In a nice, recent paper Achyuta Adhvaryu, Namrata Kala, and Anant Nyshadham take a look at how air pollution hurts productivity and what effect, if any, managers can have in mitigating these effects. The short answer is yes, pollution hurts worker productivity, and yes, managers with certain qualities do a better job of mitigating these effects as they manage their workers. That’s the short story, but how they get there is quite fascinating.
Let’s start by talking pollution. Adhvaryu and co. are focusing on fine particulate matter (PM). This is particulate matter that is less or equal to 2.5 microns in diameter, and it’s not very good for you. Basically, it accumulates in your lungs, making it hard to breathe, and it’s associated with a range of long term and also short term health problems. It comes mostly (in this setting) from industrial combustion and automobile exhaust.
On to production. Adhvaryu and co. are working with a very large garment manufacturer in Bangalore. Workers in these factories work on production lines, but have individual tasks, which happen in sequence until a finished garment comes out at the end (see the details in the paper if you want to look at your shirt in a whole new way). This gives them very granular data on individual worker output. And they can benchmark this off of international standards (yes, they exist!) for completing individual garments. Finally, one key way in which supervisors matter for productivity is in switching workers around. Given shifting demand (lines don’t produce the same garment endlessly), worker absence, and the like, workers move around a lot across, but also within lines.
Now, to get to the analysis. First of all, it turns out that Bangalore has pretty high levels of PM. The mean is around 65 micrograms per cubic meter, with a standard deviation of 45. Southern California averages between 10-20 micrograms per cubic meter. To get a sense of what conditions are like for the workers, Adhvaryu and co. install meters near the production lines they are studying (cool measurement gear alert!). Since fine PM is not a direct by-product of clothing manufacturing (coarse PM is), this gives them granular (hourly) pollution data to match to the productivity levels.
Adhvaryu and co. present a range of estimates, using some combination of line, worker, month, day, and hour-of-day fixed effects (as well as controls for coarse PM) and show that a one standard deviation increase in PM leads to about a 3.6% drop in productivity. Adhvaryu and co. then look at this effect by the difficulty of tasks and find that more difficult tasks are more heavily affected by fine PM, while simple tasks show lower or even positive effects from PM. This could suggest that supervisors might be reallocating workers in response to the effects of PM on individual worker productivity (keep in mind that base health is idiosyncratic and hence PM effects will also be idiosyncratic). So, Adhvaryu and co. look at the efficiency/PM gradient by individual production lines to get at supervisor effects and they find negative slopes for some, but not for others.
Bring on the supervisors! Adhvaryu and co. are working with administrative data, so they don’t have a wealth of worker characteristics. Nonetheless, they come up with two interesting measures. The first is supervisor experience – supervisors are experienced if they have been with the factory the median number of years (1.5) or more. The second measure they call relatability. These are supervisors who are younger than the median age of supervisors, have less than a high school education, are native speakers of the local language, and come from Bangalore.
It turns out that having a more experienced supervisor reduces the impact of fine PM on efficiency by 22-35%. And having a relatable supervisor reduces the impact of PM by 50 to 80%. So supervisors matter a lot for the impacts of these pollution shocks on productivity. Adhvaryu and co. then dig into the precise mechanism through which the supervisor might be having these effects. They start by looking at the number of times a worker is assigned a task that is different than the task she was doing in the previous hour. Overall, with increasing PM, worker reassignment across tasks increases. And this reassignment is concentrated among workers working for more relatable and more experienced supervisors. So the boss who is more in tune with his workers and/or has more experience, is able to see when workers are flagging and moves them around to compensate.
This is a neat paper. On the methodological level, it’s a nice example of how you can get creative with collecting super granular production and pollution data and making good use of existing administrative data. On the results level, it shows us a pretty clear causal chain of how pollution can affect productivity and how management matters to help offset these effects.
Let’s start by talking pollution. Adhvaryu and co. are focusing on fine particulate matter (PM). This is particulate matter that is less or equal to 2.5 microns in diameter, and it’s not very good for you. Basically, it accumulates in your lungs, making it hard to breathe, and it’s associated with a range of long term and also short term health problems. It comes mostly (in this setting) from industrial combustion and automobile exhaust.
On to production. Adhvaryu and co. are working with a very large garment manufacturer in Bangalore. Workers in these factories work on production lines, but have individual tasks, which happen in sequence until a finished garment comes out at the end (see the details in the paper if you want to look at your shirt in a whole new way). This gives them very granular data on individual worker output. And they can benchmark this off of international standards (yes, they exist!) for completing individual garments. Finally, one key way in which supervisors matter for productivity is in switching workers around. Given shifting demand (lines don’t produce the same garment endlessly), worker absence, and the like, workers move around a lot across, but also within lines.
Now, to get to the analysis. First of all, it turns out that Bangalore has pretty high levels of PM. The mean is around 65 micrograms per cubic meter, with a standard deviation of 45. Southern California averages between 10-20 micrograms per cubic meter. To get a sense of what conditions are like for the workers, Adhvaryu and co. install meters near the production lines they are studying (cool measurement gear alert!). Since fine PM is not a direct by-product of clothing manufacturing (coarse PM is), this gives them granular (hourly) pollution data to match to the productivity levels.
Adhvaryu and co. present a range of estimates, using some combination of line, worker, month, day, and hour-of-day fixed effects (as well as controls for coarse PM) and show that a one standard deviation increase in PM leads to about a 3.6% drop in productivity. Adhvaryu and co. then look at this effect by the difficulty of tasks and find that more difficult tasks are more heavily affected by fine PM, while simple tasks show lower or even positive effects from PM. This could suggest that supervisors might be reallocating workers in response to the effects of PM on individual worker productivity (keep in mind that base health is idiosyncratic and hence PM effects will also be idiosyncratic). So, Adhvaryu and co. look at the efficiency/PM gradient by individual production lines to get at supervisor effects and they find negative slopes for some, but not for others.
Bring on the supervisors! Adhvaryu and co. are working with administrative data, so they don’t have a wealth of worker characteristics. Nonetheless, they come up with two interesting measures. The first is supervisor experience – supervisors are experienced if they have been with the factory the median number of years (1.5) or more. The second measure they call relatability. These are supervisors who are younger than the median age of supervisors, have less than a high school education, are native speakers of the local language, and come from Bangalore.
It turns out that having a more experienced supervisor reduces the impact of fine PM on efficiency by 22-35%. And having a relatable supervisor reduces the impact of PM by 50 to 80%. So supervisors matter a lot for the impacts of these pollution shocks on productivity. Adhvaryu and co. then dig into the precise mechanism through which the supervisor might be having these effects. They start by looking at the number of times a worker is assigned a task that is different than the task she was doing in the previous hour. Overall, with increasing PM, worker reassignment across tasks increases. And this reassignment is concentrated among workers working for more relatable and more experienced supervisors. So the boss who is more in tune with his workers and/or has more experience, is able to see when workers are flagging and moves them around to compensate.
This is a neat paper. On the methodological level, it’s a nice example of how you can get creative with collecting super granular production and pollution data and making good use of existing administrative data. On the results level, it shows us a pretty clear causal chain of how pollution can affect productivity and how management matters to help offset these effects.
Join the Conversation