Manuscript

This is the author’s accepted manuscript without copyediting, formatting, or final corrections. It will be published in its final form in an upcoming issue of
Journal of Labor Economics, published by The University of Chicago Press. Include the DOI when citing or quoting: https://doi.org/10.1086/716347
Copyright 2021 The University of Chicago Press.

Does Minimum Wage Increase Labor Productivity?

Evidence from Piece Rate Workers*

Hyejin Ku†
University College London

This version: June 2021

Abstract: We examine worker effort as a potential margin of adjustment to a minimum wage

hike using unique data on piece rate workers who perform a homogenous task and whose
individual output is rigorously recorded. By employing a difference-in-differences strategy that
exploits the increase in Florida’s minimum wage from $6.79 to $7.21 on January 1, 2009, and
worker location on the pre-2009 productivity distribution, we provide evidence consistent with
incumbent workers’ positive effort responses.

Key words: minimum wage; incentive; effort; labor productivity

JEL Codes: J20, J38, M50

*
We thank anonymous referees, Jerome Adda, Joe Altonji, Alison Booth, Christian Dustmann, Hank Farber, Bernd
Fitzenberger, Bob Gregory, Peter Kuhn, Kevin Lang, Ed Lazear, Attila Lindner, Steve Machin, Xin Meng, Kjell
Salvanes, Chris Stanton, Uta Schoenberg, Kathryn Shaw, Lowell Taylor, Martin Weidner, Daniel Wilhelm and
seminar/conference participants at ANU, Cornell, NHH, UCL, IZA, EALE, NBER Summer Institute and AASLE, for
helpful comments and suggestions. The support of the Economic and Social Research Council (ESRC), under Project
Ref: ES/R005745/1, is gratefully acknowledged.
†
Department of Economics and CReAM, University College London, 30 Gordon Street, London, WC1H 0AX, and
Institute of Labor Economics (IZA). E-mail: [email protected].
This is the author’s accepted manuscript without copyediting, formatting, or final corrections. It will be published in its final form in an upcoming issue of
Journal of Labor Economics, published by The University of Chicago Press. Include the DOI when citing or quoting: https://doi.org/10.1086/716347
Copyright 2021 The University of Chicago Press.

1. Introduction

Worker effort as a potential margin of adjustment to a minimum wage was raised in early studies

such as Obenauer and von der Nienburg (1915) and Stigler (1946). Yet rigorous empirical

investigation on this issue has been lacking in the literature, despite the significant progress made

in recent decades on the minimum wage’s (dis)employment effect.1

In this paper, we employ a direct and high frequency measure of individual worker

productivity on a homogenous task to examine possible worker effort responses to a minimum

wage increase. In particular, we use personnel records from a large tomato farm in Florida—where

piece-rate workers hand-harvest tomatoes in the field—together with the change in the state

minimum wage from $6.79 to $7.21 on January 1, 2009. In piece rate settings, the employer must

make up any shortfall between a worker’s raw productivity (output in dollars/hour) and the

minimum wage for all work hours during a given pay period (in this context, one week).2 Hence,

firm’s compliance costs increase with the minimum wage, which may (at least in part) be offset

by the increased effort of low productivity workers.

This is a unique setting conducive to examining worker effort responses to a minimum

wage increase for several reasons. First, due to the pay scheme being piece-rate based, the

productivity of individual workers is rigorously recorded.3 Not only do the workers clock in and

out for each work spell, but an electronic system keeps track of their output in the field. Second,

the minimum wage increase of January 1, 2009 occurs within a given harvesting season (autumn

1
See Card and Krueger (1995), Brown (1999) and Neumark, Salas and Wascher (2014) for reviews, and Cengiz et
al. (2019) and Harasztosi and Lindner (2019) for more recent evidence.
2
Workers whose raw productivity (output in dollars/hour) is above the minimum wage get paid according to their
actual output.
3
One salient feature of this work environment is that the output of individual workers is readily observable, which is
conducive to the adoption of a piece rate pay scheme (Lazear 1986).

1
This is the author’s accepted manuscript without copyediting, formatting, or final corrections. It will be published in its final form in an upcoming issue of
Journal of Labor Economics, published by The University of Chicago Press. Include the DOI when citing or quoting: https://doi.org/10.1086/716347
Copyright 2021 The University of Chicago Press.

2008 season), which allows us to compare the same worker’s productivity before and after the

hike. Third, the nature of the task and workforce allows us to rule out other potential determinants

of worker productivity. In particular, hand-harvesting of fresh tomatoes is a low skilled, labor-

intensive process and there is little scope for technological adjustments (e.g. shift towards capital)

or innovation, at least in the short run (i.e. within season).4 In addition, due to the seasonal nature

of the harvesting task and high workforce turnover, firm investment in worker training is virtually

nonexistent and largely irrelevant in this sector.

In order to isolate the effects of worker effort from external determinants of labor

productivity (e.g. field lifecycle or weather), we employ a difference-in-differences (DID) strategy.

Similar to Mas and Moretti (2009), we first capture each worker’s “baseline” or “permanent”

productivity by estimating their fixed effects using data from outside our main estimation window.

We then look for possibly differential productivity changes of individual workers around January

1, 2009 by their baseline productivity. Using high productivity workers (who are always above the

minimum wage either in the old or new regime) to difference out the effects of farm-level common

shocks, we isolate the minimum wage-induced effort responses based on a disproportionate

productivity increase on the lower part of the fixed effects distribution when the minimum wage

increases. This is analogous to the approach in Cengiz et al. (2019) and Dustmann et al. (2020),

for instance, where workers on the upper part of the wage distribution are viewed as a “control”

when evaluating the effect of a higher minimum wage on low wage workers.

4
In the US the markets for fresh- and processing-tomatoes are entirely separate. Not only are different varieties of
tomatoes grown to serve each market but they are harvested differently. In particular, processing tomatoes (which
are common in California) are machine-harvested whereas fresh-market tomatoes are hand-harvested. The Florida
farm studied here serves the market for fresh tomatoes only and the incidence of hand harvesting is 100 percent.

2
This is the author’s accepted manuscript without copyediting, formatting, or final corrections. It will be published in its final form in an upcoming issue of
Journal of Labor Economics, published by The University of Chicago Press. Include the DOI when citing or quoting: https://doi.org/10.1086/716347
Copyright 2021 The University of Chicago Press.

We find evidence consistent with incumbent workers’ positive effort responses. As the

minimum wage increases by 6% ($6.79 to $7.21) on January 1, 2009, worker productivity (i.e.

output per hour) in the bottom 40th percentile of the worker fixed effects distribution increases by

about 4.6% relative to that in the higher percentiles. Examining the employment outcomes of

individual workers over time, we find that while low productivity workers have 6-10% lower

chances of being employed on any given day than high productivity workers, this existing

difference is not further amplified when the minimum wage increases. This lack of significant

employment effects attributable to the minimum wage hike may have several explanations. In a

competitive framework, the positive effort responses of subminimum wage workers should obviate

the need for reducing employment opportunities assigned to these workers. Moving beyond

competition, it is also possible that workers simply reacted to a perceived threat, even in the

absence of any actual pressure coming from the employer, and/or are driven by other motives such

as gift exchange (Akerlof 1982). Overall, our back-of-the-envelope calculation shows that the

increased productivity among the low fixed effect workers can offset about half of the projected

rise in the firm’s wage bill, suggesting a roughly equal sharing of the minimum wage cost between

the employer and the workers.

By taking advantage of rare data on piece-rate workers whose physical output (pieces per

hour) is rigorously recorded around a minimum wage hike where the piece rate itself remains the

same throughout, we are able to test for workers’ effort responses as a plausible channel of

adjustment to a minimum wage hike. Such responses are extremely difficult to detect in

observational data, since in most settings we lack data that repeatedly measure the same worker’s

productivity on the same task around a minimum wage hike. Although it is difficult to know the

exact extent to which the effort responses shown here will apply to other low wage settings, the

3
This is the author’s accepted manuscript without copyediting, formatting, or final corrections. It will be published in its final form in an upcoming issue of
Journal of Labor Economics, published by The University of Chicago Press. Include the DOI when citing or quoting: https://doi.org/10.1086/716347
Copyright 2021 The University of Chicago Press.

hypothesized effort responses do not rest on the pay scheme being piece-rate based. For instance,

a recent study by Coviello, Desseranno and Persico (2019) illustrates a minimum wage-related

productivity increase among salespeople of a retail chain, where the compensation scheme at use

is a base pay plus a performance-based commission. Moreover, even in settings where workers are

paid a fixed hourly wage, we know that the employer and co-workers can to a varying extent

assess/observe the productivity of different employees.5 It is the ability to tell apart low vs. high

productivity workers, and not the pay scheme per se, that dictates the relevance of the effort margin

as a possible response to minimum wage changes. An added advantage of our study is that it

provides rare insights on the labor market behavior and outcomes of US farm laborers, a relevant

yet understudied group when it comes to analyses of minimum wage impact. In particular, and as

shown in Appendix Table A.1, the wages of farmworkers are not dissimilar to those of workers at

fast food restaurants (see e.g. Card and Krueger 1994) or care homes (see e.g. Machin, Manning

and Rahman 2003), the subgroups often studied in leading papers in the minimum wage literature.

By providing clean evidence on the minimum wage effect on worker effort, we add to the

recent and growing literature that explores alternative channels (other than employment) through

which firms may absorb the rising labor cost associated with the minimum wage (see Clemens

2021 for a review). These channels include increased worker retention and reduced turnover

(Portugal and Cardoso 2006; Dube, Lester and Reich 2016; Gittings and Schmutte 2016), labor-

labor substitution (Giuliano 2013), changes in hiring standards (Clemens, Kahn and Meer 2021),

an increase in prices (Aaronson 2001; Aaronson, French and MacDonald 2008; Leung 2018;

5
Observable characteristics such as experience, for instance, may serve as a proxy for productivity. In Jardim et al.
(2018)’s evaluation of Seattle’s minimum wage, they find that all of the earnings increases from a higher minimum
wage accrue to the more experienced half of the low-wage workforce whereas the less experienced half saw no
significant change in earnings due to decreases in their hours worked offsetting their wage gains.

4
This is the author’s accepted manuscript without copyediting, formatting, or final corrections. It will be published in its final form in an upcoming issue of
Journal of Labor Economics, published by The University of Chicago Press. Include the DOI when citing or quoting: https://doi.org/10.1086/716347
Copyright 2021 The University of Chicago Press.

Harasztosi and Lindner 2019), and a decrease in profits (Draca, Machin and Van Reenen 2011;

Bell and Machin 2018). In particular, we speak directly to effort-driven labor productivity, by

employing reliable data on the physical output (pieces per hour) of harvesting employees around

an increase in the statutory minimum wage.6

This paper also relates to the personnel economics literature that explores how incumbent

workers’ productivity may be related to labor market conditions. In an earlier work, Rebitzer (1987)

showed that the level of unemployment raises productivity growth using US data at two-digit

manufacturing industries for 1960-1980. In addition, a recent work of Lazear, Shaw and Stanton

(2016) shows that incumbent workers may work harder during recession and when the

unemployment rates are higher. While similar in the usage of personnel records from a US firm,

Lazear, Shaw and Stanton (2016)’s study of recession effects focuses on the increased cost in case

of discharge for workers with a relatively long employment contract, whereas our analysis of

minimum wage effects concentrates on the increased risk of not being picked up for daily

employment for workers operating in a casual labor market, where daily employment is decided

on an ad hoc basis in the absence of any longer-term contract.

2. Background and Data

2.1 Minimum Wage for Piece Rate Workers

For a given pay period (here, one calendar week), consider a worker 𝑖 with a transaction profile of

(ℎ𝑖 , 𝑌𝑖 ), where ℎ𝑖 denotes the total field hours spent and 𝑌𝑖 the total output (in pieces) produced.

6
This is in contrast to approaches that are based on firm-level, revenue-based productivity such as Total Factor
Productivity (TFP). For instance, Mayneris, Poncet and Zhang (2018) document in the context of China that minimum
wage leads to the exit of low productivity firms and increases firm-level TFP conditional on survival.

5
This is the author’s accepted manuscript without copyediting, formatting, or final corrections. It will be published in its final form in an upcoming issue of
Journal of Labor Economics, published by The University of Chicago Press. Include the DOI when citing or quoting: https://doi.org/10.1086/716347
Copyright 2021 The University of Chicago Press.

Applying the constant piece rate (dollars/piece) 𝑝, the total output can be expressed as 𝑝𝑌𝑖 in

dollars. This worker’s average productivity then is 𝑝𝑌𝑖 /ℎ𝑖 ≡ 𝑝𝑦𝑖 (dollars per hour). Since the piece

rate 𝑝 remains constant throughout, there is a one-to-one correspondence between a worker’s

physical productivity 𝑦𝑖 (pieces per hour) and his productivity expressed in dollars 𝑝𝑦𝑖 .

For all hours employed during the pay period, workers whose average raw productivity is

above (below) the minimum wage are paid according to actual output (minimum wage). 7 Hence,

worker 𝑖’s hourly wage is

𝑝𝑦𝑖 if 𝑝𝑦𝑖 ≥ 𝑀𝑊
𝐻𝑜𝑢𝑟𝑙𝑦 𝑤𝑎𝑔𝑒𝑖 = {
𝑀𝑊 if 𝑝𝑦𝑖 < 𝑀𝑊

where 𝑀𝑊 denotes the minimum wage. Worker 𝑖’s total weekly earnings are

𝑝𝑌 if 𝑝𝑦𝑖 ≥ 𝑀𝑊
𝐸𝑎𝑟𝑛𝑖𝑛𝑔𝑠𝑖 = { 𝑖
ℎ𝑖 𝑀𝑊 if 𝑝𝑦𝑖 < 𝑀𝑊

so the firm’s total wage bill is

∑ 𝑝𝑌𝑗 + ∑ ℎ𝑗′ (𝑀𝑊 − 𝑝𝑦𝑗′ )

𝑗 𝑝𝑦𝑗′ <𝑀𝑊

where the first and second parts represent (i) the unadjusted wage bill for all workers and (ii) the

compliance cost for the minimum wage expended on subminimum wage workers, respectively.

When worker productivity is held constant, firm’s compliance costs increase with the

minimum wage for two reasons: first, a higher minimum wage makes the minimum wage bite for

more workers than previously, and second, it increases the gap between the (new) minimum wage

and the raw productivity of subminimum wage workers. A minimum wage increase thus creates

an incentive for firms to reduce (either at the extensive or intensive margins) the employment hours

7
This is similar to the piece rate scheme with a guarantee modelled in Lazear (2000). The difference is that in Lazear
(2000) the guarantee is chosen by the firm whereas here the minimum wage is imposed by the government.

6
This is the author’s accepted manuscript without copyediting, formatting, or final corrections. It will be published in its final form in an upcoming issue of
Journal of Labor Economics, published by The University of Chicago Press. Include the DOI when citing or quoting: https://doi.org/10.1086/716347
Copyright 2021 The University of Chicago Press.

assigned to low productivity workers. On the other hand, low productivity workers can preempt

the firm’s action by increasing their efforts and productivity, thereby (at least partially) relieving

the firm of the expanding compliance cost. 8 Whether either or both effects exist is examined

empirically below.

2.2 Setting and Data

The setting of our analysis is a large tomato farm in Florida where piece-rate workers hand-harvest

fresh tomatoes in the field. Our main data come from the personnel records of the farm covering

the 12-week autumn harvesting season from November 2008 to January 2009. Because this firm

uses one calendar week pay periods, the timeline in Figure 1 shows the harvesting periods by week.

During the 9th week of this season, in particular on January 1, 2009, the state minimum wage rose

from $6.79 to $7.21, an increase by 42 cents or 6% of the baseline minimum wage.

[Figure 1]

The minimum wage increase comes from Article X, Section 24 of the Florida Constitution.

Enacted in 2004 and first implemented in 2005, Florida’s minimum wage is indexed to inflation.

In particular, on September 30th of each year, an adjusted minimum wage rate is computed based

on the current minimum wage and the inflation rate (based on CPI-W) during the twelve months

prior to each September 1st, which is then published and takes effect on the following January 1st.9

8
Note that increased productivity of workers who are above the minimum wage makes little difference for the firm’s
labor cost since compensation is purely piece-rate based, which means that the total wage bill is determined by the
total output (pieces) harvested and the piece rate only and not by the speed or productivity (output/hour) of (above
minimum wage) workers.
9
Specifically, part (c) of Article X, Section 24 of the Florida Constitution reads “MINIMUM WAGE. Employers shall
pay Employees Wages no less than the Minimum Wage for all hours worked in Florida. Six months after enactment,
the Minimum Wage shall be established at an hourly rate of $6.15. On September 30th of that year and on each
following September 30th, the state Agency for Workforce Innovation shall calculate an adjusted Minimum Wage
rate by increasing the current Minimum Wage rate by the rate of inflation during the twelve months prior to each
September 1st using the consumer price index for urban wage earners and clerical workers, CPI-W, or a successor

7
This is the author’s accepted manuscript without copyediting, formatting, or final corrections. It will be published in its final form in an upcoming issue of
Journal of Labor Economics, published by The University of Chicago Press. Include the DOI when citing or quoting: https://doi.org/10.1086/716347
Copyright 2021 The University of Chicago Press.

As Table A.2 shows, the minimum wage hike on January 1, 2009 is relatively large in absolute

magnitude. This has to do with the high inflation rate that prevailed during the twelve months prior

to September 1, 2008, as shown in Appendix Figure A.1.

Although the farm operates several different fields and grows different tomato varieties,

due to a confidentiality agreement with the firm, this analysis is constrained to the harvesting of

two main varieties, round and grape tomatoes, which represent over 70 percent of total man hours.

All field workers are paid by piece rate based on individual output, meaning no team element in

production or compensation, and may be asked to pick either tomato variety depending on the

day’s harvesting requirements.

During the season workers stay in a living quarter located near the farm in rural Florida.

The available worker pool may change as new workers arrive and existing workers exit during the

season (see Section 4.3 for further details). There is no long-term contract and employment is

decided on a day-to-day basis. Specifically, the harvesting manager decides based on the field

capacity (i.e. how many mature crops are there to be harvested on that day, which is predetermined

by the acreage planted and the field lifecycle) and weather conditions, whether they will harvest

or not for a given day, and how many workers will be needed (based on some heuristic he has

figured out through many years of experience). As shown in Figure A.2, the number of workers

employed each day is closely related to the “field capacity”, with an R-squared of 0.7695.10 Once

the day’s harvesting plan is known, an appropriate number of workers are recruited from the

available worker pool. We do not know exactly how this is achieved in practice but in the data we

index as calculated by the United States Department of Labor. Each adjusted Minimum Wage rate calculated shall be
published and take effect on the following January 1st. For tipped Employees meeting eligibility requirements for the
tip credit under the Fair Labor Standards Act (FLSA), Employers may credit towards satisfaction of the Minimum
Wage tips up to the amount of the allowable FLSA tip credit in 2003.”
10
Since it is not possible to measure the (predetermined) field capacity, we use the actual output as a proxy.

8
This is the author’s accepted manuscript without copyediting, formatting, or final corrections. It will be published in its final form in an upcoming issue of
Journal of Labor Economics, published by The University of Chicago Press. Include the DOI when citing or quoting: https://doi.org/10.1086/716347
Copyright 2021 The University of Chicago Press.

clearly see that it is not the same number of workers or the same set of workers being employed

each day. In Section 4.3 we investigate whether workers’ employment outcomes are in any way

related to their productivity.

The harvesting workers, if working that day, arrive in the field by buses organized by crew

leaders, and once the day’s harvesting is finished, leave by the same buses. To track each worker’s

output and work hours electronically, an ID card with a magnetic chip is attached to each worker’s

bucket and scanned at the beginning and end of each work spell. Although a work day may

comprise multiple work periods, there is typically a morning and afternoon work spell with a lunch

break separating the two. During a work period, workers spread around the field to pick tomatoes

from different rows of thick tall bushes and then carry their filled buckets to a truck parked in the

middle of the field. Several “dumpers” standing on the back of the truck empty the full buckets

into a large collection bin and scan the worker’s ID card with a scanning device to add the output

unit to the system. This procedure is repeated throughout the day until the day’s designated fields

are completely picked.

Output is measured in 32 pound bucket, for which the piece rate for round (grape) tomatoes

is a constant $0.50 ($3.75) throughout. Therefore, for ease of comparison, workers’ physical output

is always converted to dollars (pieces times piece rate for the relevant variety), and productivity

(output per hour) is expressed in dollars per hour. For each variety separately, we remove the

transactions that fall in the bottom and top 1 percent of the productivity distribution to ensure that

the results are not driven by outliers. Further, we focus on workers with at least five spells

(transactions) during weeks 1-5 (so that we can obtain reliable estimates of their fixed effects).11

This results in 31,762 transactions for 974 unique workers. The average output per hour

11
Our main results are robust to alternative choices of minimum spell numbers in the vicinity.

9
This is the author’s accepted manuscript without copyediting, formatting, or final corrections. It will be published in its final form in an upcoming issue of
Journal of Labor Economics, published by The University of Chicago Press. Include the DOI when citing or quoting: https://doi.org/10.1086/716347
Copyright 2021 The University of Chicago Press.

(dollars/hour) in the sample is $9.53 with a standard deviation of $3.62. In Table A.3 we report

the mean daily employment (1 if working and 0 otherwise), daily hours worked (if working that

day), and productivity (output per hour) by quintiles of worker fixed effects (see Section 3) and by

time periods (weeks 1-5, 6-8 and 9-12).

To address the relevant question of how substantive this new $7.21 minimum wage is, the

incidence and extent of the old and the new minimum wages are tabulated in Table 1. As the

minimum wage rises from $6.79 to $7.21, the share of worker-weekly paychecks for which the

minimum wage binds rises from 12 to 16 percent. Moreover, the share of workers for whom the

minimum wage will ever bite increases from 42 to 52 percent. At the same time, the share of farm-

level employment hours assigned to worker-weeks below the minimum wage rises from 10 to 14

percent, and the minimum wage compliance cost increases from $8,340 to $13,217 (about 58%

increase).

[Table 1]

2.3 Compliance

The minimum wage is part of the Fair Labor Standards Act (FLSA), which also sets overtime,

recordkeeping, and child labor standards. Contrary to popular misconceptions, all agricultural

workers on any but small farms, while exempt from the law’s overtime pay provision, are covered

by its minimum wage requirement. 12 Since the state of Florida has its own minimum wage,

whichever one is higher binds between the federal and state minimum wages (see Table A.2).

12
An agricultural employer who does not use more than 500 man days (days on which a worker provides at least one
hour of agricultural work) in any calendar quarter of the preceding calendar year is exempt from the FLSA minimum
wage provision for the current calendar year. The farm studied here hires 300–600 workers per day and thus is not
exempt from the provision.

10
This is the author’s accepted manuscript without copyediting, formatting, or final corrections. It will be published in its final form in an upcoming issue of
Journal of Labor Economics, published by The University of Chicago Press. Include the DOI when citing or quoting: https://doi.org/10.1086/716347
Copyright 2021 The University of Chicago Press.

As with any empirical research on minimum wage, one important concern here is

noncompliance.13 The most common violation of minimum wage regulation is manipulating the

manual records of workers’ compensable hours. The recordkeeping standards at the farm studied

here, however, makes ex post manipulation of employment hours highly implausible. Workers are

clocked in and out in the field by magnetic chips. Nevertheless, we perform several tests to

eliminate this possibility, including an inspection of workers’ actual paystubs to verify that

subminimum wage workers were indeed paid the minimum wage. To illustrate, the worker whose

weekly paystub is shown in Figure A.3 worked a total of 15.28 hours over two days during the

reference week in 2008. Based on his output, his raw (unadjusted) earnings were $87.75 dollars

($29.00 + $7.50 + $11.25 + $40.00), which translates into an hourly productivity (dollars/hour) of

$5.74. Because the relevant minimum wage for this period was $6.79, the worker was paid $6.79

and not $5.74 for all 15.28 hours worked, resulting in a total earnings of $103.75 ($6.79 times

15.28). The firm’s compliance costs were thus $16 ($103.75 minus $87.75), which appears as a

line item labeled “minimum wage.”

We also check for any sign of ex post manipulation in the payroll data, in particular, any

downward adjustment of employment hours for workers whose raw hourly productivity falls

below the minimum wage. Figure A.4 plots the mean of worker-weekly total hours of employment

by 5 cents bins of worker-weekly average productivity (output per hour) for a 2 dollar window

around the relevant minimum wage. Data are pooled across weeks with week fixed effects

controlled for. The plot for weeks 1-8 (minimum wage = $6.79) is in part (a) whereas that for

weeks 9-12 (minimum wage = $7.21) is in part (b). As the figure illustrates, the individual work

13
See Ashenfelter and Smith (1979) and Clemens and Strain (2020) for discussion on employer noncompliance with
the minimum wage.

11
This is the author’s accepted manuscript without copyediting, formatting, or final corrections. It will be published in its final form in an upcoming issue of
Journal of Labor Economics, published by The University of Chicago Press. Include the DOI when citing or quoting: https://doi.org/10.1086/716347
Copyright 2021 The University of Chicago Press.

hours in any given (calendar week) pay period are smooth along the distribution of each worker’s

contemporaneous productivity. That is, there is no sign of a discontinuous drop in field hours for

workers below the productivity threshold of $6.79 (or $7.21), which could be expected if the firm

had adjusted subminimum wage workers’ field hours downward.14

On the other hand, if the firm were to make a uniform downward adjustment to everyone’s

employment hours, such adjustment could not be detected without having access to the

unadulterated records. Even if such uniform downward adjustment were to happen, it would not

threaten the analysis because the difference-in-differences strategy used examines possible

differential changes in the outcomes of low versus high productivity workers when both groups

are exposed to the same shocks at the firm level. Such shocks would include both the January 1

minimum wage hike and the (highly unlikely) uniform downward adjustment of everyone’s

employment hours.

3. Empirical Strategy

Outdoor production of agricultural crops tends to be characterized by natural fluctuations in

average productivity due to external factors such as weather conditions and the field lifecycle. It

is therefore tenuous to attribute to effort any changes in worker productivity observed before and

after a minimum wage hike. To isolate the effects of worker effort from external determinants of

labor productivity, we therefore employ a difference-in-differences (DID) strategy. Similar to Mas

and Moretti (2009), we first capture each worker’s “baseline” or “permanent” productivity by

14
Relatedly, Figure A.5 shows the McCrary plot, which tests for selective sorting around the threshold on the worker-
weekly average productivity (output per hour). Consistent with no ex post manipulation of production records by the
firm, the figure shows no discontinuity in the density of observations around the minimum wage either in the pre or
in the post period.

12
This is the author’s accepted manuscript without copyediting, formatting, or final corrections. It will be published in its final form in an upcoming issue of
Journal of Labor Economics, published by The University of Chicago Press. Include the DOI when citing or quoting: https://doi.org/10.1086/716347
Copyright 2021 The University of Chicago Press.

estimating their individual fixed effects using data from weeks 1-5 of the harvesting season. Based

on the estimated fixed effects, we classify workers into high versus low productivity types.

We then look for possibly differential productivity changes of individual workers from

weeks 6-8 to weeks 9-12 by their baseline productivity. Since low fixed effect workers are more

likely to fall below a minimum wage than high fixed effect workers when subject to a common

production environment, to the extent workers respond to the fear of selective non-employment

(or other related motives) we should expect a disproportionate increase in observed productivity

in the lower part of the fixed effects distribution as the minimum wage increases.

Based on data from weeks 1-5, we first estimate by OLS the following regression:

(1) 𝑦𝑖𝑣𝑓𝑡 = 𝜙𝑣𝑓 + 𝜓𝑡 + 𝛾1 𝒁𝑖𝑡 +𝛾2 𝑿𝑣𝑓𝑡 + 𝛼𝑖 + 𝑢𝑖𝑣𝑓𝑡 ,

where 𝑦𝑖𝑣𝑓𝑡 denotes (log) worker 𝑖’s output per hour for variety 𝑣 in field 𝑓 on day 𝑡. We include

variety-field fixed effects (𝜙𝑣𝑓 ) to capture any between-variety differences that are also field

specific, and day fixed effects (𝜓𝑡 ) to account for such day-specific common shocks as weather.

Worker fixed effects, which capture each worker’s baseline productivity, are denoted by 𝛼𝑖 . As a

result, the estimates of 𝛼𝑖 capture the differences between workers who harvest the same variety

in the same field while eliminating day-specific common shocks. Further, we also include variety-

field-day specific observed characteristics such as a cubic polynomial of the variety-field

lifecycle, 15 supervisor fixed effects (collected in 𝑿𝑣𝑓𝑡 ), and a cubic polynomial of worker

experience, measured as cumulative work hours from the beginning of the season to day 𝑡, 𝒁𝑖𝑡 .

Essentially, we want to capture in 𝛼𝑖 worker’s fixed characteristic, which we refer to as baseline

productivity, while accounting for other determinants of worker’s observed productivity.

15
The variety-field lifecycle is computed as the number of days the variety has been picked in that field by day 𝑡
divided by the total number of days it has been harvested in that field during the season.

13
This is the author’s accepted manuscript without copyediting, formatting, or final corrections. It will be published in its final form in an upcoming issue of
Journal of Labor Economics, published by The University of Chicago Press. Include the DOI when citing or quoting: https://doi.org/10.1086/716347
Copyright 2021 The University of Chicago Press.

Next, based on the estimated fixed effects 𝛼̂𝑖 , we classify workers into different bins (e.g.

percentiles or quintiles), and then estimate variants of the following regression (based on

transactions over weeks 6–12):

(2) 𝑦𝑖𝑣𝑓𝑡 = 𝛿(𝑃𝑜𝑠𝑡𝑡 × 𝐷𝑖 ) + 𝜋𝑖 + 𝜓𝑡 + 𝜙𝑣𝑓 + 𝛽1 𝒁𝑖𝑡 + 𝛽2 𝑿𝑣𝑓𝑡 + 𝑒𝑖𝑣𝑓𝑡 ,

where 𝑦𝑖𝑣𝑓𝑡 again denotes (log) worker 𝑖’s output per hour for variety 𝑣 in field 𝑓 on day 𝑡. The

variable 𝑃𝑜𝑠𝑡𝑡 assumes the value of unity if day 𝑡 belongs to weeks 9-12 and zero otherwise, while

𝐷𝑖 indicates whether worker 𝑖 is, say, in the bottom 40th percentile on the (pre-estimated) worker

fixed effects distribution.16 We include dummies indicating each percentile of the pre-estimated

worker effects 𝜋𝑖 (which subsumes 𝐷𝑖 ) and day fixed effects 𝜓𝑡 (which subsumes 𝑃𝑜𝑠𝑡𝑡 ). As in

equation (1), we also control for variety-field fixed effects 𝜙𝑣𝑓 , and for the variables in 𝒁𝑖𝑡 and

𝑿𝑣𝑓𝑡 . The DID estimate 𝛿 measures the disproportionate productivity changes of workers in the

bottom 40th percentile of the fixed effects distribution relative to those in the upper part. All

standard errors are clustered by day.17

We start by comparing the productivity changes in the bottom two quintiles with those in

the upper quintiles. Because the quintiles are based on predefined characteristics (i.e. worker fixed

effects in weeks 1-5), this method of classifying treatment status is exogenous to workers’ actual

effort choices during the analytic window (weeks 6-12). The identifying assumption for this

approach is that conditioning on the included controls, in particular, the harvesting day fixed

effects that capture farm-level common shocks specific to each day (e.g. weather), there are no

significant changes on or around January 1, 2009, other than the new minimum wage that might

16
We also consider a more flexible approach where we allow quintile-specific productivity changes.
17
Our main results are robust to clustering at the worker level.

14
This is the author’s accepted manuscript without copyediting, formatting, or final corrections. It will be published in its final form in an upcoming issue of
Journal of Labor Economics, published by The University of Chicago Press. Include the DOI when citing or quoting: https://doi.org/10.1086/716347
Copyright 2021 The University of Chicago Press.

differentially influence the effort choices of workers in the lower versus upper part of the worker

fixed effects distribution.

4. Results

4.1 Worker Fixed Effects

Based on a sample of 13,291 transactions (974 unique employees) from weeks 1-5, we first

estimate equation (1). Based on the estimated fixed effects (whose mean is zero by normalization

with a standard deviation of 0.2254), we classify workers into different bins (quintiles or

percentiles).18 We then examine the relationship between worker’s observed productivity in weeks

6-8 (the “pre” period with respect to the minimum wage hike) and their baseline productivity (i.e.

fixed effects).19

Below we present evidence that the pre-estimated fixed effects are indeed a good predictor

of worker’s observed productivity and hence the risk of falling below the minimum wage. In the

graphical analysis below, we use worker-week as the unit of observation—the unit at which

paychecks are issued and minimum wage adjustments are made—without using additional controls.

In our regression analysis (Section 4.2), we use a finer variation with an extensive list of controls

to account for external determinants of productivity.

Figure 2 plots the worker-week level productivity distribution for weeks 6–8 by quintiles

of the worker fixed effects based on weeks 1-5. Because of such external factors as weather and

field lifecycle, there is a fair amount of dispersion in worker productivity even for the same

18
The distribution of the estimated worker fixed effects is presented in Figure A.6.
19
The correlation in the observed productivity of each worker between periods is 0.6886 (between periods 1 and 2),
0.5544 (between periods 1 and 3) and 0.6597 (between periods 2 and 3), where periods 1, 2 and 3 refer to weeks 1-5,
6-8 and 9-12, respectively.

15
This is the author’s accepted manuscript without copyediting, formatting, or final corrections. It will be published in its final form in an upcoming issue of
Journal of Labor Economics, published by The University of Chicago Press. Include the DOI when citing or quoting: https://doi.org/10.1086/716347
Copyright 2021 The University of Chicago Press.

quintiles. However, on average, workers in the lower quintiles tend to have lower productivity,

suggesting that the estimated worker fixed effects (from weeks 1-5) are indeed informative. The

monotonic relationship between worker productivity (in weeks 6-8) and the pre-estimated worker

fixed effects is also visualized in Figure 3, which displays the mean of worker-weekly productivity

against the percentile of the worker fixed effects. Based on these figures, the pre-estimated worker

fixed effects (based on weeks 1-5) seem to be a good proxy for workers’ baseline productivity.20

[Figures 2 and 3]

Given the monotonicity in Figure 3, it is easy to imagine that workers in the lower part of

the fixed effects distribution are more likely to fall below the new (and old) minimum wage than

those in the upper part. Figure 4 illustrates this. Based on worker-weekly productivity observations

(from weeks 6-8), we compute for each percentile of the worker fixed effects the share of

observations that fall below the current and new minimum wages. As shown, the probability to

fall below the minimum wage is greater at the lower part of the distribution than at the higher part.

Moreover, the probability shifts upwardly as we apply the new minimum wage to the same

productivity data, and the upward shift is more pronounced on the lower part of the fixed effects

distribution than in the upper part. Therefore, low fixed effect workers have a greater incentive to

increase effort than high fixed effect workers when both are subject to the minimum wage hike on

January 1, 2009.

[Figure 4]

20
This stability can also be established using fixed effects based on weeks 1-3 and observed productivity in weeks 4-
5 albeit for a smaller sample than used here.

16
This is the author’s accepted manuscript without copyediting, formatting, or final corrections. It will be published in its final form in an upcoming issue of
Journal of Labor Economics, published by The University of Chicago Press. Include the DOI when citing or quoting: https://doi.org/10.1086/716347
Copyright 2021 The University of Chicago Press.

4.2 Minimum Wage Effect on Worker Effort

4.2.1 Main results

Panel A of Table 2 presents the estimates of equation (2) (or its variants), which contrasts the effort

responses of workers in the lower part of the distribution with those of workers further up the

distribution. Column (1) compares the productivity change in quintiles 1-2 of the worker fixed

effects distribution with that of the upper quintiles (quintiles 3-5). The estimates reveal that there

is a positive productivity change in the bottom two quintiles relative to the upper quintiles. Here,

the coefficients on 𝑃𝑜𝑠𝑡𝑡 × 𝑄𝑢𝑖𝑛𝑡𝑖𝑙𝑒 1𝑖 and 𝑃𝑜𝑠𝑡𝑡 × 𝑄𝑢𝑖𝑛𝑡𝑖𝑙𝑒 2𝑖 are not significantly different

from each other. In column (2), we then estimate the average changes among quintiles 1-2 relative

to quintiles 3-5. The estimates show that the output per hour for workers below the 40th percentile

increases a disproportionate 4.6 percent relative to workers in the comparison group (above 40th

percentile).21

[Table 2]

To address the concern that what constitutes the pre and post period may not be so clear-

cut, we exclude the transition weeks (weeks 8-9). While January 1 falls in week 9, the new and

higher minimum wage may take time to sink in with the workers rather than having an immediate

effect. On the contrary, workers may respond proactively in the week before the scheduled increase

in the minimum wage. When we focus on these pre and post periods that are more clearly apart,

the coefficient becomes larger (see columns (3) and (4)). In columns (5) and (6), we exclude the

final week from the sample as the incentives may be weaker when workers know that the season

21
In Table A.4 we present estimates using alternative ways to account for workers’ baseline productivity. Columns
(1)-(2) replicate our baseline estimates from columns (1)-(2) of Table 2. In columns (3)-(4), we include dummies for
each quintile of the pre-estimated worker fixed effects, requiring coarser information on worker types than in the
baseline. In columns (5)-(6), we include the pre-estimated worker fixed effects in levels, imposing a linear effect and
reducing the number of parameters to be estimated. As shown, the results are robust to these alternative (and less
demanding) specifications to account for the baseline productivity.

17
This is the author’s accepted manuscript without copyediting, formatting, or final corrections. It will be published in its final form in an upcoming issue of
Journal of Labor Economics, published by The University of Chicago Press. Include the DOI when citing or quoting: https://doi.org/10.1086/716347
Copyright 2021 The University of Chicago Press.

will end after that. The magnitude of the main estimates changes only slightly. Overall, these

findings suggest that the increase of the minimum wage from $6.79 to $7.21 increases the relative

productivity of workers in the bottom 40th percentile by 4-7%.

So far we focused on the dichotomous distinction between low vs. high productivity

workers. We now take a more flexible approach and estimate a variant of equation (2) where we

replace 𝑃𝑜𝑠𝑡𝑡 × 𝐷𝑖 with 𝑃𝑜𝑠𝑡𝑡 interacted with dummies for each quintile, using quintile 3 as the

reference category. The estimated coefficients along with 95th percent confidence intervals are

plotted in Figure 5. As expected, the productivity response of quintiles 1 and 2 are clearly positive

relative to that of quintile 3 (the reference category). Moreover, there are also modest increases in

the productivity of quintiles 4 and 5 albeit smaller in magnitude than that for quintiles 1 and 2 and

imprecisely estimated.22 Such responses on the upper part of the distribution may be driven by

peer pressure as in Mas and Moretti (2009) and/or high-productivity workers’ preferences for

maintaining their rank ordering as in Kuziemko et al. (2014). The pattern shown in Figure 5 also

ensures that the estimates in Table 2 are indeed driven by positive effort responses of low

productivity workers (who are more “at risk” than others) rather than potentially negative effort

responses of high productivity workers that may be activated by fairness concerns as in Breza,

Kaur and Shamdasani (2017) and Dube, Giuliano and Leonard (2019).

[Figure 5]

22
The difference between the estimated coefficients plotted in Figure 5 and the p-value associated with each test are
̂
as follows: 𝑃𝑜𝑠𝑡 × 𝑄2 − 𝑃𝑜𝑠𝑡̂ × 𝑄4 = 0.0503 (p-value = 0.004); 𝑃𝑜𝑠𝑡 ̂ × 𝑄2 − 𝑃𝑜𝑠𝑡 ̂ × 𝑄5 = 0.0392 (p-value =
̂
0.015); and 𝑃𝑜𝑠𝑡 × 𝑄2 − 𝑃𝑜𝑠𝑡̂ × 𝑄1 = 0.0068 (p-value = 0.684).

18
This is the author’s accepted manuscript without copyediting, formatting, or final corrections. It will be published in its final form in an upcoming issue of
Journal of Labor Economics, published by The University of Chicago Press. Include the DOI when citing or quoting: https://doi.org/10.1086/716347
Copyright 2021 The University of Chicago Press.

4.2.2 Robustness Checks

Our DID estimates show a disproportionate increase in observed productivity in the lower part of

the fixed effects distribution than in the upper part as the minimum wage increases from the pre to

the post periods. There are potential threats to interpreting these effects as “effort response” to the

minimum wage including composition changes and reliability of the estimated worker effects.

Below we report a series of additional robustness checks to address these concerns.

First, we investigate whether the baseline effects detected might result from changes in

worker composition (i.e. differential selection within high vs. low quintile bins over time). We

repeat our main analysis using a balanced sample of workers who worked in both the pre (weeks

6–8) and post (weeks 9–12) periods. In these estimates, reported in Panel B of Table 2, the sample

size becomes slightly smaller, but the patterns are similar to those in Panel A of the same table.23

Thus, the increase in output per worker detected is unlikely to be driven by differential

compositional changes within high vs. low quintile bins.

Next, we address concerns regarding the reliability of the estimated worker effects. In

estimating worker fixed effects in equation (1), we imposed the restriction of at least 5 spells per

worker, resulting in on average 13 spells per worker, which should be considered “large” by the

standard of panel data models (see Fernández-Val and Weidner (2018)). To check nevertheless

whether smaller spell numbers for some workers may be an issue, we vary the minimum number

of spells we require when estimating equation (1). In Table 3, column (3) replicates the result in

column (2) of Table 2. Columns (1) and (2) show results with lower numbers of minimum spells

23
Note that in either sample, the main coefficient of interest, Post*Low FE, is identified from workers who worked in
both the pre and post periods. However, in panel A, the part-season workers still contribute to the estimation of other
included controls such as day FE or variety-field FE.

19
This is the author’s accepted manuscript without copyediting, formatting, or final corrections. It will be published in its final form in an upcoming issue of
Journal of Labor Economics, published by The University of Chicago Press. Include the DOI when citing or quoting: https://doi.org/10.1086/716347
Copyright 2021 The University of Chicago Press.

and columns (4)-(6) with higher numbers. As we move from column (1) through column (6), the

DID estimate of productivity changes remains largely stable.

[Table 3]

As an additional check for the reliability of our estimated worker effects, we also conduct

a simulation exercise. In particular, we randomly draw individual effects from a distribution with

mean of zero and standard deviation of 0.2254 (see Figure A.6) and estimate the 𝛿 in equation (2)

with these randomly assigned individual effects. We repeat this exercise 500 times and plot the

distribution of the estimated 𝛿’s in Figure A.7. To assess how likely it is that the baseline estimate

of 0.046 (column (2) of Table 2) happens “by chance” even when the underlying worker effects

are completely random, the short dashed line indicates the 95th percentile of the distribution of 𝛿̂ ’s

based on randomly assigned individual effects. Our baseline estimate of 𝛿̂ =0.046 lies clearly

outside this threshold.

4.3 Employment Outcomes

Our analysis so far focused on the effort responses of workers. In this section, we examine (i)

whether the employment outcomes differ between high and low productivity workers (irrespective

of the minimum wage) and (ii) whether any pre-existing differences may be further amplified

because of the minimum wage increase.

Before we proceed, it is worthwhile understanding the evolution of farm-level employment

during the season. As Panel (a) of Figure 6 shows, the farm’s lifecycle and hence its labor demand

cycle peaks in the middle of the season. In keeping with these employment requirements, the farm

needs to build up its worker pool at the beginning of the season and shed it once the peak has

reached and as the season is winding down. In Panel (b) of Figure 6, we plot the cumulative inflows,

20
This is the author’s accepted manuscript without copyediting, formatting, or final corrections. It will be published in its final form in an upcoming issue of
Journal of Labor Economics, published by The University of Chicago Press. Include the DOI when citing or quoting: https://doi.org/10.1086/716347
Copyright 2021 The University of Chicago Press.

cumulative outflows, and the stock of workers in each week. The inflows in week K are measured

as workers whose first harvesting day occurs in week K, and the outflows in week K are defined

as those whose last harvesting day occurs in week K-1. The stock of workers in week K are the

cumulative inflows by week K minus the cumulative outflows by week K, proxying for the size of

the worker pool available for hire in week K. As the figure shows, we see large inflows at the

beginning of the season, the pace of which then slows down as the season progresses. In contrast,

worker outflows are negligible at the beginning of the season, but then steadily increases over time.

[Figure 6]

In order to understand whether worker productivity matters at all in the day-to-day

allocation of employment opportunities, we estimate the following equation:

(3) 𝑒𝑚𝑝𝑙𝑜𝑦𝑚𝑒𝑛𝑡𝑖𝑡 = 𝜂1 𝐷𝑖 + 𝜂2 (𝑃𝑜𝑠𝑡𝑡 × 𝐷𝑖 ) + 𝜓𝑡 + 𝜔𝑖𝑡 ,

where 𝑒𝑚𝑝𝑙𝑜𝑦𝑚𝑒𝑛𝑡𝑖𝑡 is worker 𝑖’s employment outcome on day 𝑡. The variable 𝐷𝑖 indicates

whether worker 𝑖 is below the 40th percentile in the baseline productivity distribution. The variable

𝑃𝑜𝑠𝑡𝑡 assumes the value of unity if day 𝑡 belongs to weeks 9-12 and zero otherwise. Day fixed

effects are absorbed in 𝜓𝑡 (which subsumes 𝑃𝑜𝑠𝑡𝑡 ). As before, the treatment status 𝐷𝑖 is based on

the worker’s pre-determined characteristic (from weeks 1-5) and is orthogonal to his

contemporaneous decisions. Standard errors are clustered by day.

In the absence of the second term, the equation estimates the simple difference between

low versus high productivity workers, i.e. whether low productivity workers are overall employed

less frequently than high productivity workers. Once we include the interaction term (𝑃𝑜𝑠𝑡𝑡 × 𝐷𝑖 ),

the coefficient 𝜂2 picks up the excess selectivity in the post period (weeks 9-12) over and above

that (𝜂1 ) in the pre period (weeks 6-8), which we may attribute to the minimum wage hike.

21
This is the author’s accepted manuscript without copyediting, formatting, or final corrections. It will be published in its final form in an upcoming issue of
Journal of Labor Economics, published by The University of Chicago Press. Include the DOI when citing or quoting: https://doi.org/10.1086/716347
Copyright 2021 The University of Chicago Press.

Table 4 displays the estimates of equation (3) (or its variants) using worker-day as the unit

of analysis. Columns (1) through (4) use daily employment (1 if working that day and 0 otherwise)

as the dependent variable. The estimate in column (1) shows that overall, workers below the 40th

percentile have a 0.025 percentage point lower probability of being employed each day than those

above the 40th percentile, which is about 6% (0.025/0.397) of the mean. Column (2) shows that the

pre-existing difference (-0.019) is about twice as large as the additional effect in the post period (-

0.011).

[Table 4]

So far, we have included all worker-days in the sample including workers who may have

already left the farm and are no longer available for hire for a given day. For instance, a worker

whose last day of employment during the season falls in week K-4 is unlikely to be available for

hire in week K. By including such workers and recording them as not working, we may overstate

the true extent of non-working status at this farm. In columns (3) and (4), we therefore exclude

worker-days from the sample if the worker’s last day of employment at this farm during this season

occurs in any week prior to the week of the present worker-day. The mean employment probability

in this restricted sample is 0.642 as opposed to 0.397 in the full sample. The estimates in columns

(3) and (4) show that while low productivity workers are 8.4% (0.054/0.642) less likely to be

employed than high productivity workers in general, there is no strong evidence that this pre-

existing employment gap widens when a higher minimum wage takes effect. Based on columns

(2) and (4), a possible reduction in the employment of low productivity workers attributable to the

minimum wage hike is at most 4.6% (0.030/0.642).

22
This is the author’s accepted manuscript without copyediting, formatting, or final corrections. It will be published in its final form in an upcoming issue of
Journal of Labor Economics, published by The University of Chicago Press. Include the DOI when citing or quoting: https://doi.org/10.1086/716347
Copyright 2021 The University of Chicago Press.

Further, we examine in columns (5) and (6) the intensive margin effect (i.e. total hours

worked conditional on working that day) on employment.24 While low productivity workers work

5.7% (0.323/5.596) fewer hours than high productivity workers, we do not find any evidence of a

further reduction in the hours worked by low productivity workers after the minimum wage hike.

This is consistent with the fact that harvesting workers, once in the field, tend to work the same

hours until the day’s harvesting finishes (and leave by the same buses on which they arrived in the

morning), leaving little scope for intensive margin adjustments.

Based on Table 4, it appears that while low productivity workers in general have lower

chances of being employed than high productivity workers, this existing employment gap is not

further widened because of the higher minimum wage. This lack of significant employment effects

attributable to the minimum wage hike may be reconciled based on various grounds. In a

competitive framework, the positive effort responses of subminimum wage workers should obviate

the need for reducing employment opportunities assigned to these workers. The fact that not all

workers whose raw productivity falls below the minimum wage are discharged is also in line with

earlier findings of Holzer, Katz and Krueger (1991) that some positive rents (for workers) are

associated with minimum wage jobs. Moving beyond competition, it is also possible that workers

simply reacted to a perceived threat—even in the absence of any actual pressure coming from the

employer—and/or are driven by fairness concerns between the worker and the employer in the

spirit of gift exchange (Akerlof 1982).

24
Columns (5) and (6) include worker-days with non-zero employment hours only, hence the smaller number of
observations than in other columns.

23
This is the author’s accepted manuscript without copyediting, formatting, or final corrections. It will be published in its final form in an upcoming issue of
Journal of Labor Economics, published by The University of Chicago Press. Include the DOI when citing or quoting: https://doi.org/10.1086/716347
Copyright 2021 The University of Chicago Press.

4.4 Discussion

The results of our analysis above indicate that in response to the January 1, 2009 minimum wage

hike, the productivity of workers in the bottom 40th percentile of the productivity distribution

increased by a disproportionate 4.6% relative to those in the higher percentiles (column (2), Table

2). In the absence of such worker responses, a higher minimum wage means a higher labor cost

for the firm because of higher associated compliance costs. If, however, some subminimum wage

workers increase their efforts, it may (at least partially) offset these rising costs.25 We examine

these alternatives in Table 5. Based on data for weeks 1–8, when the prevailing minimum wage is

$6.79, we compute a compliance cost by subminimum wage workers of $8,340.26

[Table 5]

We then consider the consequence of a minimum wage increase from $6.79 to $7.21. In

the absence of worker effort responses this will raise the firm’s compliance cost by $4,877 (from

$8,340 to $13,217). However, as earlier analyses show, low productivity workers may increase

their efforts, which would bring the firm’s compliance cost down to $10,660, which is $2,557 (or

52 percent) less than the projected increase of $4,878. On the other hand, any additional saving

implied by possible employment adjustment is rather minor (at most $374 or 8%) even if we allow

for a significant reduction in the employment hours assigned to low productivity workers by 4.6%

(0.030/0.642 based on column (4), Table 4). Overall, this calculation indicates that increased

worker productivity can offset about half of the projected rise in the minimum wage compliance

25
The positive productivity response to the minimum wage may suggest the new minimum wage functioning as an
efficiency wage (Shapiro and Stiglitz 1984; Rebitzer and Taylor 1995). However, efficiency wage is a concept
applicable to settings where workers are paid a fixed wage (salary) while their effort or output is difficult to monitor.
In our context, workers are compensated by pure piece rate and their output is readily observable by the employer,
which obviates the firm’s need to employ efficiency wage. See Shapiro and Stiglitz (1984) and Katz (1986) for further
discussion on the characteristics of workplaces conducive to adoption of efficiency wage.
26
For this counterfactual exercise, we fix the production schedule at the period of pre-minimum wage hike so as not
to confound the minimum wage effect with a seasonality effect.

24
This is the author’s accepted manuscript without copyediting, formatting, or final corrections. It will be published in its final form in an upcoming issue of
Journal of Labor Economics, published by The University of Chicago Press. Include the DOI when citing or quoting: https://doi.org/10.1086/716347
Copyright 2021 The University of Chicago Press.

cost, suggesting a roughly equal sharing of the projected increase between the employer and the

workers.

5. Conclusions

By employing a direct and high frequency measure of individual-level productivity on a

homogenous task in the context of Florida’s minimum wage hike on January 1, 2009 (from $6.79

to $7.21), we examine worker effort responses as a possible margin of adjustment to a minimum

wage hike. When the statutory minimum wage increases, workers in the lower part of the

productivity distribution face increased risk of falling below the minimum wage relative to those

on the upper part of the distribution. Low productivity workers may therefore increase their efforts

(and hence productivity) to preempt possible discharge. We find that in response to the 42 cents

or 6% increase in the minimum wage, worker productivity (i.e. output per hour) in the bottom 40th

percentile of the worker fixed effects distribution increases by about 4.6% relative to that in the

higher percentiles, suggesting that productivity increases driven by worker effort may help

mitigate the higher labor costs associated with the minimum wage.

Several cautions are warranted. First, this margin of adjustment can only work within

relatively low ranges of the minimum wage. If the minimum wage continues to rise to a higher

level, workers may no longer be able to keep up their effort (and productivity) with the minimum

wage due to physical or cognitive limits. At that point, the firm may adopt an entirely different

personnel policy (or different production technology) than observed here and worker effort

responses may no longer be a valid channel to absorb the rising labor cost associated with the

minimum wage.

25
This is the author’s accepted manuscript without copyediting, formatting, or final corrections. It will be published in its final form in an upcoming issue of
Journal of Labor Economics, published by The University of Chicago Press. Include the DOI when citing or quoting: https://doi.org/10.1086/716347
Copyright 2021 The University of Chicago Press.

Second, we focus here on workers who do not have a long-term contract and are hired on

a day-to-day basis within a harvesting season. If workers had extended-term contracts, the

incentive structure in place may look quite different. On the one hand, the fact that the job is more

or less guaranteed—at least for the fixed term—may reduce the incentive to increase effort in

response to the minimum wage. On the other hand, the job is worth more (in present discounted

values) than a daily laboring, hence the workers may find a greater incentive to increase effort to

keep it.

Although a plausible channel of adjustment to the minimum wage, incumbent workers’

effort responses have been largely overlooked in the literature, probably because in most settings,

measuring individual-level productivity around a minimum wage hike—without convolution with

task and workforce composition—is difficult. This work, although focused on a particular firm in

a particular industry, opens a new avenue for future research, particularly in terms of whether labor

productivity serves as a mechanism for adjusting to the minimum wage in other firms or industries.

26
This is the author’s accepted manuscript without copyediting, formatting, or final corrections. It will be published in its final form in an upcoming issue of
Journal of Labor Economics, published by The University of Chicago Press. Include the DOI when citing or quoting: https://doi.org/10.1086/716347
Copyright 2021 The University of Chicago Press.

References
Aaronson, Daniel, 2001. “Price Pass-Through and the Minimum Wage,” Review of Economics and
Statistics, 83(1): 158–69.

Aaronson, Daniel, Eric French, and James MacDonald, 2008. “The Minimum Wage, Restaurant
Prices, and Labor Market Structure,” Journal of Human Resources, 43(3): 688-720.

Akerlof, George A., 1982. “Labor Contracts as Partial Gift Exchange,” Quarterly Journal of
Economics, 97(4): 543-569.

Ashenfelter, Orley, and Robert S. Smith, 1979. “Compliance with the Minimum Wage Law,”
Journal of Political Economy, 87(2): 333–350.

Bell, Brian, and Stephen Machin, 2018. “Minimum Wages and Firm Value,” Journal of Labor
Economics, 36(1): 159-195.

Breza, Emily, Supreet Kaur, and Yogita Shamdasani, 2017. “The Morale Effects of Pay Inequality,”
Quarterly Journal of Economics, 133(2): 611-663.

Brown, Charles, 1999. “Minimum Wages, Employment, and the Distribution of Income,”
Handbook of Labor Economics, 3: 2101–2163.

Card, David, and Alan B. Krueger, 1994. “Minimum Wages and Employment: A Case Study of
the Fast Food Industry in New Jersey and Pennsylvania,” American Economic Review,
84(4): 772-793.

Card, David, and Alan B. Krueger, 1995. Myth and Measurement: The New Economics of the
Minimum Wage, Princeton, NJ: Princeton University Press.

Cengiz, Doruk, Arindrajit Dube, Attila Lindner, and Ben Zipperer, 2019. “The Effect of Minimum
Wages on Low-wage Jobs,” Quarterly Journal of Economics, 134(3): 1405-1454.

Clemens, Jeffrey, 2021. “How Do Firms Respond to Minimum Wage Increases? Understanding
the Relevance of Non-employment Margins,” Journal of Economic Perspectives, 35(1):
51-72.

Clemens, Jeffrey, Lisa B. Kahn, and Jonathan Meer, 2021. Dropouts Need Not Apply? The
Minimum Wage and Skill Upgrading. Journal of Labor Economics, 39(S1): S107-S149.

Clemens, Jeffrey, and Michael R. Strain, 2020. “Understanding “Wage Theft”: Evasion and
Avoidance Responses to Minimum Wage Increases,” No. w26969, National Bureau of
Economic Research.

27
This is the author’s accepted manuscript without copyediting, formatting, or final corrections. It will be published in its final form in an upcoming issue of
Journal of Labor Economics, published by The University of Chicago Press. Include the DOI when citing or quoting: https://doi.org/10.1086/716347
Copyright 2021 The University of Chicago Press.

Coviello, Decio, Erika Deserranno, and Nicola Persico, 2019. “Minimum Wage and Individual
Worker Productivity: Evidence from a Large US Retailer,” working paper.

Draca, Mirko, Stephen Machin, and John Van Reenen, 2011. “Minimum Wages and Firm
Profitability,” American Economic Journal: Applied Economics, 3(1): 129–151.

Dube, Arindrajit, Laura Giuliano, and Jonathan Leonard, 2019. “Fairness and Frictions: The
Impact of Unequal Raises on Quit Behavior,” American Economic Review, 109(2): 620-
63.

Dube, Arindrajit, T. William Lester, and Michael Reich, 2016. “Minimum Wage Shocks,
Employment Flows, and Labor Market Frictions,” Journal of Labor Economics, 34(3):
663-704.

Dustmann, Christian, Attila Lindner, Uta Schönberg, Matthias Umkehrer, and Philipp vom Berge,
2020. “Reallocation Effects of the Minimum Wage: Evidence From Germany,” Quarterly
Journal of Economics, forthcoming.

Fernández-Val, Iván and Martin Weidner, 2018. “Fixed Effect Estimation of Large T Panel Data
Models,” Annual Review of Economics, 10(1): 109-138.

Gittings, R. Kaj, and Ian M. Schmutte, 2016. “Getting Handcuffs on an Octopus: Minimum Wages,
Employment, and Turnover,” ILR Review, 69(5): 1133-1170.

Giuliano, Laura, 2013. “Minimum Wage Effects on Employment, Substitution, and the Teenage
Labor Supply: Evidence from Personnel Data,” Journal of Labor Economics, 31(1): 155–
194.

Harasztosi, Péter, and Attila Lindner, 2019. “Who Pays for the Minimum Wage?,” American
Economic Review, 109(8): 2693-2727.

Holzer, Harry J., Lawrence F. Katz, and Alan B. Krueger, 1991. “Job Queues and Wages,”
Quarterly Journal of Economics, 739-768.

Jardim, Ekaterina, Mark C. Long, Robert Plotnick, Emma Van Inwegen, Jacob Vigdor, and Hilary
Wething, 2018. “Minimum Wage Increases and Individual Employment Trajectories,” No.
w25182, National Bureau of Economic Research.

Katz, Lawrence F., 1986. “Efficiency Wage Theories: A Partial Evaluation,” NBER
Macroeconomics Annual, 1: 235-276, MIT Press.

Kuziemko, Ilyana, Ryan W. Buell, Taly Reich, and Michael I. Norton, 2014. ““Last-place
Aversion”: Evidence and Redistributive Implications,” Quarterly Journal of Economics,
129(1): 105-149.

Lazear, Edward P., 1986. “Salaries and Piece Rates,” Journal of Business, 405–431.
28
This is the author’s accepted manuscript without copyediting, formatting, or final corrections. It will be published in its final form in an upcoming issue of
Journal of Labor Economics, published by The University of Chicago Press. Include the DOI when citing or quoting: https://doi.org/10.1086/716347
Copyright 2021 The University of Chicago Press.

Lazear, Edward P., 2000. “Performance Pay and Productivity,” American Economic Review, 90(5):
1346-1361.

Lazear, Edward P., Kathryn L. Shaw, and Christopher Stanton, 2016. “Making Do with Less:
Working Harder During Recessions,” Journal of Labor Economics, 34(S1): S333–S360.

Leung, Justin H., 2018. Minimum Wage and Real Wage Inequality: Evidence from Pass-through
to Retail Prices. Review of Economics and Statistics, 1-47.

Machin, Stephen, Alan Manning, and Lupin Rahman, 2003. “Where the Minimum Wage Bites
Hard: Introduction of Minimum Wages to a Low Wage Sector,” Journal of the European
Economic Association, 1(1): 154-180.

Mas, Alexandre, and Enrico Moretti, 2009. “Peers at Work,” American Economic Review, 99 (1):
112-45.

Mayneris, Florian, Sandra Poncet, and Tao Zhang, 2018. “Improving or Disappearing: Firm-level
Adjustments to Minimum Wages in China,” Journal of Development Economics, 135: 20-
42.

Neumark, David, J. M. Ian Salas, and William Wascher, 2014. “Revisiting the Minimum Wage-
Employment Debate: Throwing Out the Baby with the Bathwater?,” ILR Review,
67(3_suppl): 608-648.

Obenauer, Marie Louise and Bertha Marie von der Nienburg, 1915. “Effect of Minimum-wage
Determinations in Oregon,” Bulletin of the US Bureau of Labor Statistics, July 1915 (176).

Portugal, Pedro, and Ana Rute Cardoso, 2006. “Disentangling the Minimum Wage Puzzle: An
Analysis of Worker Accessions and Separations,” Journal of the European Economic
Association, 4(5): 988–1013.

Rebitzer, James B., 1987. “Unemployment, Long-term Employment Relations, and Productivity
Growth,” Review of Economics and Statistics, 627-635.

Rebitzer, James B., and Lowell J. Taylor, 1995. “The Consequences of Minimum Wage Laws
Some New Theoretical Ideas,” Journal of Public Economics, 56 (2): 245-255.

Shapiro, Carl, and Joseph E. Stiglitz, 1984. “Equilibrium Unemployment as a Worker Discipline
Device,” American Economic Review, 74(3): 433–444.

Stigler, George J., 1946. “The Economics of Minimum Wage Legislation,” American Economic
Review, 36(3): 358–365.

29
Figure Click here to access/download;Figure;Figures.docx
This is the author’s accepted manuscript without copyediting, formatting, or final corrections. It will be published in its final form in an upcoming issue of
Journal of Labor Economics, published by The University of Chicago Press. Include the DOI when citing or quoting: https://doi.org/10.1086/716347
Copyright 2021 The University of Chicago Press.

Figure 1: Timeline of the 2009 minimum wage hike

Minimum
Pre (6.79) ← → Post (7.21)
Wage:

Week: 1 2 3 4 5 6 7 8 9 10 11 12
estimate worker FE Jan 1

The harvesting season being investigated spans 12 weeks from November 2008 to January 2009,
during which Florida's state minimum wage rose from $6.79 to $7.21 on January 1, 2009, a date
that falls in week 9 of the analytic window. The pre period is defined as weeks 1-8; the post
period as weeks 9-12. The estimation of worker fixed effects is based on transactions during the
initial 5 weeks; that of worker effort responses on those during weeks 6-12.
This is the author’s accepted manuscript without copyediting, formatting, or final corrections. It will be published in its final form in an upcoming issue of
Journal of Labor Economics, published by The University of Chicago Press. Include the DOI when citing or quoting: https://doi.org/10.1086/716347
Copyright 2021 The University of Chicago Press.

Figure 2: Distribution of worker productivity by quintiles of worker fixed effects

Worker-weekly output per hour during weeks 6-8. The worker fixed effects as pre-estimated by
equation (1) using transactions during weeks 1-5. The dashed vertical line indicates the old and new
minimum wages of $6.79 and $7.21, respectively.
This is the author’s accepted manuscript without copyediting, formatting, or final corrections. It will be published in its final form in an upcoming issue of
Journal of Labor Economics, published by The University of Chicago Press. Include the DOI when citing or quoting: https://doi.org/10.1086/716347
Copyright 2021 The University of Chicago Press.

Figure 3: Average productivity by percentiles of worker fixed effects

This figure plots the mean of worker-weekly productivity during weeks 6-8 by percentiles of worker
fixed effects. The worker fixed effects are pre-estimated by equation (1) using transactions during
weeks 1-5. The coefficient (SE) of the fitted line is 0.0473 (0.0021) and the R-squared is 0.8302.
This is the author’s accepted manuscript without copyediting, formatting, or final corrections. It will be published in its final form in an upcoming issue of
Journal of Labor Economics, published by The University of Chicago Press. Include the DOI when citing or quoting: https://doi.org/10.1086/716347
Copyright 2021 The University of Chicago Press.

Figure 4: Propensity to fall below the minimum wage by worker fixed effects

This figure plots for each percentile of worker fixed effects, the share of worker-weekly productivity
observations during weeks 6-8 that fall below the current ($6.79) and new ($7.21) minimum wages,
respectively. The worker fixed effects are pre-estimated by equation (1) using transactions during
weeks 1-5.
This is the author’s accepted manuscript without copyediting, formatting, or final corrections. It will be published in its final form in an upcoming issue of
Journal of Labor Economics, published by The University of Chicago Press. Include the DOI when citing or quoting: https://doi.org/10.1086/716347
Copyright 2021 The University of Chicago Press.

Figure 5: Productivity change from pre- to post-MW hike by quintiles of worker fixed effects

Based on transactions during weeks 6-12. Post=1 if week 9 or later. This figure plots the changes in
productivity (output per hour) from Pre to Post periods by quintiles of worker fixed effects (relative to
quintile 3). It plots the DID coefficients from estimating a variant of equation (2), where we replace
Post x D with Post x dummy for each quintile except for quintile 3. Regression includes percentile
dummies for pre-estimated worker effects, day FE, a cubic polynomial of worker experience, variety-
field FE, a cubic polynomial of variety-field life-cycle, and supervisor FE. Standard errors are
clustered by day.
This is the author’s accepted manuscript without copyediting, formatting, or final corrections. It will be published in its final form in an upcoming issue of
Journal of Labor Economics, published by The University of Chicago Press. Include the DOI when citing or quoting: https://doi.org/10.1086/716347
Copyright 2021 The University of Chicago Press.

Figure 6: Employment and worker flows during the season

(a) Employment

(b) Cumulative inflows, cumulative outflows and stock of workers

Panel (a) shows the total number of workers employed in each week. Panel (b) shows the cumulative
inflows, cumulative outflows, and stock of workers in each week. The inflows in week K are
measured as workers whose first harvesting day occurs in week K, and the outflows in week K are
defined as those whose last harvesting day occurs in week K-1. The stock of workers in week K are
the cumulative inflows by week K minus the cumulative outflows by week K, proxying for the size of
worker pool available for hire in week K.
Tables
This is the author’s accepted manuscript without copyediting, formatting, or final corrections. It will be published in its final form in an upcoming issue of
Journal of Labor Economics, published by The University of Chicago Press. Include the DOI when citing or quoting: https://doi.org/10.1086/716347
Copyright 2021 The University of Chicago Press.

Table 1: Incidence and extent of the new minimum wage

MW=$6.79 MW=$7.21
Minimum wage bites for the following share of:
worker-weekly paychecks 0.118 0.158
workforce (for whom MW ever bites) 0.422 0.521
employment hours 0.096 0.135
Minimum wage compliance cost:
in dollars 8,340 13,217
Notes: Both columns are based on 5,400 worker-weekly observations (974 unique workers)
for weeks 1-8. The first column applies the low MW of $6.79 (the current MW in weeks 1-8)
and the second column applies the high MW of $7.21 (the new MW to take effect in weeks 9-
12).
 This is the author’s accepted manuscript without copyediting, formatting, or final corrections. It will be published in its final form in an upcoming issue of
Journal of Labor Economics, published by The University of Chicago Press. Include the DOI when citing or quoting: https://doi.org/10.1086/716347
Copyright 2021 The University of Chicago Press.

Table 2: Worker output per hour by quintiles of worker fixed effects

(1) (2) (3) (4) (5) (6)

Dependent var.: log (output per hour)
All Exclude transition weeks Exclude final week

A. Full sample
Post x Low FE 0.046*** 0.070*** 0.043**
(0.016) (0.020) (0.017)
Post x Quintile 1 0.045* 0.080*** 0.044*
(0.024) (0.028) (0.025)
Post x Quintile 2 0.052*** 0.065*** 0.046***
(0.012) (0.016) (0.012)
Observations 18,471 18,471 12,675 12,675 17,445 17,445
R-squared 0.565 0.564 0.577 0.577 0.536 0.536

B. Balanced sample
Post x Low FE 0.043** 0.068*** 0.040**
(0.016) (0.019) (0.017)
Post x Quintile 1 0.043* 0.080*** 0.041
(0.024) (0.028) (0.026)
Post x Quintile 2 0.049*** 0.061*** 0.044***
(0.012) (0.015) (0.012)
Observations 16,756 16,756 11,247 11,247 15,730 15,730
R-squared 0.567 0.567 0.574 0.574 0.538 0.538
Notes: Panel A is based on the full sample. Panel B is based on the balanced sample including workers who worked in both the
pre (weeks 6-8) and post (weeks 9-12) periods. Based on transactions during weeks 6-12. Post=1 if week 9 or later. Transition
weeks refer to weeks 8 and 9. The quintiles are based on the worker fixed effects estimated based on equation (1) using data
from weeks 1-5. Low FE indicates worker fixed effects are in the bottom 40th percentile. All regressions include percentile
dummies for pre-estimated worker effects, day FE, a cubic polynomial of worker experience, variety-field FE, a cubic
polynomial of variety-field life-cycle, and supervisor FE. Robust standard errors clustered by day in parentheses. *** p<0.01, **
p<0.05, * p<0.1
This is the author’s accepted manuscript without copyediting, formatting, or final corrections. It will be published in its final form in an upcoming issue of
Journal of Labor Economics, published by The University of Chicago Press. Include the DOI when citing or quoting: https://doi.org/10.1086/716347
Copyright 2021 The University of Chicago Press.
 This is the author’s accepted manuscript without copyediting, formatting, or final corrections. It will be published in its final form in an upcoming issue of
Journal of Labor Economics, published by The University of Chicago Press. Include the DOI when citing or quoting: https://doi.org/10.1086/716347
Copyright 2021 The University of Chicago Press.

Table 3: Imposing restrictions on the minimum number of spells when estimating worker fixed effects

(1) (2) (3) (4) (5) (6)

Dependent var.: Log (output per hour)

Post x Low FE 0.050*** 0.049*** 0.046*** 0.034** 0.043** 0.041**

(0.016) (0.015) (0.016) (0.016) (0.016) (0.016)

Observations 19,446 19,249 18,471 17,827 17,460 16,752

R-squared 0.566 0.568 0.564 0.567 0.570 0.568
Minimum #spells 3 4 5 6 7 8
Average #spells 12 13 13 14 14 15
Notes: Column (3) replicates the result in column (2) of Table 2. Columns (1) and (2) show results
with lower numbers of minimum spells and columns (4)-(6) with higher numbers. Based on
transactions during weeks 6-12. Post=1 if week 9 or later. The quintiles are based on the worker
fixed effects estimated based on equation (1) using data from weeks 1-5. All regressions include
percentile dummies for pre-estimated worker effects, day FE, a cubic polynomial of worker
experience, variety-field FE, a cubic polynomial of variety-field life-cycle, and supervisor FE. Robust
standard errors clustered by day in parentheses. *** p<0.01, ** p<0.05, * p<0.1
 This is the author’s accepted manuscript without copyediting, formatting, or final corrections. It will be published in its final form in an upcoming issue of
Journal of Labor Economics, published by The University of Chicago Press. Include the DOI when citing or quoting: https://doi.org/10.1086/716347
Copyright 2021 The University of Chicago Press.

Table 4: Daily employment outcomes by low versus high fixed effect workers

(1) (2) (3) (4) (5) (6)

Dependent var.:
Daily employment Daily hours worked
(extensive margin) (intensive margin)
Exclude workers who likely Conditional on working
All workers exited the farm that day
Mean of D.V.: 0.397 0.642 5.596

Post x Low FE -0.011 -0.030 0.028

(0.020) (0.027) (0.125)
Low FE -0.025** -0.019 -0.066*** -0.054*** -0.312*** -0.323***
(0.010) (0.016) (0.013) (0.015) (0.070) (0.109)

Day FE Yes Yes Yes Yes Yes Yes

Observations 36,038 36,038 22,313 22,313 14,323 14,323
R-squared 0.106 0.106 0.167 0.167 0.340 0.340
Notes: Based on worker-day level data for weeks 6-12. Post=1 if week 9 or later. Low FE indicates pre-estimated
worker fixed effects are in the bottom 40th percentile. Daily employment is 1 if the worker is working that day
and 0 otherwise. Daily hours worked are total hours worked by the worker conditional on working that day.
Columns (1) and (2) include all workers in our sample. Columns (3) and (4) exclude worker-days if the worker's
last day of employment at this farm during this season occurs in any week prior to the week of the present
worker-day. Columns (5) and (6) restrict sample to worker-days with non-zero employment hours. All
regressions include day FE. Post is subsumed in day FE. Robust standard errors clustered by day in parentheses.
*** p<0.01, ** p<0.05, * p<0.1
 This is the author’s accepted manuscript without copyediting, formatting, or final corrections. It will be published in its final form in an upcoming issue of
Journal of Labor Economics, published by The University of Chicago Press. Include the DOI when citing or quoting: https://doi.org/10.1086/716347
Copyright 2021 The University of Chicago Press.

Table 5: Implication of worker effort responses on the firm's minimum wage compliance cost

(1) (2) (3) (4)

MW=6.79 MW=7.21

Minimum wage compliance cost ($) 8,340 13,217 10,660 10,286

Worker effort response No Yes Yes
Change in the allocation of employment hours No No Yes

Implied reduction in compliance cost ($) N/A 2,557 2,931

Notes: Based on 5,400 worker-weekly observations (974 unique workers) for weeks 1-8.
Columns (3) and (4) apply productivity increase of 4.6 percent for workers in the bottom 40th
percentile. Column (4) additionally applies a decreased share of employment hours for workers
in the bottom 40th percentile by 4.6 percent.
This is the author’s accepted manuscript without copyediting, formatting, or final corrections. It will be published in its final form in an upcoming issue of
Journal of Labor Economics, published by The University of Chicago Press. Include the DOI when citing or quoting: https://doi.org/10.1086/716347
Copyright 2021 The University of Chicago Press.
Manuscript Appendix (File Built into Manuscript)
This is the author’s accepted manuscript without copyediting, formatting, or final corrections. It will be published in its final form in an upcoming issue of
Journal of Labor Economics, published by The University of Chicago Press. Include the DOI when citing or quoting: https://doi.org/10.1086/716347
Copyright 2021 The University of Chicago Press.

Appendix
This is the author’s accepted manuscript without copyediting, formatting, or final corrections. It will be published in its final form in an upcoming issue of
Journal of Labor Economics, published by The University of Chicago Press. Include the DOI when citing or quoting: https://doi.org/10.1086/716347
Copyright 2021 The University of Chicago Press.

Figure A.1: Monthly Consumer Price Index for urban wage earners and clerical workers (CPI-W)

Source: US Bureau of Labor Statistics (Series ID: CWUR0000SA0). 1982-84=100. August for each
year is marked on the horizontal axis.
This is the author’s accepted manuscript without copyediting, formatting, or final corrections. It will be published in its final form in an upcoming issue of
Journal of Labor Economics, published by The University of Chicago Press. Include the DOI when citing or quoting: https://doi.org/10.1086/716347
Copyright 2021 The University of Chicago Press.

Figure A.2: Daily number of workers employed and daily total output

This figure plots the relation between the number of workers employed each day and the "field
capacity" proxied by the total output. The label for x-axis (total output) is suppressed to not reveal the
farm’s day-specific scale of operation. R-squared associated with the regression line is 0.7695.
This is the author’s accepted manuscript without copyediting, formatting, or final corrections. It will be published in its final form in an upcoming issue of
Journal of Labor Economics, published by The University of Chicago Press. Include the DOI when citing or quoting: https://doi.org/10.1086/716347
Copyright 2021 The University of Chicago Press.

Figure A.3: Example worker paystub

Employee ID: ABCXYZ

From: Nov 16, 2008 To: Nov 22, 2008
Date Type Hours Rate Pieces Earnings
Nov 16, 2008 Minimum Wage 16.00
Nov 16, 2008 Round 5.33 0.5 58 29.00
Nov 16, 2008 Grape 1.67 3.75 2 7.50
Nov 17, 2008 Grape 3.65 3.75 3 11.25
Nov 17, 2008 Round 4.63 0.5 80 40.00
Total 15.28 103.75
This is the author’s accepted manuscript without copyediting, formatting, or final corrections. It will be published in its final form in an upcoming issue of
Journal of Labor Economics, published by The University of Chicago Press. Include the DOI when citing or quoting: https://doi.org/10.1086/716347
Copyright 2021 The University of Chicago Press.

Figure A.4: Worker-weekly total hours of employment against worker-weekly average productivity

(a) weeks 1-8 (MW=$6.79)

(b) weeks 9-12 (MW=$7.21)

This figure plots the mean of residual of worker-weekly total hours of employment (after accounting
for week fixed effects) by 5 cents bins of worker-weekly average productivity (output per hour) for a
2 dollar window around the relevant minimum wage ($6.79 in panel (a) and $7.21 in panel (b)).
Quadratic fit with 95 percent confidence interval is shown on either side of the minimum wage. It
shows that the employment hours in the record are smooth around the minimum wage with no sign of
a discontinuous drop before the minimum wage.
This is the author’s accepted manuscript without copyediting, formatting, or final corrections. It will be published in its final form in an upcoming issue of
Journal of Labor Economics, published by The University of Chicago Press. Include the DOI when citing or quoting: https://doi.org/10.1086/716347
Copyright 2021 The University of Chicago Press.

Figure A.5: McCrary test: density of worker-weekly productivity around the minimum wage threshold

(a) weeks 1-8 (MW=$6.79)

(b) weeks 9-12 (MW=$7.21)

This figure shows the McCrary plot, which tests for selective sorting around the threshold on the
worker-weekly average productivity (output per hour). The vertical line is the relevant minimum
wage, $6.79 in panel (a) and $7.21 in panel (b). The figure shows no discontinuity in the density of
observations around the minimum wage.
This is the author’s accepted manuscript without copyediting, formatting, or final corrections. It will be published in its final form in an upcoming issue of
Journal of Labor Economics, published by The University of Chicago Press. Include the DOI when citing or quoting: https://doi.org/10.1086/716347
Copyright 2021 The University of Chicago Press.

Figure A.6: Distribution of worker fixed effects

This figure shows the distribution of worker fixed effects as estimated by equation (1) using 13,291
observations from 974 unique workers during weeks 1-5. The mean (standard deviation) of the
estimated FE is 0.0000 (0.2254).
This is the author’s accepted manuscript without copyediting, formatting, or final corrections. It will be published in its final form in an upcoming issue of
Journal of Labor Economics, published by The University of Chicago Press. Include the DOI when citing or quoting: https://doi.org/10.1086/716347
Copyright 2021 The University of Chicago Press.

Figure A.7: Distribution of estimated productivity change based on simulated worker effects

This figure plots the histogram of the estimated productivity change from the pre to the post period
(parameter δ in equation (2)) based on simulated worker effects. Worker effects are drawn from the
normal distribution with mean (standard deviation) of 0.0000 (0.2254), the sample mean and SD
among the estimated fixed effects). Based on 500 reps. The short dashed line shows the 95th
percentile.
 This is the author’s accepted manuscript without copyediting, formatting, or final corrections. It will be published in its final form in an upcoming issue of
Journal of Labor Economics, published by The University of Chicago Press. Include the DOI when citing or quoting: https://doi.org/10.1086/716347
Copyright 2021 The University of Chicago Press.

Table A.1: Lowest wage occupations in Florida

Hourly wages
10th 25th 75th 90th
Occupation title percentile percentile median percentile percentile mean

A. All Occupations 8.02 10.06 14.58 22.58 34.26 18.96

B. Major occupation groups

Food preparation and serving related occupations 7.41 7.71 8.82 11.22 14.69 10.09
Farming, fishing, and forestry occupations 7.43 7.80 8.86 11.35 15.32 10.34
Personal care and service occupations 7.51 8.10 9.75 12.99 18.80 11.63

C. Detailed occupations
Combined food preparation and serving workers, including fast food 7.35 7.57 8.10 9.15 10.80 8.66
Dining room and cafeteria attendants and bartender helpers 7.35 7.57 8.16 9.33 11.41 8.81
Cooks, fast food 7.36 7.57 8.20 9.28 10.83 8.65
Dishwashers 7.36 7.57 8.23 9.27 10.78 8.62
Amusement and recreation attendants 7.37 7.62 8.39 9.95 12.69 9.27
Counter attendants, cafeteria, food concession, and coffee shop 7.38 7.64 8.39 9.45 11.28 8.83
Cashiers 7.34 7.54 8.44 9.41 11.18 8.83
Waiters and waitresses 7.37 7.61 8.56 11.21 14.54 9.91
Bartenders 7.38 7.62 8.66 11.19 15.29 10.11
Farmworkers and laborers, crop, nursery, and greenhouse 7.41 7.73 8.66 10.58 13.16 9.63
Notes: This table shows the mean and various percentiles of hourly wages for the lowest paying occupations in Florida. Based on May 2009 estimates
for Florida from Occupational Employment Statistics (OES) provided by the US Bureau of Labor Statistics. Panel A shows the statistics for all
occupations in the state. Panel B lists the three lowest paying major occupation groups, out of 22 in the Standard Occupational Classification (SOC).
Panel C lists the ten lowest paying detailed occupations with at least 20,000 employment count. There are about 800 detailed occupations in the
SOC.
This is the author’s accepted manuscript without copyediting, formatting, or final corrections. It will be published in its final form in an upcoming issue of
Journal of Labor Economics, published by The University of Chicago Press. Include the DOI when citing or quoting: https://doi.org/10.1086/716347
Copyright 2021 The University of Chicago Press.

Table A.2: Minimum wage in Florida, 2000-2015

Change
in
Federal Florida Florida
Minimum Minimum Minimum
Wage Wage Wage Florida Effective Date
* 2000 $5.15 $5.15
2001 $5.15 $5.15 $0.00
2002 $5.15 $5.15 $0.00
2003 $5.15 $5.15 $0.00
2004 $5.15 $5.15 $0.00
** 2005 $5.15 $6.15 $1.00 05/02/2005 12/31/2005
2006 $5.15 $6.40 $0.25 01/01/2006 12/31/2006
2007 $5.85 $6.67 $0.27 01/01/2007 12/31/2007
2008 $6.55 $6.79 $0.12 01/01/2008 12/31/2008
2009 $6.55 $7.21 $0.42 01/01/2009 7/23/2009
*** 2009 $7.25 $7.25 $0.04 7/24/2009 12/31/2009
*** 2010 $7.25 $7.25 $0.00 01/01/2010 12/31/2010
*** 2011 $7.25 $7.25 $0.00 01/01/2011 5/31/2011
**** 2011 $7.25 $7.31 $0.06 06/01/2011 12/31/2011
2012 $7.25 $7.67 $0.36 01/01/2012 12/31/2012
2013 $7.25 $7.79 $0.12 01/01/2013 12/31/2013
2014 $7.25 $7.93 $0.14 01/01/2014 12/31/2014
2015 $7.25 $8.05 $0.12 01/01/2015 12/31/2015
Source: Florida Department of Economic Opportunity, October 2015
* 2000-2004, The Federal Minimum Wage
** Florida enacted a state minimum wage (Florida Minimum Wage Amendment
approved through election ballot on November 2, 2004).
*** Florida defaulted to the Federal minimum wage
**** Legal ruling raising the minimum wage to $7.31
This is the author’s accepted manuscript without copyediting, formatting, or final corrections. It will be published in its final form in an upcoming issue of
Journal of Labor Economics, published by The University of Chicago Press. Include the DOI when citing or quoting: https://doi.org/10.1086/716347
Copyright 2021 The University of Chicago Press.

Table A.3: Mean employment, hours worked, and productivity

(1) (2) (3) (4)

daily employment

exclude workers
who likely exited daily hours if
all workers the farm worked that day output per hour

A. weeks 1-5
Quintile 1 0.49 0.53 4.90 7.59
Quintile 2 0.51 0.55 4.96 8.85
Quintile 3 0.53 0.59 5.06 9.97
Quintile 4 0.55 0.57 5.03 11.13
Quintile 5 0.52 0.55 4.96 11.94
All 0.52 0.56 4.99 10.00

B. weeks 6-8
Quintile 1 0.47 0.59 4.98 8.16
Quintile 2 0.50 0.64 5.33 8.91
Quintile 3 0.48 0.69 5.36 9.77
Quintile 4 0.53 0.66 5.41 10.83
Quintile 5 0.49 0.63 5.45 11.88
All 0.49 0.64 5.31 9.93

C. weeks 9-12
Quintile 1 0.26 0.57 5.76 6.84
Quintile 2 0.31 0.61 5.90 7.59
Quintile 3 0.30 0.68 6.03 7.86
Quintile 4 0.36 0.70 6.25 8.76
Quintile 5 0.30 0.65 6.14 9.70
All 0.31 0.64 6.03 8.21

D. all weeks
Quintile 1 0.41 0.56 5.12 7.62
Quintile 2 0.44 0.59 5.30 8.56
Quintile 3 0.45 0.64 5.37 9.41
Quintile 4 0.48 0.63 5.45 10.39
Quintile 5 0.44 0.59 5.39 11.37
All 0.44 0.60 5.33 9.54
Obs. 56,976 41,966 25,252 31,762
Notes: Based on 974 unique workers, this table shows the mean employment, hours, and productivity (output
per hour) by quintiles of individual FE, separately for the weeks 1-5, 6-8, and 9-12, and for the overall season.
Observations in columns 1-3 are at the worker-day level. Observations in column 4 are at the spell level (more
than one spell is possible for a given day). Column 1 includes all possible worker-days. Column 2 excludes
worker-days from week K if the worker's last day of employment during the season falls in any week prior to
week K. Column 3 reports daily hours worked conditional on working that day. Column 4 reports output per
hour measured in $/hr.
 This is the author’s accepted manuscript without copyediting, formatting, or final corrections. It will be published in its final form in an upcoming issue of
Journal of Labor Economics, published by The University of Chicago Press. Include the DOI when citing or quoting: https://doi.org/10.1086/716347
Copyright 2021 The University of Chicago Press.

Table A.4: Using different specifications in accounting for the baseline productivity of workers

(1) (2) (3) (4) (5) (6)

Dependent var.: log (output per hour)
Percentile Quintile Linear
Post x Low FE 0.046*** 0.039** 0.048***
(0.016) (0.016) (0.014)
Post x Quintile 1 0.045* 0.037 0.061***
(0.024) (0.024) (0.021)
Post x Quintile 2 0.052*** 0.048*** 0.044***
(0.012) (0.012) (0.010)
Baseline productivity Yes Yes Yes Yes Yes Yes
Day FE Yes Yes Yes Yes Yes Yes
Observations 18,471 18,471 18,471 18,471 18,471 18,471
R-squared 0.565 0.564 0.543 0.543 0.559 0.558
Notes: This table employs different specifications to account for the baseline productivity of workers. Columns (1)-(2) use the
same specification as in Table 2 and include dummies for each percentile of the pre-estimated worker effects. Columns (3)-(4)
include dummies for each quintile of the worker effects. Columns (5)-(6) include the pre-estimated worker effects in levels.
Based on transactions during weeks 6-12. Post=1 if week 9 or later. All regressions include day FE, a cubic polynomial of
worker experience, variety-field FE, a cubic polynomial of variety-field life-cycle, and supervisor FE. Robust standard errors
clustered by day in parentheses. *** p<0.01, ** p<0.05, * p<0.1