Embedded Files
ECON6015_Lecture 4.pdfPolitical Reforms: Quotas in the Legislature
Why have quotas in the legislatures? such as:
gender
race
language
demographic groups, etc
ethnic minorities
A basic premise of representative democracy is that all those subject to policy should have a voice in its making. The inclusion of all parties in a nation should be present.
This involves systematic change by changing rules who can be the elective candidate by changing the set of eligible people.
Tradeoff between what is fair and what is efficient (Pareto Efficient)
Don't always turn out as expected, which revolves around implementation issues
However, policies enacted by electorally accountable governments often fail to reflect the interest of disadvantaged minorities
The Role of History - explains how things are, culturally, economically, institutionally
Many countries have experimented with mandates which seek to increase minority representation in the political process. India in the 1950s, South Africa in the 1990s.
Political Quotas: A Specific Context
The Indian Setting
The Role of History: the centuries old caste system
Caste System: form of social hierarchy
Hierarchical, denied (basic) rights to some
Education, assets, land-holdings, "voice" in society
1950's India put in place wide-ranging affirmative action policies, part of it was quotas in the legislature.
There were 10 or so states, each state had to follow the quota sustem.
Political Quotas in India: Pande (2003)
Central government and state governments required to adhere to quota system of including the ethnic minority.
Their paper focused on state legislature, and whether including them had any affect of the outcome on policy.
Prior to every state election, specified jurisdictions are declared reserved for disadvantaged castes and for tribes. Important to understand why and how certain constituencies were chosen."Only those from that minority can run". Therefore, only members of the group which benefit from reservation can stand for election. However, the entire electorate votes over the set of candidates.
Panda (2003): Quotas in Indian States
The idea is that legislative capture by non-minority individuals adversely affects the policy interest of minority groups. Therefore, it seemed to be serving the poor purpose.
Pande exploits the institutional features of political reservation to examine the role of mandated political representation in providing disadvantaged influence over policy-making.
No definitive answer in social sciences, why does it work here but not there?
Is there any change in policy? Yes
Political reservation has increased transfers to groups which benefit from the mandate.
Quotas in Indian States: The Procedure
Law requires that the percentage of seats reserved for SC/ST legislator be equal to their percentage in the state’s population.
In selecting reserved jurisdictions, preference is given to jurisdictions with a higher population share of the group in whose favour reservation is being practiced, while ensuring a sufficient dispersal of reserved jurisdictions within the state.
SCs make up roughly 16% of the Indian population, and STs another 8% which varies among the many states (higher percentage in some and lower in others.
Scheduled Castes (SCs): in the number x schedule of Indian constitution these people are listed by name. Designated minority
Tribe Castes (TC's): very marginalised, not part of same social hierarchy, but exceptionally poor and marginalised. Mostly lived in forests did not engage with the modern economy.
These are broad categories
Census updates the population every 10 years. The reservations change proportionally to the census.
This takes effect at the next state election after the census. The elections take place every 5 years (there will only be elections in every 3/4 states who will go through elections), making it staggered and aiding in econometric analysis.
Pande exploits this institutional feature of reservation to isolate its impact.
Data and Variables
A panel data data set of the 16 major Indian states which spans the period 1960-1992.
Our unit of observation is a state in a particular year These states account for over 95% of the Indian population.
Our unit of observation is a state in a particular year. We are going annually, and the population numbers come every 10 years due to the census and staggered elections (interpolation). This means we look at the population ten years back, then now and make an assumption as to how it has grown.
Typically occurs in the early parts of the decade, and the population changes in your data, and the constituencies which were reserved don't change until the next election.
The fact that difference states go to election in different years and the census updating is every 10 years creates a random shock to the reservation status.
Measures the political reservation afforded to a group in a state as the fraction reserved for that group in the state.
Outcome (y) Variables (Regressions)
Total Spending: total amount of real state per capita expenditures (the state level).
Education
Land reform
SC/ST job quotas and targeted welfare spending
The above are the summary tables of the research, the fact that we use 16 states over 52 years generates 500+ observations.
Main point of interest is what percentage of the constituencies in that state is reserved to the SC and ST groups.
For every year you can look at the percentage of constituencies which are reserved for either the SC or ST groups and run a (fixed effects) regression.
Fixed Effects Regression: you have these two variables and 9 dummies (one for each year) to look at annual trends, and you have got a dummy for each of the 16 states.
The year dummies captures any kind of shock which happen to all of the states in a particular year (national drought for example). State dummies are used as a way to say "for any state when the proportion of seats changes what is the effect on spending?" Therefore, you are comparing a state over time, not across different states.
A cross section regression analysis would just exploit the variation across different states, would be looking at all the states in one point in time.
If you run this, you are saying that "I am comparing as across states there is differentiation in number of reserved, what effect does this have on spending?" That is less desirable as two states may be very different as there are lots of dimensions. So, what you are describing to the voter about reservation could just be the other dimensions. Making this exercise more credible.
This is why they use as many years as possible (1960-92) to make it more credible, as opposed to just one year.
Given the fact that there is lack of coordination between updating and actual population changes as it is staggered is great, that is what is getting the random shock to reservation. Populations change but reservations haven't changed as they are still part of the old election cycle, and it is every 5 years when it changes.
Therefore, whether or not you have elections in that state, in that year is not determined by the population changes, rather the 5 year rule. So, that gives you an automatic exogenous change in the reservation status.
Pande's Causal Identification Strategy
Some concerns over reverse causation, reservation changes on the rule of population, and population only changes on census (once every 10 years). Therefore, all of these discontinuities built into the system is great from an econometric perspective as it generates the non-smooth transition which helps you to get to causal identification.
After running the first regression we see some effects, but we haven't identified causal effects yet, as there are some important variables such as:
What are the factors that determine how many of the constituencies are reserved or not?
What if the population of minorities also independently affected the total spending and not just through their affect on spending. Then not having them would be introducing an omitted variable bias. Therefore, you need to account for the actual population. Once you have accounted for the population you know how reservation changes according to all of those rules.
But even better, if you can do the interpolation of the population would make it even better, and that is what Pande does, taken it a step further in column (4) by introducing other controls, such as:
population density
state income per capita lagged one period (year)
whether there was an election year or not
These controls help control for what could potentially be affecting the pattern of spending, and avoiding having any omitted variables.
This makes the fourth column the most reliable column. Therefore, we can look at the top two rows and look at the causal effect of having these quotas on spending.
Changes happen in a very discontinuous way whenever by census coming, and then the state election has to come, and so on and so on. This discontinuous change helps us identify the causal effects.
Pande's
Understanding the Data
There are two sets of data we inspect:
SC Reservation:
-0.04 = coefficient: captures effect of this variable on that variable
(0.007) = standard error: robust as they have been adjusted for heteroskedasticity.
Standard errors: give you a sense as to how noisy, the lower the standard error the better, the ratio of coefficient to standard errors is the t statistic.
The bigger the ratio the more confident you are that this coefficient is not zero and that it is something else. If you are indeed confident (95%, 99%, etc) then you indicate with the stars found at the bottom of Table 6. Unless there are stars, you can safely ignore them, only focus on the stars.
The Results
Looking for ST Reservation in the Row and (1) in the Column:
The interpretation is that as you increase reservation for the ST group, the total spending goes up.
0.023/0.003 indicates that with 99% chance we are happy to say that there is not a null effect, there is actually something.
So, the interpretation when looking at the table as that when you look at total spending there does not seem to be much of an effect from the SC group but the ST group seems to push it up.
The next four columns (5-8), are the same sort of coefficients but the y variable has changed to Education from Total Spending. So, this is to show how much the state is spending on education per capita.
We now see a different effect, ST reservation seems to reduce education.
Whereas, SC seems to have no effect on this such as the previous (total spending).
The final is Land Reform Policies:
In India in the 1950's, the distribution in land holdings was very unequal. Clearly the government declared this as not good so implemented land reforms to improve redisrubtion. Having been implemented in different states at different times, not very effective in some, quite effective in some others.
So, with this you seek to find what the impact is of having leaders from these minority groups on the way that land reforms are being implemented
There appears to be no effect at all from both ST and SC.
Pande's conclusion from this table was that:
There isn't a very clear pattern when it comes to the effect of having these quotas on general policy outcomes (affect everyone in the state).
The question now comes about "what about policy specific to those minority groups, does reservation affect that?"
So, we look at policy targeted outcomes in Table 7:
There is still a huge debate over reservation for minorities in jobs in public offices (job quotas). Even now, people are debating and discussing this. In public hospitals, should you have this system. The question is, do having people from these minority groups speed up the implementation of these jobs quotas in public offices.
It turns out that the answer is Yes, if it for the reservation of the scheduled caste groups, there is a big and strong impact on these job quotas:
Coefficients are large, and very significant
However, welfare spending for the SC groups have no discernible effects
But for welfare spending on the ST group, having ST reservation does in fact increase welfare spending.
Therefore, having reservation has different effects on targeted policy.
For job quotas, SC leaders push for it, and ST leaders not.
For welfare spending, ST leaders seem to push for it, and SC leaders not
Therefore, there is heterogeneity in response. There are two minorities and they are behaving differently once you give them space in the legislator.
So, broadly speaking, there is an effect, but important to keep aware that there is a lot of heterogeneity.
SC/ST Reservations
Chin and Prakash (2011) empirically asses the impact of political reservation in the state-level legislature on overall state-level poverty.
Follow up paper to Pande (2003) same data, same strategy, same identification as Pande
Their y variable is different, they use poverty
Their point is that both these minority groups come from a historically disadvantaged arena.
So, is there an effect on reservation on poverty?
Again, they find heterogeneity in response.
Findings: ST Reservation reduces poverty while SC reservation has no impact.
Comparing this with Pande's Table 7, SC reservation was not affecting SC welfare spending but ST reservation spending was. Welfare spending is typically for the poor households which is how we explain it.
"These results are in line with Pande (2003), who found that ST and SC reservations in state legislative assemblies have different policy effects, with the former increasing spending on ST welfare programs and the latter increasing the number of state government jobs set aside for minorities. Welfare programs primarily target the poor whereas reserved jobs are open to even better off minorities, so it is not unexpected given Pande's results that ST reservation would reduce poverty whilst SC reservation would not.
Some Related Issues: A Theory
Is it then possible that the gains to some minority groups (say the SC's) from political reservation may not be uniform?
"So now you change the identity of the leaders in the legislature, and different stuff happens, is this surprising?"
More importantly you have two minorities, for one of them you get a big effect on job quotas, but nothing on poverty or welfare. But for the other its the other way around...
How do you reconcile this? Why is this happening, what is the underlying theoretical justification?
Mitra (2017) presents a theory to explain this, although heavily critiqued.
Adds an explanation to the heterogeneity to Pande and Chin and Prakesh
It is simple, and we can explain it using the skills that we have developed over the last couple of weeks, it builds upon the probabilistic voting model (being a swing group is what gives you a lot of transfers), and that is what this paper utilities.
The mechanism underlying the theory builds on these key insights from these well-known models of redistributive politics.
When parties compete for votes by promising transfers across different groups, the group with the least ideological bias (swing voters) is most favoured by all parties. Politicians understand that these are the people to go for, and give them relatively more transfers.
Mitra (2017): Homophily
Mitra (2017) creates a two-stage game:
Introduced an initial first stage, fielding candidates, the party must choose from these field candidates from ethnic groups.
Lets say that there are two ethnic groups, L (low castes) minority, and H majority (high castes)
Free choice in states which have not been reserved, and not in those which have been. Therefore, with reservation you are forced to pick the minority.
In the second stage, the fielded candidates propose redistribution policies making promises and targeting more swing voters.
The innovation is to attempt to understand what goes into ideology.
When you vote for the candidate you are looking at the characteristics of the candidate alongside the characteristics of the party, a combination of these two things.
Therefore, every voter feels a positive bias from their own ethnic group.
So, if an L group (minority) sees a party fielding a candidate from their L group, all else constant, they will vote for that group.
Homophily: all else constant, when you see someone from your own group (race, language, religion), you will vote for that person.
Some assumptions:
Two parties, L and R (one more pro labour other pro capital).
What reservation does, is by law it makes the two candidates look identical on the ethnic dimension. When you see two candidates in a reserved constituency it doesn't matter if you are from the majority or minority group, the two candidates look identical. In that case, voting takes place on party-wise ideology.
Pro L votes for L
Pro R votes for R
However, in a constituency which is not reserved, any iteration of candidates can occur.
One could be fielding a candidate from the majority the other from minority, and that creates the tension.
So, reservation affects the one component of the ideological bias and can therefore change who the swing group is. The moment you change the ideological bias, you change the identity of who swing is.
So, in a reserved district you have both parties with identical candidates, the ethnic bias is irrelevant and you vote on your party affiliation. Clearly, the swing group is a middle-class (assuming L is for low-income, and R is for Rich)
However, if you contrast this with the case that the case that the L group fields a minority candidate and R fields a majority group. Here in determining who is swing, both components of the ideology is going to matter.
If I am a poor voter, but I belong to the majority group and looking at this combination. So, poor so likes the party but doesn't like the candidate as much and vice versa.
This makes the voter indifferent, a swing voter, a poor-majority voter, and because they are swing, they will be targeted and get most of the benefits.
Once you introduce reservation you are killing this and taking us to a world where you are benefitting middle income groups.
This is what is happening in SC groups, which is why they are not increasing welfare spending, they don't care about the poor. They are increasing job quotas to really go to the middle-income in the job quotas.
If you know why these people have an ethnic bias, would would anyone ever field a candidate from the minority group, is this not shooting yourself in the foot?
Parties want to have an inclusive image because maybe they believe in it or because it is convenient.
You are not short of examples to show parties fielding candidates from minority groups in countries like the Uk and others.
The moment you bring in this inclusive aspect you justify it, closing the story.
Gender Quotas in Politics
No need to argue that it is causal, it genuinely is.
Chattopadhyay and Duflo (2004) provide evidence from a Randomised Policy Experiment in India.
In the mid early 1990's the Indian constitution realised that what they do for minorities they should do for women as well. Village councils in India have authority over local public goods decisions so should implement this too.
In 1993, an Indian Constitutional amendment mandated representation for women and minorities.
They fully randomised by looking at village councils in a state and every third village council will be headed by a women, ignoring all the features of the village.
This is an ideal setting as you can look at the effects of a female leader in a village council on budgeting.
How do you know that the spending is representing the preferences of the women vs men?
The researchers do a survey to make use of revealed preferences. They ask "what have you not been happy about in terms of spending at the village level?" And they find different answers dependent on gender
Preferences over Local Public Goods
The researchers go to two states, one district in each of the two states (caveat), not a big sample but get away with it as it is fully randomised.
In both of these places, West Bengal and Rajasthan it is found that women strongly prefer drinking water. Often as women typically have to travel miles and miles to have drinking water in their homes, and feel like more should be done to have this closer to their villages.
Men appear to have different ideas as those in West Bengal prefer education and irrigation (crops), whereas men in Rajasthan preferred roads to bring their goods and crops to the markets.
No men bothered with drinking water as it was not part of their household chores.
The researchers look in areas reserved for women, and expect that we should see more investment in drinking water everywhere, less investment in school and irrigation in West Bengal and less investment in roads in Rajasthan.
The Main Findings
The results suggest that there is more spending on public goods which show preferences of women.
These results are not reverted in the second cycle as women elected for the second time invest in a very similar way to women elected in the first cycle.
There is no "backlash" in places where men come back in power after the end of reservation. These men did not rule back on the changes.
There is another study, they find that after this men revise their views on women leaders that they may better or at least no worse than men.
While there is clearly a redistribution towards women, one cannot conclude that the allocation is welfare improving as this depends on the preferences for roads, schools, wells, etc.
Page updated
Google Sites
Report abuse