This is the 6th in this year’s job market series
In the narrow, residential streets of Tiruppur, South India, nearly every other door opens to a small garment factory. At 9 am, when these factories open, many sewing machines remain idle. Their tailors haven’t arrived yet.
Tiruppur is not unusual: worker absences and lateness are common across the developing world. This matters more as firms grow, because moving from individual production to assembly lines requires worker coordination. I study whether late arrivals reduce worker and firm output, and why firms tolerate this behavior.
Factories in Tiruppur are organized as production lines which need coordination
My job market paper focuses on small garment factories in Tiruppur, located inside residential neighborhoods. These factories typically employ 6 workers, 85% of whom live within a 15-minute commute of their firm. Production in these firms is organized in batch processing lines where each worker performs one operation (e.g., joining shoulder seams) and passes bundles of garments to the next worker. This organization of production requires coordination between workers. For example, if the worker who joins shoulder seams arrives late, workers performing later tasks sit idle. This can delay order completion.
However, workers commonly arrive late. Using novel, high-frequency data on the internal functioning of these firms, I show that workers arrive late at work on 43% of worker-days, and arrival times range from 3 hours before to 3 hours after firms’ start time. (Figure 1) These late arrivals reflect workers exercising their preference for flexible arrival times: workers are on average 37 years of age, 45% are female, 94% are married, and 84% have at least one child younger than 14 years of age. Thus, these workers are at a stage of life with substantial care responsibilities at home. Even though owners can learn which workers tend to be late, this unpredictability makes it harder to plan production around worker arrival times.
The puzzle: firms allow lateness even when production lines require coordination
This setting presents a puzzle. The production line setup implies gains from workers arriving on time, but workers arrive late and unpredictably. This may imply either (1) there are no gains from timely worker arrival, or (2) firms cannot enforce timings or provide large-enough incentives given workers’ value for flexibility.
Firm owners believe that late arrivals hurt their firm and would like workers to arrive at the same time. Yet, they do little to coordinate arrival times. First, most pay workers operation-specific piece-rates with no direct incentives for finished garments. This contract does not provide explicit incentives to arrive on time. Second, firms neither reward punctuality with bonuses nor penalize lateness.
An experiment to test the puzzle
I designed an experiment with 92 firms comprising 600 workers to test whether reducing lateness affects worker and firm outcomes. To do so, I offered workers in 46 randomly selected treatment firms a daily incentive worth 20% of their average daily earnings if they arrived before the start time for four weeks. Control workers received unconditional transfers equivalent to realized incentive payments in treatment firms, so comparisons between treatment and control outcomes don’t reflect income effects. The experiment lasted 6 weeks with 2 pre-incentive weeks and 4 incentive weeks.
Small, informal firms in developing countries often do not maintain any personnel or production records. Over six weeks, I collected daily data on worker attendance, arrival times, work hours, along with weekly firm output, quantity of new orders procured, and their delivery timelines. Uniquely, I recorded each worker’s output for each operation stitched in the day. This helps study the precise mechanisms behind the relationship between worker lateness and firm outcomes.
Comparing production outcomes between treatment and control firms tests whether there are gains from coordinating worker arrival times. Since the bonus was offered as a choice to workers, their decision to arrive on time reveals how much they value flexibility. Workers who forgo the incentive reveal that they value flexibility at more than 20% of their current earnings.
Incentives induced late workers to arrive on time, but output increased for everyone
Workers in treatment firms are nearly 40% more likely to arrive on or before time (21 p.p., compared to a control mean of 58%). This implies that 4 out of 5 arrive on time during incentives compared to 3 out of 5 workers during pre-incentive weeks, conditional on being present. This increase is driven by workers who typically arrived late before the experiment (henceforth, “late”). Workers who were early at baseline (“early” workers) did not change their arrival times. There is no change in whether workers show up to work at all, and a small increase in daily work hours which approximately offsets how late workers were before the incentives.
As a result, workers in treatment firms produce 24% more output per week, equivalent to ~930 additional pieces per week. Output rose by a similar amount for all workers, including “early” workers who did not change their arrival times. This suggests treatment generated positive spillovers within the firm.
Spillovers are concentrated in firms where late arrivals created bottlenecks in production lines
These spillovers could either reflect a better coordinated production line, or other potential mechanisms such as team morale. I use the positioning of “late” workers on the production line to identify firms where lateness is more likely to create bottlenecks along the line. I classify firms as ‘Bottleneck’ firms if any worker who performs upstream tasks in the production line at baseline is a “late” worker. In remaining firms, “late” workers perform downstream tasks (‘Non-Bottleneck’ firms). Incentives improve timely arrival for “late” workers in both sets of firms, while “early” workers don’t change their arrival times.
Effects on worker output are larger in Bottleneck firms (Figure 2). Relative to control, workers in these firms produce 32% more in weekly output, compared to 12% more in Non-Bottleneck firms. Strikingly, spillovers to “early” workers are only present in Bottleneck firms: both “early” and “late” workers gain ~33% in weekly output. In Non-Bottleneck firms, gains are concentrated among “late” workers only. Therefore, spillovers reflect gains from a better coordinated production line.