John J. Horton
Price floors and employer preferences:
Evidence from a minimum wage experiment

American Economic Review, 2025, vol. 115, No. 1

Seiro Ito

TL; DR

  • Im many studies: MW↑ ⇒ \(\Delta\) employment \(\simeq 0\)
  • One reason: Oligopsony
  • This paper: Labor-labor substitution
    • Unskilled employment↓
    • Skilled employment↑
    • Net effect \(\simeq 0\)
    • Hours↓ ← Skilled requires fewer hours
  • MW experiment on an online labor market: $2, $3, $4 per hour
    • Skill/productivity proxy = past wage rates, past hours

Why are disemployment impacts of minimum wage modest?


Firms can adjust in other dimensions (even under non-oligopsony)


MW↑

  • Employee head counts↓
  • Employment hours↓
  • Nonmonetary compensation↓
  • Profits↓
  • Output prices↑
  • Labour productivity ↑ (labour-labour substitution)

A problem: Not all adjustments are observable in firm surveys…

Previous literature

Insufficient data

Even within the “low wage” category/type, more productive workers get jobs while less productive workers lose out

  • Net impact \(\simeq 0\): Averaging gives muted disemployment impacts
  • Previous studies did not show this ← Lack of productivity proxy

Experiment design

  • Online job platform: Post and bid
  • Random assignment of hourly MW: $2, $3, $4
  • Applicants: Not aware of MW, but a bid < MW not accepted
  • Employers: Not aware of MW, see applicant profiles
  1. Firms post hourly job opening ← immediate treatment arm assignment
    • Arms: $2 (8%), $3 (8%), $4 (8%), control (75%) assigned to firms, not jobs
    • Job description, skills required, expected duration, bid closing time
  2. Workers bid the hourly wage← bid < MW not accepted
  3. Workers with the lowest bid receive a contract

No reference to the name of the platform

Data

  • N=156,656 (employers)
  • No info on dates, experiment duration, DID uses Jan 2014 - Jan 2015
  • No info on employers except job category
  • Reshaped data: Category monthly panel, individual (\(\pm\) 2 week window) panel
    • DID: Announcement/imposition impacts on wage, employment, etc.

I wonder if nondisclosure agreement has to do with…

No descriptive statistics shown, difficult to understand: all posted jobs \(\subset\) all jobs

What are the differences:

  • Sample vs. population of all jobs
  • Outsourced jobs vs. internalized jobs

In DID, are the trends of higher paid jobs appropriate as the control group?

Headline results

  • Headcounts↓ small
  • Employment probability↓ by 12% for workers < MW4
  • Hours↓ by 27% (for low wage jobs, MW4)
  • Employment substitution towards more productive workers

Threats to internal validity … unlikely

  1. Failed randomization
  2. Applicant attrition before wage bidding
  3. Applicant sorting across arms (MW levels)
    • # of applicants balance between arms
    • Practically difficult to know the employer’s treatment assignment
  4. Employer sorting through time (reposts the job later)
    • # of filled posts: before < after ← data show oppositte
  5. Employer sorting to other platform
    • # of posts: before > after← not supported by data

I would worry more about the common trend violation for DID/ES estimates

Where to look for the impacts

What are the “low paying jobs”?

  1. Administrative support jobs: Lowest mean wage in all job categories
  2. Predictive model (LASSO) of wages using pre-experiment data: Pay less than $5

Mechanism behind the post-bid

An example case:

A \(\succ\) B \(\succ\) C \(\succ\) No hire

wage\(_{A}\) < wage\(_{r}\) < wage\(_{B}\) < MW < wage\(_{C}\)

  • wage\(_{A}\) is not considered
  • B’(MW, \(y_{B})\) \(\prec\) C(wage_{C}, \(y_{C})\)

Why do applicants bid different \(\frac{w}{y}\)?

  • Different reservation wages
  • Not take own characteristics in bids

Estimation

Experiment on employer \(j\)

\[ y_{j}=\beta_{0}+\beta_{2}MW2_{j}+\beta_{3}MW3_{j}+\beta_{4}MW4_{j}+\epsilon_{j} \tag{1} \]

  • MWs+control are MECE: Goldsmith-Pinkham, Hull, and Kolesár (2024) bias applies
  • Hereafter, I will assume all the estimates are legit…

Confirmatory event study (platform-wide MW) of treated category, category \(i\)

\[ y_{it}=\sum_{t=s}^{S}\beta_{t} A\small{DMIN}+\alpha_{i}+\delta_{t}+\gamma_{i}t+\epsilon_{it} \tag{2} \]

Applicant impact DID (platform-wide MW), applicant \(i\), posted job \(j\), wage band \(k\)

\[ y_{ij}=\sum_{k\in K}\beta_{k}\left(P\small{OST}_{ij}\times \normalsize{P}\small{RE}\normalsize{W}\small{AGE}\normalsize{B}\small{AND}^{k}_{i}\right)+c_{i}+\epsilon_{ij} \tag{3} \]

Results

Employment↓

Hours↓↓

Earnings \(\simeq 0\)

Sample size is large1


More for

  • MW4
  • Low predicted wage jobs

Results

Hired workers’ past wage↑, cumulative past earnings↑

  • Not so much for Admin jobs

Hired workers’ profile rate↑

Hired workers’ markup rate of bids↑


More for

  • US based workers
  • Low predicted wage jobs
    • Not so much for Admin jobs
  • Effects on Non-US workers: Because of many less productive workers or country effect?

Results

Wage↑ sources …All three need to be accounted for

  1. Selection (by firms slashing less productive, low wage jobs)
  • This canot be the only reason
  • If so, closure of lowest 10% of all jobs ⇒ 10%↑ in wages ← control group elasticity = 1
  • MW3 Admin: \(\Delta\)# of jobs = -2%, mean wage↑ \(\simeq\) +8% (All jobs is -2%, +25%…), more than selection happens here
  1. Substitution (Increase in worker productivity and pay higher wages)
  • Employ higher paid applicants (Fig 5, past wage +15% for MW4, past earnings +27% for MW4)
  1. Markup (No change in worker productivity but pay higher wages)
  • Markup rate↑ (Fig 5, +10% or higher for all MWs)
  • A windfall gain for hired workers

Confirmatory DID (Event study)

Online, partial, secret RCT


Leaves questions

  1. Is L-L substitution useful margin of adjustments in equilibrium? yes
  2. Does employer knowledge on MW matter? no
  3. Do quantity and contents of jobs change after MW imposition? yes
  4. Do the worker impacts remain the same in equilibrium? yes

Confirmatory DID

Platform wide imposition of MW $3/hour

  • Announcement
  • Imposition (2.5 months later)
  1. Single diff (Before-after): Compare market outcomes
  2. Double diff (DID): Use the categories that show little difference in wages as the control

Reconstruct (reshape) data: Category panel data

\[ y_{it}=\sum_{t=s}^{S}\beta_{t} ADMIN+\alpha_{i}+\delta_{t}+\gamma_{i}t+\epsilon_{it} \]

Confirms most of experiment results

Jobs posted↓, hours↓, past earnings ↑, past hours↑

Labor-labor substitution

Confirmatory DID

  • Jobs posted↓ (mainly hourly jobs↓)
  • \(\Delta\) Fraction filled \(\simeq 0\) or ↓ (-.02 vs. -.01 in experiment)
  • Mean hours ↓ (-.44 vs. -.08 in experiment)…why larger is not discussed
  • \(\Delta\) Log mean earnings \(\simeq 0\); (same in experiment)
  • Log past mean wage of hired worker↑ (+.07 vs. +.09 in experiment MW3)1
  • Log past mean earnings of hired worker↑ (+.2 vs. +.05 in experiment)
  • Log past hours of hired worker↑

Applicant impact DID

Applicant \(i\), posted job \(j\), wage band \(k\) \[ y_{ij}=\sum_{k\in K}\beta_{k}\left(P\small{OST}_{ij}\times \normalsize{P}\small{RE}\normalsize{W}\small{AGE}\normalsize{B}\small{AND}^{k}_{i}\right)+c_{i}+\epsilon_{ij} \tag{3} \]

Wrong x-axis labels:

(0, 1], (1, 2], (2, 3], …

  • No spillover of MW to higher wages
  • Pr[hiring]↓ for pre-period wage < MW and higher
  • # of hires↓ for pre-period wage < MW and higher
  • # of bids↓ for pre-period wage < MW … they gave up?

Reshape (= estimate DID to control typical new applicant wage growth: bid low → bid high) to

long applicant panel

[-14 days, +14 days] ×

{-1 year, imposition year} ×

{all jobs applied} ×

{wage, results}

short applicant panel

[-14 days, +14 days] ×

{-1 year, imposition year} ×

{# of jobs applied, # of hires}

Conclusions

Experimental findings

  1. Wages of hired workers increase
  2. Prob of hiring decreases at a sufficiently high MW
  3. Hours decrease
  4. Reduction in hours can be explained in part by labor-labor substitution

Observational findings

  1. Wages of hired workers increase
  2. Employers posted fewer jobs “that would have likely paid lower wages”
    • They showed reduction in all jobs combined, not only the low paid jobs
  3. Hours decrease
  4. Firms substitute toward more productive workers
  5. Prob of hiring for low wage workers reduced

Conclusions

Labor-labor substitution is an important margin of adjustment for firms

  • Less productive gets less jobs, more productive gets more jobs
    • Heterogeneity in disemployment impacts even within a cateory/type
    • Averaging labor-labor substitution gives muted impacts

higher minimum wages should be paired with more targeted exemptions for lower-productivity workers likely to be displaced.




Workers of different productivity being paid the marginal product in a competitive market?

  • Possible, firms can choose the hour-productivity pair that suits them best
  • If the labor market is thin, some jobs may not get filled, foreseeing this, firms may not post jobs

感想

  • 最低賃金↑ ⇒ 雇用労働の熟練↑, 雇用時間量↓ を実証的に示したのは画期的
    • 南アの農場主が対応策として同じことを話していた
    • Oligopsonyとの関係? オンライン労働市場M Turkは買い手寡占的(Dube et al. 2020)
  • 低技能労働が代替されて失職、全労働の純効果はゼロに近いことを示したのも画期的
    • Cengiz et al. (2019): # of jobs lost below MW \(\simeq\) # of jobs gained above MW
    • Data: Pooled cross section ⇒ Estimates are counts, not individuals moving from below to above MW, laid off workers can be out of labor force

感想

  • 記述統計すらないので結果がどの母集団に一般化できるのか不明
    • Single taskのアウトソース作業 vs. フルタイム雇用: 後者では生産性の差はそこまでないかも
      • 植樹・幡種、剪定、収穫などの単作業の仕事は労働の高熟練化に帰結しやすい?
      • 論文での労働の高熟練化は資本装備率の差も含まれていないか
    • 統御群=高賃金仕事はDID推計で適切か
  • 最低賃金規制は変な政策
    • Single taskのアウトソース作業は他の生産工程から分離
    • 労働需要+最低賃金規制=部品需要+最低時間単価規制
    • 最低時間単価を規制する政策=他にはない変わった政策
    • 低効率な部品供給者を閉め出す(=労働時間を過少申告させる)効果あり
  • 著者が雇用されていたプラットフォームのデータ、情報非開示合意書NDAでexternal validityを示す手段を失ったのでは?
  • プラットフォームが利潤のために実施した実験だからIRB不要、は斬新

References

Allegretto, Sylvia A., Arindrajit Dube, and Michael Reich. 2011. “Do Minimum Wages Really Reduce Teen Employment? Accounting for Heterogeneity and Selectivity in State Panel Data.” Industrial Relations: A Journal of Economy and Society 50 (2): 205–40. https://doi.org/10.1111/j.1468-232X.2011.00634.x.
Card, David. 1992. “Using Regional Variation in Wages to Measure the Effects of the Federal Minimum Wage.” ILR Review 46 (1): 22. http://search.ebscohost.com.kras1.lib.keio.ac.jp/login.aspx?direct=true&db=buh&AN=9301271052&lang=ja&site=ehost-live.
Card, David, and Alan B Krueger. 1995. “Time-Series Minimum-Wage Studies: A Meta-Analysis.” The American Economic Review 85 (2): 238–43.
Cengiz, Doruk, Arindrajit Dube, Attila Lindner, and Ben Zipperer. 2019. “The effect of minimum wages on low-wage jobs.” The Quarterly Journal of Economics 134 (3): 1405–54. https://doi.org/10.1093/qje/qjz014.
Clemens, Jeffrey, and Michael Wither. 2019. “The Minimum Wage and the Great Recession: Evidence of Effects on the Employment and Income Trajectories of Low-Skilled Workers.” Journal of Public Economics 170: 53–67. https://doi.org/https://doi.org/10.1016/j.jpubeco.2019.01.004.
Dube, Arindrajit, Jeff Jacobs, Suresh Naidu, and Siddharth Suri. 2020. “Monopsony in Online Labor Markets.” American Economic Review: Insights 2 (1): 33–46. https://doi.org/10.1257/aeri.20180150.
Dube, Arindrajit, T. William Lester, and Michael Reich. 2010. “Minimum Wage Effects Across State Borders: Estimates Using Contiguous Counties.” The Review of Economics and Statistics 92 (4): 945–64. https://doi.org/doi.org/10.1162/REST_a_00039.
Fairris, David, and Leon Fernandez Bujanda. 2008. “The Dissipation of Minimum Wage Gains for Workers Through Labor-Labor Substitution: Evidence from the Los Angeles Living Wage Ordinance.” Southern Economic Journal 75 (2): 473–96. https://doi.org/10.1002/j.2325-8012.2008.tb00915.x.
Giuliano, Laura. 2013. “Minimum Wage Effects on Employment, Substitution, and the Teenage Labor Supply: Evidence from Personnel Data.” Journal of Labor Economics 31 (1): 155–94. https://doi.org/10.1086/666921.
Goldsmith-Pinkham, Paul, Peter Hull, and Michal Kolesár. 2024. “Contamination Bias in Linear Regressions.” American Economic Review 114 (12): 4015–51. https://doi.org/10.1257/aer.20221116.
Katz, Lawrence F., and Alan B. Krueger. 1992. “The Effect of the Minimum Wage on the Fast-Food Industry.” ILR Review 46 (1): 6. https://doi.org/10.1177/001979399204600102.
Katz, Lawrence F., and Kevin M. Murphy. 1992. “Changes in relative wages, 1963-1987: Supply and demand factors*.” The Quarterly Journal of Economics 107 (1): 35–78. https://doi.org/10.2307/2118323.
Meer, Jonathan, and Jeremy West. 2015. “Effects of the Minimum Wage on Employment Dynamics.” Journal of Human Resources. https://doi.org/10.3368/jhr.51.2.0414-6298R1.
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): 608–48. https://doi.org/10.1177/00197939140670S307.
Powell, David. 2022. “Synthetic Control Estimation Beyond Comparative Case Studies: Does the Minimum Wage Reduce Employment?” Journal of Business & Economic Statistics 40 (3): 1302–14. https://doi.org/10.1080/07350015.2021.1927743.
Rambachan, Ashesh, and Jonathan Roth. 2023. “A more credible approach to parallel trends.” The Review of Economic Studies, February. https://doi.org/10.1093/restud/rdad018.