you, mother of child 1, and mother of child 2 should add up to Rs. 120. Note that child 1 and child 2 are both “moderately malnourished/suffer from grade 2 level of malnourishment.”

After marking your decisions please fold your response sheet and place it in front of you. The experimenter will collect them.

Child 1’s name is [Announced by the experimenter]……………………………………..

Child 2’s name is [Announced by the experimenter] ……………………………………..

CIRCLE THE OPTION YOU WANT TO CHOOSE BELOW:

OPTIONS

A I want to keep Rs. 40 for myself I want to give Rs. 40 to the mother of child 1 I want to give Rs. 40 to the mother of child 2

31

B I want to keep Rs. 40 for myself I want to give Rs. 0 to the mother of child 1 I want to give Rs. 80 to the mother of child 2

C I want to keep Rs. 40 for myself I want to give Rs. 80 to the mother of child 1 I want to give Rs. 0 to the mother of child 2

D I want to keep Rs. 40 for myself I want to give Rs.. ……...(write amount here) to the mother of child 1 I want to give Rs. ……..(write amount here) to the mother of child 2 [Note: the three amounts must add up to Rs. 120]

General Instructions for Tasks 1 and 3

Thank you for your participation. You will be paid Rs. 50 for your participation.

Each of you will participate in two tasks in this experiment. At the end of the two tasks, in addition to the fixed fee, we will randomly choose one task for which some of you will also receive additional payments.

On the table at the front of the room are 2 cards, each with 1, or 2 written on them. They are kept facing down so that the numbers are not visible. We will ask one of you to pick one of the cards and announce the number written on that card. That particular task will then be used for making the additional payments for this session. For example, if the worker picks a card with 1 written on it, task 1 will be used for making the additional payments. After a specific task has been chosen for payments, 25 per cent of you will be randomly chosen for the additional payments. Here is a bag that contains tokens with all subject IDs written on them. One of the workers will be asked to pick a certain number of these tokens and announce the IDs. For example, if there are 40 participants in a session, we will pick 10 IDs and pay the participants who have those IDs. These randomly chosen participants will be paid according to their decision made on the task that was chosen for the additional payments.

We are about to read the instructions for the two tasks. Please listen carefully. It is important that you read the instructions for the tasks before making a decision. If you do not understand, you will not be able to participate effectively. In case you have a question while reading the instructions, please raise your hand and we will escort you to the nearby room and clarify your query.

Each page has an ID# on it. Do not show this ID# to any other participant or allow it to be visible to anyone during or after this experiment. There should be no talking or discussion of the task amongst you while waiting for the experiment to begin. We request you to remain seated, and not do anything unless instructed by the experimenter. Also do not look at others’ responses once the experiment begins. You might be disqualified otherwise.

If you are ready, then we will proceed.


Each of you will make decisions that will decide your own payouts and the payouts of mothers of two randomly chosen pre-school children. The children will be chosen through a lottery from an Anganwadi Center in an urban slum of Chandigarh that is not in your block. Note that only mothers of children identified as “moderately malnourished/suffer from grade 2 level malnourishment” are eligible to receive the payouts from the experiment.


32

In this task you have inherited Rs. 120. Your task is to decide on how much of that money you want to keep for yourself, how much of it do you want to give to the two randomly chosen mothers of grade 2 level malnourished children.

Please choose only one of the following four options: A, B, C, or D. Notice that in all the options, the sum of the allocations between you, mother of child 1, and mother of child 2 should add up to Rs. 120. Note that child 1 and 2 are both “moderately malnourished/suffer from grade 2 level of malnourishment.”



OPTIONS

A I want to keep Rs. 40 for myself I want to give Rs. 40 to mother of child 1 I want to give Rs. 40 to mother of child 2

B I want to keep Rs. 40 for myself I want to give Rs. 0 to mother of child 1 I want to give Rs. 80 to mother of child 2

C I want to keep Rs. 40 for myself I want to give Rs. 80 to mother of child 1 I want to give Rs. 0 to mother of child 2

D I want to keep Rs. 40 for myself I want to give Rs. ……(write amount here) to mother of child 1 I want to give Rs. ……(write amount here) to mother of child 2 [Note: the two amounts must add up to Rs. 80]


Each of you will make decisions that will decide your own payouts and the payouts of mothers of two randomly chosen pre-school children. The children will be chosen from our sample through a lottery. They belong to an Anganwadi Center in an urban slum of Chandigarh that is not in your block.


In this task you have inherited Rs. 120. Your task is to decide on how much of that money you want to keep for yourself, how much of it you want to give to the two randomly chosen mothers of pre-school children.

Please choose only one of the following four options: A, B, C or D. Notice that in all the options, the sum of the allocations between you, mother of child 1, and mother of child 2 should add up to Rs. 120.


Child 1’s name is [Announced by the experimenter]……………………………………..

Child 2’s name is [Announced by the experimenter] ……………………………………..


33

OPTIONS

A I want to keep Rs. 40 for myself, I want to give Rs. 40 to the mother of child 1 I want to give Rs. 40 to the mother of child 2

B I want to keep Rs. 40 for myself, I want to give Rs. 0 to the mother of child 1 I want to give Rs. 80 to the mother of child 2

C I want to keep Rs. 40 for myself, I want to give Rs. 80 to the mother of child 1 I want to give Rs. 0 to the mother of child 2

D I want to keep Rs. 40 for myself, I want to give Rs. ……...(write amount here) to the mother of child 1 I want to give Rs. ……..(write amount here) to the mother of child 2 [Note: the three amounts must add up to Rs. 80]

religious discrimination among childcare workers ......(chandigarh and new delhi) among 183...

Documents