SOCIAL GROUP IDENTITY, ELECTORAL RESERVATIONS, AND ENTREPRENEURSHIP - - PowerPoint PPT Presentation

SOCIAL GROUP IDENTITY, ELECTORAL RESERVATIONS, AND ENTREPRENEURSHIP

Venkata Krishna Nadella Indiana University, Bloomington October 11, 2017 UNU-WIDER Helsinki, Finland

OUTLINE

• Caste as social groups
• Quota system and electoral reservations
• Literature and Research question
• Data and variables
• Baseline comparisons
• Analysis: Matched-pair (policy feature) and RDD (politics)
• Working hypotheses

CASTE AS SOCIAL GROUPS

Caste groups are endogamous social groups whose membership is confined by birth (Ketkar, 1909) Hundreds of localized caste groups across the country; with identity formation routed in history of human displacement (Ambedkar, 1944) Genome research traces intermarriage restrictions to 70 generations ago with asymmetric mixing between immigrant and indigenous populations irrespective of ancestry (Basu et al., 2016) Socio-cultural norms restricting intermarriage among large populations has created a system of social stratification, with strong in-group identities, most dominant in rural India (Karve, 1961; Srinivas, 1976)

THE QUOTA SYSTEM: RESERVATIONS

Indian constitution recognizes dalit (SCs) and adivasi (ST) populations for targeted affirmative-action policies through the quota system, which includes

• Public, and ‘publicly-funded’, education
• Public sector employment including administrative civil service
• Parliamentary and state elections

ELECTORAL RESERVATIONS

Electoral reservations for SCs and STs exist at the national (parliamentary) and state (assembly) levels of government and are the most comprehensive mandate to ensure proportional political representation  The delimitation commission assigns reservation status and constituency boundaries after every decennial population census  In reserved constituencies, only members from SC or ST, are nominated to contest elections while the entire electorate votes  Changes to the number and status of reserved constituencies have been made four times since Independence - 1952, 1962, 1972 & 2002.  Reserved constituencies drafted in 1972 remained in place for elections between 1976 to 2008 (the time period of this study).

ONE VOTE ONE VALUE

“On the 26th of January 1950, we are going to enter into a life of

• contradictions. In politics we will have equality and in social and

economic life we will have inequality. In politics we will be recognizing the principle of one man one vote and one vote one value. In our social and economic life, we shall, by reason of our social and economic structure, continue to deny the principle of one man one value.” (emphasis added)

• B.R. Ambedkar addressing the Constituent Assembly of

India on 25th November, 1949

Source: Constituent Assembly of India, Volume XI http://parliamentofindia.nic.in/ls/debates/vol11p11.htm

LITERATURE: TWO STYLIZED FACTS

• 1. Electoral quotas are effective in aggregating policy preferences of

disadvantaged groups with regards to ‘low spillover goods’

Public spending (Pande, 2003) Public good provision (Besley et al., 2004; Chattopadhyay and Duflo, 2004) Poverty alleviation (Chin and Prakash, 2011)

• 2. Despite welfare gains over time, there are no relative gains of

political representation on ‘long-run development outcomes’ of targeted groups

Development indicators (Jensenius, 2015) Literacy rates (Bhavnani and Jensenius, 2017)

RESEARCH QUESTION/S

Do electoral quotas improve allocation to productive entrepreneurship among disadvantaged social groups? Policy (and programmatic) evaluation approach focusing on

1. Target population: SC and ST populations in reserved constituencies 2. Treatment Condition: a) quota assignment, and b) effect of politics 3. Programmatic Aspects of Quota Policy and Implementation: Freezing of reserved status for large periods of time, and method of picking reserved constituencies

DATA CHALLENGE

Secondary data sources, at the individual, household, enterprise, and industry levels, are sampled at the district level and cannot be disaggregated at geographies below district boundaries. As a consequence, reliable estimates for policy evaluations using secondary data sources can only be obtained at the district levels This challenge can be circumvented by using population and economic censuses of India, which focus on complete enumeration of household populations, with metadata on names of villages, towns, blocks, sub-districts, and other administrative boundaries.

CONSTITUENCY LEVEL DATASET

Construction of the constituency level dataset is similar to Asher & Novosad (2017) by matching village and town names across waves of the economic census and population census datasets, AND identifying the location of these villages and towns within state assembly constituency boundaries using GIS metadata provided by ML Infomap (a private mapping firm based in New Delhi)

CENSUS DATA & VARIABLES

VILLAGES MATCHED (1990‐2005) TOWNS MATCHED (1990‐2005)

EC-1990 EC-1998 EC-2005 PC-1991 PC-2001 Individual Enterprise variables Socio-economic variables PC-1971 Policy Change 1976

BASELINE COMPARISONS: 1991 POPULATION CENSUS

Variables Overall Constituency Type GEN SC ST Rural Public Goods (Share villages), 1991 paved road access 0.61 0.63 0.56 0.53 power supply 0.80 0.83 0.77 0.64 primary schools 0.86 0.87 0.84 0.86 Urban public goods (real values), 1991 paved roads (kms.) 9.36 10.12 7.38 6.08 # electricity connections 2381 2547 2238 1072 # primary schools 6.85 7.30 6.26 3.73

GROWTH AND URBANIZATION (1990-2005)

Variables Overall Constituency Type GEN SC ST Period 0 (1990-1998) Annualized Employment Growth (change in log emp) 0.020 0.020 0.021 0.024 Urbanization 1998 (prop. urban jobs) 0.21 0.22 0.20 0.16 Period 1 (1998-2005) Annualized Employment Growth (change in log emp) 0.029 0.030 0.030 0.021 Urbanization 2005 (prop. urban jobs) 0.31 0.33 0.28 0.24

EMPIRICAL SETTING (1): MATCHED PAIR

The Delimitation commission of India assigns reservation status to constituencies, proportional to the population share within a state,

• n the basis of two criteria (Jensenius, 2015):
• 1. Highest proportion of SCs in the district in the state (rank)
• 2. Preference to geographical spread within the state (spread)

This treatment assignment feature of the delimitation commission meant that constituencies within a district (in a state) that were very similar to the treated constituency in all aspects (with the exception atleast one of the above criteria) failed to receive treatment.

EMPIRICAL SETTING (2): MATCHED PAIR

To evaluate the intended effect of treatment (reservation assignment)

• n the treated (reserved) units, I compare the outcomes in treated

units by matching each treated constituency with a untreated (general) constituency. This approach is similar to Jensenius (2015) Data: Matched 279 reserved constituencies to general constituencies Treatment: SC Reservation Status for constituencies Outcomes of Interest:

• 1. Self-employment rates of Scheduled Caste populations
• 2. Employment by SC firms
• 3. Relative Gaps in SC and Non-SC economic outcomes

PRETREATMENT SAMPLE BALANCE

Variable (1971 Population Census) General Reserved Difference p-value (1) (2) (2)-(1) Literacy Rate (Overall) 24.80 24.49

• 0.30

0.24 SC Literacy Rate 11.91 11.97 0.06 0.789 Non-SC Literacy Rate 28.09 27.85

• 0.25

0.365 Employment Rate (Overall) 32.83 32.84 0.01 0.896 SC Employment Rate 37.12 37.09

• 0.03

0.856 Non-SC Employment Rate 31.82 31.82 0.00 0.986 Agri Labor (Overall) 21.25 20.96

• 0.28

0.352 SC Agri Labor 53.24 52.74

• 0.51

0.485 Non-SC Agri Labor 20.40 20.09

• 0.31

0.357 SC Percent in Constituency 20.30 21.79 1.49 <0.001 SC Population 31,576 33,669 2,094 <0.001 Total Population 155,392 154,062

• 1,330

<0.001

SC ECONOMIC OUTCOMES

1990 1998 2005 Difference p-value Difference p-value Difference p-value SC Self-employment Rates 0.011 0.011 0.003 0.188 0.002 0.498 SC Firms (1-9 employees) 0.001 0.250 0.001 0.149 0.002 0.018 SC Firms > 10 emps.

• 0.001

0.317

• 0.001

0.540 0.000 0.483 SC Firms > 20 emps. 0.000 0.648

• 0.001

0.330 0.000 0.503 All SC Firms employment 0.012 0.009 0.004 0.119 0.004 0.195 Notes: Standard errors clustered at state level

GAP IN ECONOMIC OUTCOMES

1990 1998 2005 Difference p-value Difference p-value Difference p-value Self-employment Rates

• 0.008

0.230

• 0.005

0.474 0.004 0.439 Firms (1-9 employees)

• 0.001

0.800

• 0.008

0.084

• 0.005

0.276 Firms > 10 emps

• 0.014

0.097 0.002 0.699

• 0.012

0.038 Firms > 20 emps

• 0.016

0.036 0.002 0.716

• 0.010

0.049 All Firms employment

• 0.024

0.009

• 0.011

0.229

• 0.013

0.065

Non SC minus SC comparisons. A negative sign means that gap is lower in reserved cons.

Notes: Standard errors clustered at state level

EMPIRICAL SETTING: RDD

Political favoritism is a key mechanism through which ruling party constituencies gain favors from winning coalitions (Asher & Novosad, 2017) Specific mechanisms include local regulations (occupational licensing, health and safety standards), public contracting (big-ticket projects, grants), and political control over bureaucratic discretion (management of public infrastructure and rule of law) Instrument: Using the discontinuity in close elections for two time periods between the three waves of the economic census i.e.;1990- 1998; and 1998-2005

RDD: INTERNAL VALIDITY (1)

Sample RD balance

Ruling Party Constituencies Opposition Constituencies RD Estimate t-Stat Baseline Log Employment 2.352 2.352

• 0.001

0.989 Urbanization 0.249 0.243 0.006 0.835 Share villages with power supply 0.847 0.838 0.009 0.791 Share villages with paved road 0.588 0.611

• 0.023

0.512 Share villages with Primary Schools 0.873 0.860 0.013 0.82 Share of agricultural land irrigated 0.195 0.122 0.070 0.011 # Urban Schools 8.797 9.659

• 0.860
• 0.85

# Urban Electricity Connections 3183 2975 208 0.46 Urban Roads (in kms) 13.314 13.122 0.192 0.14

RDD: INTERNAL VALIDITY (2)

.01 .02 .03 .04

• 50

50

Margin of Victory Annualized log Emp. Growth

POLITICS AND ECONOMIC GROWTH (1)

Variables OLS RDD (local linear) RDD (global polynomial) (1) (2) (3) Ruling party status 0.004* 0.014*** 0.009** (0.002) (0.004) (0.004) Margin of victory

• 0.012
• 0.144

(0.010) (0.155) Margin x ruling 0.012 0.072 (0.013) (0.229) Baseline log employment

• 0.019***
• 0.021***
• 0.019***

(0.002) (0.004) (0.002) State-Election Fixed Effects Yes Yes Yes Controls Yes Yes Yes R-Squared 0.233 0.216 0.233 N 3,521 742 3,521

Estimates of the effect of Ruling party status on Employment Growth

• Dep. Var.: change in ann. log employment; Period 0 (1990-1998) & Period 1 (1998-2005)

Notes: Standard errors clustered at state-period level

POLITICS AND ECONOMIC GROWTH (2)

Variables General Reserved (SC &ST) SC-Reserved ST-Reserved (1) (2) (3) (4) Ruling party status 0.017*** 0.002

• 0.013

0.038 (0.006) (0.022) (0.025) (0.039) Margin of victory

• 0.189

0.239 0.594

• 0.880

(0.178) (0.486) (0.726) (0.917) Margin x ruling 0.084

• 0.324
• 0.562

0.213 (0.330) (0.480) (0.859) (1.130) Baseline log employment

• 0.020***
• 0.030***
• 0.033***
• 0.021

(0.005) (0.007) (0.010) (0.028) State-Election Fixed Effects Yes Yes Yes Yes Controls Yes Yes Yes Yes R-Squared 0.240 0.196 0.226 0.101 N 570 167 107 60

• Dep. Var.: change in ann. log employment; Period 0 (1990-1998) & Period 1 (1998-2005)

Notes: Standard errors clustered at state-period level

POLITICS AND GROUP OUTCOMES

Variables Overall General SC-Reserved (1) (2) (3) Ruling party status 0.054** 0.059* 0.025 (0.026) (0.032) (0.044) Margin of victory

• 0.907
• 1.005

0.137 (0.873) (1.146) (0.889) Margin x ruling

• 0.018
• 0.230
• 0.498

(0.911) (1.080) (1.654) Period 1 Dummy

• 0.078***
• 0.068***
• 0.084**

(0.006) (0.008) (0.035) Baseline log employment

• 0.022*
• 0.013
• 0.055*

(0.012) (0.013) (0.030) State-Election Fixed Effects Yes Yes Yes Controls Yes Yes Yes R-Squared 0.240 0.196 0.226 N 570 167 107

RDD local estimates of the effect of Ruling party status on Employment Gaps

• Dep. Var.: change in employment gap between SC and Non-SC firms in constituency;

Period 0 (1990-1998) & Period 1 (1998-2005)

Notes: Standard errors clustered at state-period level

WORKING HYPOTHESES

• Centrality Hypotheses: It is plausible that SC and ST politicians

are not central to their political parties and hence do not transfer as many favors downs to their constituents (Jensenius, 2017)

• Elite Capture Hypotheses: It is plausible that benefits in reserved

constituencies go to a small number of firms, which leads to no significant changes on the median SC/ST voter

• Mobilization Hypotheses: Another explanation can be that there

is more caste-based mobilization in general constituencies, which does not exist in reserved constituencies