Syndicate content

Add new comment

Party politics affects the spatial distribution of growth: Guest post by Paul Novosad

This is the sixth of our series of posts by PhD students on the job market.

Does politics affect the distribution of local economic growth? Do politicians use their power on net to promote growth or extract rents? My job market paper (“Politics and Local Economic Growth: Evidence from India,” joint with Sam Asher), presents evidence that politician identity has a big effect on local economic outcomes in India, and that regulation plays a key role in politicians' influence on the economy.

There is a great deal of evidence that in many countries, political considerations play an important role in the allocation of government inputs, such as public goods, operating licenses or credit from state banks. Connections between big business and politicians are the norm in developing countries. Nearly all of the research in this area focuses on the allocation of specific government inputs, or on the gains to individual politically connected firms. There is considerably less evidence on whether political behavior has an important effect on the broader economy.

We look at state elections in India, to test whether areas that are represented by politicians in the state majority party, the party that controls government resources, have better economic outcomes than areas represented by politicians with less access to state power. The literature suggests that alignment of local politicians will have an important effect on government inputs. Rather than measuring specific government inputs, we want to know the effect of the complete set of politicians' actions on total local economic outcomes. To preview our results, we find that political alignment has a very large effect on local employment and the share prices of local firms.

Data: Town/village-level census of Indian firms: 1990-2005

There are two main challenges to answering this question. The first relates to data. In general, economic data tends to be aggregated to units that do not correspond to political boundaries. In India, the standard economic datasets report firm outcomes at the level of districts, which are 10 times larger than state political constituencies, and thus difficult to match to state politicians.

With my coauthor, Sam Asher, I spent almost two years of my PhD building a time series dataset from the Economic Census of India, which is a complete enumeration of non-farm firms undertaken in 1990, 1998 and 2005. We built a location panel of 600,000 villages and 5,000 towns, with detail on non-farm employment and the industrial categorization of both formal and informal firms in these towns and villages. The dataset covers 24 million firms in 1990 and 42 million in 2005, and is the basis of several other projects of ours on local economic growth.

Empirical Strategy: Regression discontinuity around close elections

The second challenge is that political alignment may be correlated with unobservable factors that affect growth. For example, if a farmers' party is in the majority, rural locations are more likely to be represented by majority-aligned politicians. We don't want to argue that the difference between urban and rural growth is driven by politicians; clearly there are many unobservable variables that could bias inference here.

We control for unobservable factors by using a regression discontinuity design around close elections. Our identifying assumption is that when an election is very close, the random factors become important relative to the lead of any one candidate, and the winner is as good as randomly assigned. Our empirical strategy compares locations where majority-aligned politicians narrowly won with locations where they narrowly lost. We show that these locations are alike on all ex ante observables, and argue that they are likely to be alike on unobservables as well.

A concern with close election regression discontinuity is raised by Grimmer et al (2011), who find that powerful politicians in the U.S. are disproportionately the winners of close elections. If true, narrowly winning and losing candidates may in fact differ on unobservables, weakening our identification strategy. To address this concern, we run a battery of tests showing that constituencies narrowly won by aligned parties are similar on observables to constituencies that are narrowly lost. Second, because we focus on alignment to the majority party in a highly dynamic period of Indian politics, we can control for many of the characteristics that candidates would be likely to sort on. We find nearly as many incumbents from non-aligned parties as we do in aligned parties, suggesting that the identification strategy is valid even if incumbents are disproportionately able to win close elections. Adding controls for incumbency or membership in any of India's major parties does not affect our results.

Main Findings

·         Annualized employment growth is 1.6 percentage points higher, over a seven year period, in locations represented by majority-aligned politicians. This is big - it amounts to about 12% over the 7-year survey period.

·         Growth is fully accounted for by private sector firms.

·         No effects on public sector employment or construction of public infrastructure (we look at roads, electricity, schools, hospitals and irrigation), which are two of the areas on which we might expect politicians to have the greatest influence.

·         Stocks show cumulative abnormal returns of 12-15% in the month following a win for an aligned candidate. This suggests that the employment effect is indeed value creating, and affects a broad range of firm sizes, from the average small firm in the Economic Census, to large publicly traded firms.

·         Regulatory practice plays an important role. Effects are larger for firms that have more interactions with government officials. We describe this part in more detail.

Regulatory policy and regulatory practice

Using international data from the World Bank Enterprise survey, we classified industries according to their dependence on government officials. We use questions like “What % of time does senior management spend dealing with government regulation” and “Did you need to apply for an operating license in the last two years.” We apply this industry-level measure to firms in our dataset, and interact a “high government dependence” dummy variable with the standard variables in the regression discontinuity. We find that firms are significantly more affected by the affiliation of the local politician when they are in industries that have a high dependence on government officials. Some examples of these industries are alcohol, tobacco and mining. The first two are subject to high excise taxes where avoidance is likely high; and mining requires multiple approvals from many branches of government.

This result stands in contrast to the usual approach to regulation in the literature, which centers on which characteristics of regulatory policy are optimal. Our findings suggest that within a constant regulatory system, politicians are controlling which regulations are and are not enforced. This fits a local picture of politicians who have a measure of control over bureaucrats, and play key roles as mediators in their constituencies. We argue that India's high level of regulation effectively creates low cost opportunities for politicians to manipulate local enforcement for political gain. A caveat on this result is that the regulatory intensity of firms is not randomly allocated, so our results on the interacted regression could be driven by unobserved characteristics of highly regulated firms.

Finally, we find that our identified effect appears to be driven mainly by reduced growth in locations where the majority party narrowly lost, rather than increased growth in locations where the majority party won. This makes sense if regulation is a tool that more easily suppresses economic growth than promotes it. This effect is also consistent with a common story from India that a powerful party will seek to “punish” voters who throw out its incumbent. We discuss some other possible explanations for the effect in the paper.

In conclusion, we find that politics has an outsize impact on the distribution of local economic growth. Our research suggests an under-explored regulatory channel by which politicians in power are able to affect not only laws but how they are enforced. The opportunities that regulation creates for politicians may help explain why onerous laws on the functioning of businesses have persisted for so long in India.

Paul Novosad is a PhD student at Harvard University and is on the job market this year.