Working papers compilation - I

1. Outsourcing creates a trade-off - outsourced workers experience large wage declines while domestic outsourcing may raise aggregate productivity. This paper finds, 

Three implications arise. First, more productive firms are more likely to outsource to save on higher wage premia. Second, outsourcing raises output at the firm level. Third, contractors endogenously locate at the bottom of the job ladder, implying that outsourced workers receive lower wages. Using firm-level instruments for outsourcing and revenue productivity, we find empirical support for all three predictions in French administrative data. After structurally estimating the model, we find that the rise in outsourcing in France between 1996 and 2007 raised aggregate output by 3% and reduced the labor share by 0.7 percentage points. A 9% minimum wage increase stabilizes the labor share and maintains two thirds of the output gains.

The point is then about an appropriate minimum wage that can stabilise labor share without significantly denting output gains.   

2. Another paper discusses the economic, social and development impact of Covid 19 by summarising the findings of various studies done so far. It has a nice summary of all the various kinds of micro-impacts, especially across low and middle income countries (LMICs). 

3. One more paper highlighting the importance of access to opportunities in the form of big push like investments to help people break out of the poverty traps. The paper studies a 11 year panel in rural Bangladesh on the impact of an asset transfer and finds,

People stay poor because they lack opportunity. It is not their intrinsic characteristics that trap people in poverty but rather their circumstances. This has three implications for how we think about development policy. The first is that big pushes that enable occupational change can play a role in alleviating the global poverty problem. Small pushes will work to elevate consumption but will not free people from the poverty trap. The magnitude of the transfer needed to achieve occupational change may be much larger than is typical with current interventions, though importantly it can be time-limited. The fiscal cost of permanently getting people out of poverty through a large, time-limited transfer might therefore actually be lower than relying on continual transfers that raise consumption but have no effect on the occupations of the poor.

The second is that big push policies can have long-lasting effects. Our analysis of long-run dynamics indicates that the asset, occupation and consumption trajectories of above-threshold beneficiaries diverge from those of below-threshold beneficiaries over time. This finding is important as it indicates that, by engendering occupational change, one-time pushes can have permanent effects.

The third is that poverty traps create mismatches between talent and jobs. We have shown that misallocation of labor is rife among the poor in rural Bangladesh. Indeed, we show that the vast majority of the poor in rural Bangladesh are not engaged in the occupations where they would be most productive. They are perfectly capable of taking on the occupations of richer women but are constrained from doing so by a lack of resources. The value of eliminating misallocation is an order of magnitude larger than the cost of moving all the beneficiaries past the threshold. This is important as it implies that poverty traps are preventing people from making full use of their abilities and indeed it is the mass squandering of people’s abilities that is the key tragedy of poverty.

Its empirical findings comparing across programs,

Assuming the household works each of the 100 days they are entitled to, the value of NREGA is 0.13 of annual per-capita expenditure. BRAC typically offers entry microloans between 100 USD and 200 USD, which correspond to 0.18 and 0.3 of average annual per-capita expenditure. Thus, two of the main programs designed to tackle poverty are too small-scale to make a long-term difference for the majority: our simulation suggests that they would allow fewer than 20% of households to escape poverty... In a first set of simulations, we resimulate the model under the assumption that all households are given a transfer equal to an increasing percentage of annual per capita consumption expenditure, until the point at which misallocation equals zero. This exercise suggests that the value of misallocation — measured as before against the maximum payoff available at the upper mode of the distribution of productive assets excluding land — would be zero if all ultra-poor households were given a transfer equal to 3.95 times the average level of baseline per capita consumption expenditure among ultra-poor households.

It's headline policy finding,

Our results point to the existence of a poverty threshold such that households with a starting level of productive assets below that threshold are trapped in poverty while households who are able to get past the threshold accumulate capital and approach the asset level of the richer classes. This allows them to switch occupations from casual laborers to the more productive business activity of livestock rearing, which in turn facilitates further asset accumulation. The existence of such a poverty threshold has important implications for policy design. Transfer programs that bring a large share of households above the threshold will see large effects on average, while transfers that fall short of this might have small effects in the long run.

The takeaway is that a large enough cash or asset transfer can provide the big push to get people over the threshold and into an enabling path to access different livelihood opportunities. There are at least two problems. One, the fiscal cost of such transfers (3.95/0.13 = 30 times the NREGS transfer) is prohibitive and clearly off the table. Two, more importantly, the economic system's ability to absorb such large shocks (even if staggered in a reasonable manner) by providing the requisite economic opportunities in a sustainable manner is deeply questionable. 

And I am not even talking about the numerous and unanticipatable second and further order consequences of such large asset or cash transfers. 

The point is that cash or asset transfer based pathways out of poverty are at best marginal and unscalable interventions and there is no substitute to sustained economic growth and broad-based development for poverty elimination. 

4. This paper examines the impact of distortions in land rental markets across Indian states on their agriculture productivity. In 2010, the real-value added per Indian worker in non-Agriculture activities was 32% of that in the US, whereas the ratio was just 5% in use of agriculture workers. Besides, the variation in GDP per workers in agriculture across states in 2011-12 is a factor of 13.5. The paper's findings,

First, we show that an efficient reallocation of land can substantially increase agricultural productivity in all states, even relative to Punjab, the state with the least distorted land market in our sample. On average, an efficient reallocation of land increases agricultural productivity by 33 percent (15 percent relative to Punjab). In Tamil Nadu and Karnataka, the increase in agricultural productivity is 89 and 49 percent (63 and 34 percent relative to Punjab)... Such an increase in agricultural TFP would have a much larger effect on agricultural labor productivity because of the reallocation of labor away from agriculture and other productivity enhancing effects such as better selection into agriculture, investment in productivity, the adoption of modern technologies, among others... Second, we decompose the contribution between farm-and state-specific distortions and find that farm distortions contribute to about one-third of the reallocation gains, whereas state-level land wedges contribute the remaining two-thirds. We also show that an efficient reallocation of land would involve substantial increases in both the share of farmers renting (participation in the rental market) as well as the share of land operated by the most productive farms... The largest TFP gains are in states with the least active rental markets.

The paper has an informative table summarising the status of tenancy reforms in various Indian states, including the nature of restrictions on leasing land.

5. How does going public impact the performance of companies?

Public attention to a firm may provide valuable monitoring, but it may also have a dark side by constraining management’s decisions and distracting it. We use inclusion in the S&P 500 index as a positive shock to public attention. Media coverage, Google searches, SEC downloads, SEC comment letters, shareholder proposals, analyst coverage, and lawsuits increase following inclusion. Post-inclusion performance falls and is negatively related to the increase in attention. Included firms’ investment and payout policies become more similar to those of index peers and the increase in similarity is positively related to the size of the attention increase.

6. The moral hazard from seat belt use is more than offset by its safety benefits, 

Using data from the Fatality Analysis Reporting System for the period 1983-1997, Cohen and Einav (2003) found that mandatory seatbelt laws were associated with a 4 to 6 percent reduction in traffic fatalities among motor vehicle occupants. After successfully replicating their two-way fixed effects estimates, we (1) add 22 years of data (1998-2019) to capture additional seatbelt policy variation and observe a longer post-treatment period... investigate pre-treatment trends and explore lagged post-treatment effects. Consistent with Cohen and Einav (2003), our updated estimates show that primary seatbelt laws are associated with a 5 to 9 percent reduction in fatalities among motor vehicle occupants.

7. Gabriel Kreindler has a paper examining the likely impact of congestion pricing on traffic congestion in Bangalore,

I study the peak-hour traffic congestion equilibrium in Bangalore. To measure travel preferences, I use a model of departure time choice to design a field experiment with congestion pricing policies and implement it using precise GPS data. Commuter responses in the experiment reveal moderate schedule inflexibility and a high value of time. I then show that in Bangalore, traffic density has a moderate and linear impact on travel delay. My policy simulations with endogenous congestion indicate that optimal congestion charges would lead to a small reduction in travel times, and small commuter welfare gains. This result is driven primarily by the shape of the congestion externality. Overall, these results suggest limited commuter welfare benefits from peak-spreading traffic policies in cities like Bangalore.

The relative lack of impact from congestion pricing in Bangalore is understandable and important to be borne in mind. In most developing country contexts, infrastructure augmentation by way of new roads, widenings etc continue to remain relevant and higher priority than ideas like traffic congestion. This however does not mean that traffic congestion policies are not important. In specific areas, where the demand elasticity of response is likely higher, congestion pricing can have significant impacts. 

8. A new working paper finds that Amazon systematically manipulates its algorithms to favour its private label brands in its search results.

We study whether Amazon engages in self-preferencing on its marketplace by favoring its own brands (e.g., Amazon Basics) in search. To address this question, we collect new micro-level consumer search data using a custom browser extension installed by a panel of study participants. Using this methodology, we observe search positions, search behavior, and product characteristics. We find that Amazon branded products are indeed ranked higher than observably similar products in consumer search results... All specifications shown, as well as a number of additional checks, including specifications with interaction terms and machine learning approaches, indicate that carrying an Amazon brand is a meaningful predictor of greater prominence in search. The effect of Amazon brands tends to be 30% to 60% as large as the effect of sponsoring.