This is our 2nd blog post for our Job Market Paper Series blog for 2025-2026.
Roshan Saha is a PhD candidate in Applied Economics at Auburn University. He works at the intersection of caste, climate and development economics. His dissertation examines the impact of caste on access to credit, and how rainfall shocks affect wages of agricultural laborers in India.
Imagine you are a farmer in rural India. The monsoon season is approaching, and your livelihood depends on it. But contrary to the assumptions of many economic models, you likely don’t have access to 30 years of precise rainfall data to calculate long-term averages. Instead, you remember the drought from two years ago or the flood from last year. You make decisions based on what you’ve experienced recently: behavior economists and psychologists call ‘recency bias’ or availability heuristics (Tversky & Kahneman, 1974).
In India, approximately 55% of net sown area relies on rainfall (Shagun, 2023). Despite significant improvements, many farmers still lack access to real-time, actionable forecasts. There are various reasons for the low adoption of modern rainfall forecast infrastructure: accessibility, difficulty in interpretation, and trust and risk aversion (Hegde, 2023). This information gap forces them to rely on heuristics to form expectations about the weather.
In my job market paper, I (along with Mykel Taylor and Sunjae Won) ask: How do rainfall shocks—defined by recent experiences rather than long-term historical averages—affect the wages of male and female agricultural field laborers in India? To explain the possible channels through which this effect operates, we also investigate the role of irrigation infrastructure and agricultural markets, at the district-level, in moderating the effect of rainfall shock on wages.
Redefining Rainfall Shocks
Most existing economic literature defines a ‘rainfall shock’ by comparing observed rainfall to historical (approximately 30-years) averages. Scientifically, this is the most appropriate method to estimate shocks. But it assumes farmers can accurately retrieve information regarding historical rainfall distributions. When operating under uncertain conditions, this assumption is not plausible, and agents often weigh recent events more heavily (Malmendier & Wachter, 2024).
To capture this reality, I propose an alternate measure of rainfall shocks using deviations from the previous 3-year rainfall distribution in a district during the sowing and harvest seasons (implying 6 observations for each district-year). The choice of 3-year lag is based on findings from an earlier study suggesting that optimal lag-length of the effect of monsoon rainfall shocks on yield, wages, and food prices in India is 3 years (Brey & Hertweck, 2022); and data constraints. I define a positive (or negative) shock based on whether current rainfall exceeds the 80th percentile (or falls below the 20th percentile) of the previous three years’ distribution. Specifically, the shocks are defined for two critical phases: the sowing season (May–July) and the harvest season (October–December). This is calculated by aggregating the rainfall across the two seasons and then identifying deviations from the 80th and 20th percentiles of the rolling 3-year historical rainfall distribution in each district.
Using information on daily wages, crop area, irrigated area, monthly rainfall and number of agricultural markets from the TCI-ICRISAT database, we construct a district-level panel dataset comprising 542 districts across India from 1993 to 2017. I use a two-way fixed effects model to estimate the impact of rainfall shocks on wages of agricultural laborers.
Impact of the Shocks
My analysis reveals that the sequence of rainfall shocks is very crucial to understanding their impact on rural agricultural labor markets- defined at the district level (Jayachandran (2006); Kaur (2019)). Male field laborers face the steepest wage penalties under a negative sowing–positive harvest rainfall shock combination, whereas female field laborers experience the most pronounced wage reductions under a positive sowing–negative harvest rainfall shock combination. b However, the negative sowing and positive harvest combination has a significant effect on both male and female wages. When districts experience dry shock during sowing followed by excess rain during harvest, wages decline by 4.1 percent for female and 5.2 percent for male field labor.
Why is this specific combination so salient for wages? A dry sowing season hampers crop establishment, reducing the demand for labor early in the season. Then, when the harvest arrives, excessive rain causes waterlogging and crop damage, further suppressing the demand for harvesting labor. To identify the mechanism driving this wage decline, I complement the district-level analysis with household data from the India Human Development Survey (IHDS). Using an agricultural household model, I account for the fact that these households are both producers and consumers, meaning their labor supply and wage outcomes are jointly determined. Through a simultaneous equation model (3SLS), I find that the ‘income effect’ dominates the ‘substitution effect’ during a negative sowing/positive harvest shock. The loss in farm profits drives households to increase labor supply resulting in lower wages.
Potential channels
The paradox of irrigation
One might expect irrigation to protect workers from these shocks. If a farmer has a tube well, surely a dry sowing season matters less? On the contrary, I find that irrigation intensifies the negative impact on wages. In districts with high tubewell irrigation coverage, particularly those relying on tube wells, the wage drop during shock years is steeper (Figure 1).
Why? Areas with better irrigation often attract migrant labor from neighboring districts (Mahajan, 2017). When a regional shock hits, districts with greater irrigation acreage might face excess labor supply as workers migrate in search of stable work. This oversupply of labor drives local wages down further than in less developed districts. It highlights an unintended consequence of infrastructure development: while irrigation protects yields, it may inadvertently increase labor market volatility for agricultural workers.
Presence of agricultural markets
I also explore the role of village agricultural markets in mediating the effect of rainfall shocks on wages. Before diving into the results, lets try to understand why this matters. The relationship between market power and prices that farmers receive has been well documented in the growing body of industrial organization literature (Bekkerman & Taylor, 2020; Chatterjee, 2023; Sexton & Xia, 2018; Sexton & Zhang, 2001). The general conclusion is that lower market power, greater competition, and easier and/or cheaper access to markets improve the price that farmers receive for their produce. Such improved prospects for farmers when faced with favorable conditions to sell their output can also translate to higher wages for agricultural labor (Jacoby, 2016). Jacoby (2016) find that nominal wages for manual labor across rural India increase with higher agricultural prices. If access to markets can ensure better prices, then we can argue that wages might increase in such situations (we also explore the wage-price pathway explicitly in the paper but don’t discuss it here for brevity).
Figure 2 shows the effect of the number of agricultural markets in a district on wages. We find that relative to districts experiencing normal rainfall, the wages are lower in districts with fewer markets (25th percentile), but higher in districts with more markets (90th percentile and beyond) when faced with the negative sowing and positive harvest rainfall shock combination.
Policy Implications
The findings from the paper offer three key takeaways for policymakers:
- Accessible forecasts: Since farmers rely on recent memory (heuristics), providing actionable, real-time rainfall forecasts could help them make better ex-ante decisions, potentially smoothing out these shocks.
- Rethinking Safety Nets: Since irrigation doesn’t protect wages (and may even worsen wage drops due to migration), employment guarantee schemes need to be responsive to rainfall shock patterns, even in well-irrigated districts.
- Market Access: Results from my paper suggest that promoting competition and market access can act as a buffer, preserving the value of labor even in bad years.
By understanding how farmers perceive rainfall shocks—rather than how models assume they do—we can design better safety nets for the millions of workers whose livelihoods depend on the monsoon. A recent study has shown that access to real-time forecasts can help farmers make better decisions based on the changing weather (Sanders, 2025). The estimates from my job market paper can be used as a benchmark to compare the effects of access to real-time forecasts on labor supply decisions with those based on alternative heuristics.
* Banner image was generated using Perplexity AI’s image generating model. The image is rendered in Claude Monet’s impressionist style, depicting rice field workers during monsoon season in India.
