Economic Modelling 148 (2025) 107086

Contents lists available at ScienceDirect

Economic Modelling
journal homepage: www.journals.elsevier.com/economic-modelling

Minimum wage and unemployment in Russia: A new look on an old

constructI
Gi Khan Ten, Shun Wang ∗
Xi’an Jiaotong-Liverpool University, China

ARTICLE INFO ABSTRACT

Dataset link: Minimum Wage and Unemployme We investigate the unemployment effects of the minimum wage policy in Russia, focusing on possible reasons
nt in Russia: A New Look on an Old Construct ( for the heterogeneity of the effects documented in previous literature. Using region-level data, we show that the
Original data) policy increases both the unemployment rate among young workers and the rate of informality. We corroborate
JEL classification:
these findings by leveraging a sudden increase in the minimum wage in Kamchatka as a natural experiment.
J1 Next, we show that the magnitude of employment responses to minimum wage changes depends on the
J2 elasticity of capital-labor substitution, with stronger effects observed in industries where capital and labor
J3 are more substitutable. When substitution is not feasible, employers respond to the policy by hiring workers
J6 informally. Consistent with revealed separations and informal recruitment, we find limited income effects of
the policy. These findings highlight the importance of accounting for existing production technologies and the
Keywords:
Minimum wage extent of non-compliance when raising the wage floor.
Unemployment
Informality
Capital-labor substitution
Russia

1. Introduction Azar et al. (2024) show that employment responds positively to a

higher minimum wage in U.S. counties characterized by sufficiently
The minimum wage continues to be a controversial policy tool. high labor-market concentration.1
Some studies show that the policy reduces income inequality by raising In addition to labor-market concentration, two additional forces —
the bottom tail of the income distribution while the disemployment
the focus of this paper — may moderate the effect of the minimum
effect is either absent or negligible (e.g., Katz and Krueger, 1992; Dube
wage on unemployment. First, the existence of an informal labor sector
et al., 2010; Draca et al., 2011; Giuliano, 2013; Cengiz et al., 2019).
Other studies show that higher labor costs destroy jobs; therefore, the can act as a buffer, allowing firms to bypass the wage floor by em-
minimum wage policy potentially harms rather than protects workers ploying workers off the books. A key feature of such employment is its
(e.g., Neumark and Wascher, 2010; Meer and West, 2016; Clemens and invisibility to local legislators, which allows firms to retain workers by
Wither, 2019). A nascent strand of literature shows that the minimum paying below the minimum wage. Second, the substitutability between
wage could even have a positive impact on employment. In a perfectly capital and labor, i.e., the shape of the two-input production isoquant,
competitive labor market, the elasticity of labor demand with respect may influence how firms respond to higher labor costs. In industries
to the minimum wage is non-positive. However, higher labor-market where capital and labor are easily interchangeable, firms may opt to
concentration, with monopsony being an extreme case, can result in
substitute labor with capital, exacerbating the disemployment effects of
increased employment following a minimum wage hike. For example,

I The authors would like to thank Sushanta Mallick (Editor), the Associate Editor who handled this manuscript as an anonymous referee, and one anonymous
referee for their valuable feedback. We are also grateful to the participants of the 15th Western Economic Association International Conference (Tokyo), the
16th International Symposium on Human Capital and Labor Markets Conference (Beijing), and seminar attendees at Xi ’an Jiaotong-Liverpool University and the
KDI School of Public Policy and Management for their helpful comments and suggestions. Gi Khan Ten gratefully acknowledges financial support from the Bisa
Research Grant provided by Keimyung University in 2022 (Project No. 20220254)
∗ Corresponding author.
E-mail address: [email protected] (S. Wang).
1
Studies also show that labor-market concentration affects wages directly (e.g., Bassanini et al., 2024) and moderates the impact of minimum wages on
self-employment (e.g., Glasner, 2021).

https://doi.org/10.1016/j.econmod.2025.107086
Received 8 April 2024; Received in revised form 22 March 2025; Accepted 23 March 2025
Available online 1 April 2025
0264-9993/© 2025 Elsevier B.V. All rights are reserved, including those for text and data mining, AI training, and similar technologies.
G.K. Ten and S. Wang Economic Modelling 148 (2025) 107086

a higher wage floor. Conversely, in sectors where capital-labor substi- capital for labor and hire workers off the books are two key factors
tution is not feasible, the minimum wage may lead to less pronounced moderating the employment effects of the minimum wage policy.
disemployment effects. Finally, we assess whether the Russian minimum wage policy aids
This paper investigates the aforementioned possibilities in the con- the bottom tail of the country’s income distribution by replicating
text of the minimum wage policy in Russia. Four key findings emerge. Dube’s (2019) analysis of U.S. data. Our results show that the policy’s
First, the policy does not affect the overall unemployment rate, but it capability to aid economically disadvantaged families in Russia is, at
has a small positive impact on the unemployment rate among workers best, limited. Job losses and informal recruitment are two revealed
under 30. Second, the share of informally employed workers responds labor-market responses that limit the minimum wage policy’s ability
positively to a higher minimum wage. Third, job separations among to raise the incomes of the poorest.
young workers are concentrated in industries that can substitute capital Our findings contribute to several strands of minimum wage liter-
for labor more easily. Fourth, consistent with job losses among young ature. First, they help explain the disparities in estimated employment
workers and the increased informalization of the labor market, the effects across generations of studies. For example, Cengiz et al. (2019)
minimum wage policy has minimal capacity to increase household utilize a long panel of U.S. states and report no overall employment
income. Altogether, these findings suggest that incompliance — man- effects, while Harasztosi and Lindner (2019) document minimal em-
ployment effects in Hungary. These findings differ from earlier studies
ifested by employers’ ability to recruit workers informally — and
that suggest higher minimum wages harm employment (e.g., Sabia
strong complementarity between labor and capital may explain the lack
et al., 2012; Neumark et al., 2014). While such disparities are often
of detectable disemployment effects from the minimum-wage policy
attributed to differences in empirical specifications (e.g., Allegretto
observed in some past studies.
et al., 2011, 2017), our results suggest that cross-country and tem-
We estimate the unemployment effects of minimum wages by ex-
poral differences in labor-market characteristics, such as capital-labor
ploiting two different empirical designs. First, we fit a two-way panel
substitutability and informality, also play a significant role.
fixed-effects model of the regional unemployment rate, accounting
Second, we contribute to the literature on firms’ adjustments to
for potential reverse causality between the local minimum wage and higher minimum wages. In their comprehensive review, Dube and Lind-
the outcome variable. The Russian institutional context provides a ner (2024) discuss various adjustments highlighted in the literature,
plausibly valid instrumental variable (IV): Bodman’s Winter Severity including reductions in fringe benefits, firm exit, price increases, input
Index, which quantifies human perception of winter severity (Bodman, cost reductions, and profit declines. In this regard, the most relevant
1916). The choice of the instrument stems from the observation that studies to our paper examine changes in capital investments in response
labor unions bargain for higher wages when winter ‘‘feels’’ harsher to higher wages (Chen, 2019; Harasztosi and Lindner, 2019; Hau et al.,
than in the past. Our results show that a 1% increase in the regional 2020; Baek et al., 2021). Relatedly, Lordan and Neumark (2018) show
real minimum wage increases the unemployment rate among workers that automatable jobs are particularly vulnerable to minimum wage
aged below 30 by 0.05 percentage points. This estimate aligns with hikes. However, some studies report either no effect or the opposite
the lower end of minimum wage effects reported in the literature. To finding (e.g., Haepp and Lin, 2017; Gustafson and Kotter, 2018). Our
further validate our findings, we rule out obvious channels through results suggest that these discrepancies can be partly explained by
which our proposed instrument may affect the regional unemployment industry-level differences in production isoquants.
rate. Consistent with most past studies, our analysis finds no impact of Finally, our results expand on the literature linking growing in-
the minimum wage on older workers (e.g., Portugal and Cardoso, 2006; formality as a response to minimum wages. Studies show that non-
Neumark et al., 2014). compliance with minimum wage laws limits the policy’s effective-
We substantiate these findings by exploiting a sizable (63%) in- ness in Colombia (Pérez, 2020), Ecuador (Wong, 2019), Honduras
crease in the minimum wage in the Kamchatka region, following the (Ham, 2018), and Indonesia (Comola and Mello, 2011). Our find-
long tradition of investigating the unemployment effects of the min- ings suggest that informal recruitment increases with higher minimum
imum wage policy in quasi-experimental frameworks.2 We leverage wages, particularly where the possibility of capital-labor substitution is
this unexpected minimum wage increase in Kamchatka, comparing its limited.
employment trends with those in the geographically adjacent control The remainder of this paper is organized as follows. Section 2
region: Chukotka. Our findings suggest that workers younger than 30 introduces the institutional context and describes the data. Section 3
faced a three-percentage-point higher probability of unemployment discusses our empirical strategies. Section 4 presents the unemployment
following the 63% increase in the local minimum wage. Differently effects of the minimum wage. Section 5 investigates its income effects.
put, a 1% increase in Kamchatka’s minimum wage increased the unem- Section 6 concludes.
ployment rate among young workers by approximately 0.05 percentage
2. Institutional context and data
points, a result strikingly comparable to that obtained from the region-
level data. Additionally, we show that workers in Kamchatka were
2.1. Minimum wage policy in Russia
more likely to be employed informally after the wage floor was raised.
We next examine cross-industry heterogeneity in the employment The Russian minimum wage policy shares at least two features with
effects of the minimum wage among youth. Our analysis shows that that of the United States. More specifically, first, Russia updates its
younger workers are more likely to face separations in industries wage floor regularly. However, federal subjects (regions henceforth)
where capital and labor are more easily substitutable. These findings have the right to set their minimum, which must be no lower than
align with one of the Hicks–Marshall laws of derived demand, which the national level, and Fig. 1 shows the dynamics of the monthly
predicts that the wage elasticity of labor demand is higher when nominal minimum wage for full-time workers by region and month
other production inputs can replace workers (Allen, 1938; Hamermesh, between 2009 and 2012 (83 regions). Second, with the exception of
1993). Furthermore, industries with little or no employment declines seven regions, the Russian minimum wage does not vary within a
in response to the minimum wage hike tend to employ more workers given region in a given period.3 In six of the seven exceptional cases,
informally. These results suggest that employers’ ability to substitute

3
There are 26 regions that have the national minimum wage above the
2
Kamchatka, located in Russia’s Far East, is known for its remote, sparsely national one at least at one point in time within our study period. Figure
populated terrain and geographic isolation from much of the country. As A1 of the Online Appendix A shows the dynamics of minimum wage in four
detailed in Section 3, it has a geographically adjacent region that serves as categories of regions, with the first (the last) category having the highest
a natural candidate for the comparison group. (lowest) level of the average minimum wage in our study period.

2
G.K. Ten and S. Wang Economic Modelling 148 (2025) 107086

Fig. 1. Nominal minimum wage, by region.

Notes: The figure illustrates the dynamics of the natural logarithm of nominal monthly full-time equivalent minimum wages by month from January 2009 to December 2012.

the difference is either between the public and private sectors (Altai a tendency for colder regions to have higher wage floors, on average;
Krai, Kurgan Oblast, Kursk Oblast, Tula Oblast, and Tyumen Oblast) or and (iii) the introduction of new minimum wages typically occurring
among the public, private, and agricultural sectors (Volgograd Oblast). in the first quarter of the year.
Among these six regions, the largest wage difference between public First, as illustrated in Fig. 1, significant disparities exist in minimum
and private sector workers is 45%, to the detriment of the public sector wage levels across regions. This heterogeneity is, in part, a result of
(Tula Oblast, in 2012). In the remaining region, Karelia, the minimum varying living costs among regions. For instance, Moscow, known as the
wage varies across cities, favoring workers in colder areas. most expensive city in Russia, has the highest wage floor. On average, it
We obtain data on the regional minimum wage from publicly avail- surpasses the federal minimum wage by 4.2 sample standard deviations
able trilateral agreements involving the government, labor unions, and over the entire study period.
employers’ associations. For the aforementioned seven regions, where Second, besides the living cost, another important factor influencing
the minimum wage varies by sectors and/or by cities, we proceed as the variations in both the level and growth of regional minimum wages
follows. In six regions with sector-specific wage floors, we construct is the local winter weather. Specifically, in colder regions, workers
a weighted average minimum wage, using the share of workers in have historically received higher minimum wages. This phenomenon
each sector as weights. In the region with city-specific wage floors, can be attributed to the past implementation of the Northern Multiplier
we weight the minimum wage by the share of the population in each policy, designed to compensate individuals for working under extreme
city. The data needed to construct sectoral or city-level weights is weather conditions. Consequently, in several regions, particularly in
publicly available from the Russian Federal Statistics Service. Finally, the North and Far East, employers were previously mandated to pro-
we collapse the minimum wages to the region-year level. vide minimum wages equal to the state-level wage floor multiplied
To understand the relative size of the Russian minimum wage, by a region-specific multiplier. This multiplier ranged from 1.15 to 2
it is useful to compare the minimum-to-average wage ratio in the and remained in effect until 2007. As a result of the historical pres-
country — a variant of the Kaitz index — with its values observed in ence of the Northern Multiplier, the minimum wage in some regions
countries frequently appearing in the literature. For example, in 2012, consistently exceeded the national level. For example, three regions
the Russian Kaitz index was 0.28.4 This value is slightly higher than with wage floors just below Moscow’s are Yamalo-Nenets, Nenets, and
that of the United States (0.27) during the same period, as reported by Khanty-Mansiysk Autonomous Okrugs, all of which belong to Northern
the International Labor Organization, but considerably lower than in Russia.
other countries frequently mentioned in the minimum wage literature, Region-specific minimum wages are established through trilateral
such as Canada (0.4), Hungary (0.4), and the United Kingdom (0.39). negotiations involving government representatives, labor unions, and
In the following section, we describe three key features of the dynamics employers’ associations. For instance, business representatives may
of minimum wages that are relevant to our econometric identification: propose that the government refrain from increasing the minimum
(i) substantial cross-regional differences in minimum wage levels; (ii) wage if the local unemployment rate is on the rise in a particular region.
Conversely, labor unions advocate for supporting workers during severe
winter conditions, even in areas where the Northern Multiplier has
4
The authors’ calculations using the data obtained from the Federal never been applied (e.g., Southern Russia). The link between win-
Statistics Service. ter severity and the proposed minimum wage level is mediated by

3
G.K. Ten and S. Wang Economic Modelling 148 (2025) 107086

expectations of rising living costs and the desire to provide moral Table 1
Summary statistics.
compensation for harsh working conditions. Consequently, one may
observe an increase in wage floors in certain regions where no other Mean
(Standard deviation)
economic factors, such as steeply rising current consumer prices or a
Panel A: Unemployment rates and informality
low recent unemployment rate, could justify high minimum wages.5
Overall 0.080
This observation is central to our baseline analysis, which we elaborate (0.043)
on in Section 3.
15–19 years old 0.086
Lastly, as Fig. 1 shows, new minimum wages are typically intro- (0.108)
duced at the beginning of the year. This observation likely makes the 20–29 years old 0.069
introduction of the new minimum wage in the middle of the year less (0.072)
anticipated by employers. Based on this conjecture, as a robustness 30+ years old 0.053
check, we investigate whether unemployment responds to a plausibly (0.028)
unexpected sharp minimum wage increase in Kamchatka, which is Share of informally employed 0.125
one of the Russian regions that suddenly increased its wage floor by (0.106)
about 63% in the middle of the year—we elaborate further on this in Panel B: Control variables
Log (MW) 8.440
Section 3.2.
(0.204)
Log (GDP per capita) 12.150
2.2. Data (0.617)
CPI 1.078
Regional Characteristics.—Our investigation begins with an analysis (0.019)
of regional unemployment rates. We obtain regional indicators, includ- Log (Working-age population) 13.520
ing the overall unemployment rate, age-specific unemployment rates (0.844)
(categorized into three age groups: 15–19, 20–29, and 30+), and the Notes: This table presents the means and standard deviations of key variables used in
proportion of the labor force engaged in informal employment, from the baseline region-level analysis. Panel A reports summary statistics for the outcome
the ‘‘Regional statistics’’ section of the Russian Federal Statistics Ser- variables, while Panel B provides statistics for the regressors. All monetary values are
vice’s portal. The Russian Federal Statistics Service defines individuals adjusted to 2010 prices.

working in the informal sector as those employed during the reporting

period in an informal enterprise. An informal enterprise is defined as
an organization that produces goods and services for market sale but all regions annually. Details on the variables used in our analysis are
lacks the legal status of a corporate entity. The key feature of informal provided in Section 4.
employment relevant to our study is its operation outside official Survey of Income and Program Participation, 2014–2017.—Our exam-
oversight. Consequently, the minimum wage floor is, in principle, not ination of the household income effect of the minimum wage policy
binding in such labor arrangements. closely follows Dube (2019). The original study utilizes the U.S. Current
We also collect regional data on real GDP per capita, consumer Population Survey, whereas we use the Russian Survey of Income and
price index, working-age population, health-related metrics (such as Program Participation (SIPP), a comparable dataset to the one used
the regional pupil-to-doctor ratio and per capita disease rates), net in Dube (2019). Conducted annually by the Russian Statistics Service,
migration inflow, agricultural output normalized by GDP, electricity the SIPP collects detailed information on household income sources,
production, and average household utility expenses from the same income levels, and participation in government programs across all
data source. Additionally, we retrieve nominal minimum wages from Russian regions. The number of surveyed individuals varies, covering
trilateral agreements involving labor unions, employers’ associations, 45,000 families in 2014 and up to 160,000 in 2017.
and the government. Finally, we obtain data on temperature and wind
speed from the Hydrometeorological Centre of Russia. Our dataset 3. Empirical strategy
contains the full set of variables for the years 2009 to 2012. Table 1
presents summary statistics. We begin by examining the response of the overall and age-specific
To investigate the relationship between industry-level elasticity of unemployment rates to changes in the minimum wage. Additionally, we
substitution and the employment effects of the minimum wage, we include informal employment as an outcome variable, given its role as
collect additional data on employment indicators (including youth a source of income for individuals unable to secure formal employment
and informal employment), capital investments, average wages, and (Porta and Shleifer, 2008; Comola and Mello, 2011). Furthermore, we
value-added per worker for each region-industry-year cell from the investigate whether the employment impact of the minimum wage pol-
same state’s statistics service. Additionally, we use two survey datasets, icy varies with the industry-specific elasticity of substitution between
which are publicly available from the ‘‘Federal Statistical Surveys on capital and labor. Finally, we assess the implications of the policy for
Socio-Demographic Issues’’ section of the Russian Federal Statistics household income.
Service’s portal, as detailed below.
Russian Labor Force Survey, 2010–2013.—To test the robustness of 3.1. Region-level analysis
our baseline findings from the regional data, we conduct a further
investigation of one region’s higher minimum wage introduction using Our baseline model leverages the region-year panel dataset de-
the Russian Labor Force Survey (RLFS). Conducted by the Russian scribed in the previous section. We estimate the following two-way
State Statistics, the RLFS surveys approximately 70,000 people from fixed-effects model:

𝑦𝑖𝑡 = 𝛽 log 𝑀𝑊𝑖𝑡 + 𝑋𝑖𝑡′ 𝜌 + 𝜗𝑖 + 𝜏𝑡 + 𝑢𝑖𝑡 , (1)

5
Examples include Kemerovo Oblast in southwestern Siberia, Leningrad
Oblast in the European part of Russia, Krasnodar Krai in Southern Russia,
where 𝑖 and 𝑡 index regions and years, respectively. The outcome
and Ivanovo Oblast in the Central Federal District. None of these regions, due variable, 𝑦𝑖𝑡 , represents one of the following: (i) the annual average
to their geographic location, has ever been subject to the Northern Multiplier overall unemployment rate, (ii) age-specific unemployment rate (with
policy, yet their minimum wages have consistently remained above the federal age groups 15–19, 20–29, and 30+), or (iii) the informality rate,
floor. defined as the proportion of the labor force employed informally.

4
G.K. Ten and S. Wang Economic Modelling 148 (2025) 107086

The primary regressor, log 𝑀𝑊𝑖𝑡 , is the natural logarithm of the geographic location and proximity to water bodies) and common year-
annual average real minimum wage. 𝑋𝑖𝑡′ is a vector of control variables, specific shocks, the remaining variation in the index appears to be
𝜗𝑖 is the region fixed effect, 𝜏𝑡 is the year fixed effect, and 𝑢𝑖𝑡 is the nearly random. Although higher net levels of the Bodman index are
error term. Consistent with prior literature, our baseline specification primarily concentrated in the northern part of Russia, changes within
includes control variables such as the regional Consumer Price Index regions do not seem to depend on the region’s geographic location.
(CPI), the natural logarithm of regional GDP per capita, and the natural Additionally, we examine a list of potential factors that might mediate
logarithm of the working-age population (Burkhauser et al., 2000; Dube the observed relationship between our IV and the unemployment rate:
et al., 2010; Meer and West, 2016). regional health indicators (measured as the population-to-doctor ratio
A potential concern with the OLS estimation of Eq. (1) is reverse
and the number of registered diseases per 1,000 people), net migration
causality. It is unlikely that regional legislators would raise the local
inflow per 10,000 people, average household utility expenses, labor
minimum wage if the local unemployment rate is already high. To
force participation rate, agricultural productivity index (measured as
address this concern, we employ an instrumental variable for our key
agricultural output over GDP), and electricity production (measured in
regressor, log 𝑀𝑊𝑖𝑡 . Our primary instrument is the regional average
Winter Severity Index for the three months preceding each March of billions of kWh). We document that none of the indicators mentioned
a given year. For instance, in the case of a given region observed in the above is correlated with the Bodman index, lending further support to
year 2009, we assign the value of the Winter Severity Index averaged our exclusion restriction assumption.
over December 2008, January 2009, and February 2009. To formally test the exclusion restriction, we introduce an addi-
One of the earliest measurements of winter severity was developed tional IV borrowed from the existing literature: the mean of mini-
by the Swedish scientist Gösta Bodman. In his summary of the research mum wages in geographically adjacent regions (Neumark and Wascher,
expedition to the South Pole (1901–1903), Bodman (1916) quantified 1992; Rybczynski and Sen, 2018). The underlying rationale is that
the subjective human perception of winter severity with the following local legislators are likely influenced by minimum wage policies in
equation, now known as the Bodman index: neighboring regions. While passing the overidentification test does not
𝑆 = (1 − 0.04 × 𝑇 ) × (1 + 0.27 × 𝑉 ), (2) prove that the Bodman index is excluded from Eq. (1), we interpret the
absence of a statistically significant correlation between the IVs and the
where 𝑆 denotes the Bodman index, 𝑇 represents the temperature error term—as is the case with our findings—as additional evidence
measured on the Celsius scale (◦ 𝐶), and 𝑉 is the wind speed in meters supporting the validity of the exclusion restriction assumption.
per second (m/s). Higher values of 𝑆 indicate a worse perception of
winter. The larger coefficient of 𝑉 (0.27) compared to that of 𝑇 (0.04)
aligns with Bodman’s rationale, which emphasizes that wind speed 3.2. A natural experiment in the Kamchatka region
carries greater weight than temperature in determining the human
perception of coldness. This is because wind speed can lead to breathing Our analysis of the regional panel data heavily relies on assumptions
difficulties, skin damage, and mechanical pressure on the human body.
regarding the validity of our IV, and we provide a range of diagnostic
For instance, if the temperature is −25◦ 𝐶 and the wind speed is 1 m/s,
tests to assess their plausibility. To further corroborate our baseline
the Bodman index value is equal to 2.54. This value is nearly equivalent
findings, we follow a well-established tradition of investigating the
to a situation where the temperature is −5◦ 𝐶 and the wind speed is
employment effects of minimum wages in quasi-experimental settings.
4.1 m/s.
The Bodman index is a good candidate for a valid instrumental An ideal natural experiment would supply a pair of regions that
variable for the minimum wage if it affects the level of minimum wages would have had parallel employment trends had one of them not
(relevance) while having no direct impact on the regional unemploy- introduced a higher minimum wage. With this in mind, we pick our
ment rates (exclusion restriction). Thus, we first discuss the first-stage ‘‘treated’’ and ‘‘control’’ regions that satisfy the following criteria:
relationship between the regional Bodman index and the minimum
wage level. We then present our arguments justifying our exclusion 1. The minimum wage increase must be sizable so that possible em-
restriction assumption. ployment effects are more likely to be detectable. This criterion
Fig. 1 shows that the minimum wage consistently exceeds the excludes those cases in which regions simply adjust local wage
federal floor in certain regions. As elaborated in Section 2.1, this floors to inflation.
phenomenon finds its explanation in the presence of the Northern 2. The intervention must occur outside of the first calendar quarter.
Multiplier. In the North and Far East regions, characterized by cold This condition decreases the likelihood of policy anticipation on
winters, workers receive higher minimum wages irrespective of labor the employers’ side, as in most cases, new minimum wages are
market conditions. Consequently, the Northern Multiplier functions as introduced in the first quarter of the year.
a region-specific constant wage multiplier, which is fully absorbed by 3. The initial minimum wage in the ‘‘treated’’ region should follow
region fixed effects. However, the historical existence of this multiplier the federal minimum, as regions with a long tradition of higher
provides a rationale for the first-stage causal relationship between
minimum wages are less likely to have labor market outcomes
the Bodman index and regional minimum wages. A worsening per-
comparable to those that follow the state’s wage floor.
ception of winter severity equips local labor unions with bargaining
power to advocate for higher minimum wages as compensation for
In June 2012, the local minimum wage rose from 4,611 RUB
more challenging working conditions. Indeed, as detailed in the results
(Federal Minimum Wage) to 7,500 RUB in the Kamchatka region.
section, consistent with our conjecture, our data reveal a strong and
As shown in Fig. 1, the timing and magnitude of this increase are
positive first-stage relationship between the Bodman index and regional
minimum wages. highly unusual. First, most other regions typically update their local
We now argue that the Bodman index is unlikely to have a direct minimum wages during the first calendar quarter. Second, prior to June
impact on local unemployment rates. Fig. 2 displays the distribution 2012, Kamchatka had consistently aligned its minimum wage with the
of the OLS residuals of the Bodman index in 2012, the most recent federal floor. The steep mid-year deviation from the federal minimum,
cross-section in our dataset. These residuals are obtained after re- as depicted in Fig. 1, is nearly unprecedented. These factors make it
gressing the Bodman index on region and year fixed effects using all plausible that Kamchatkan employers did not anticipate this policy
regions observed from 2009 to 2012. The resulting map suggests that, change. Chukotka, which is the only neighboring region of Kamchatka,
once we account for regional time-invariant characteristics (such as supplies a natural comparison unit, as it is geographically adjacent

5
G.K. Ten and S. Wang Economic Modelling 148 (2025) 107086

Fig. 2. Geography of Bodman index.

Notes: The figure plots the residuals (in year 2012) resulting from regressing the Bodman index on region and year fixed effects. Darker zones indicate higher (residual) winter
severity.

and strictly follows the state’s wage floor throughout the entire study groups of workers: young (𝑌 ) and older (𝑂). We use 30 as the threshold
period.6 age—i.e., workers younger than 30 are categorized as young, whereas
It is natural to wonder about the possible forces driving the unusual older workers are viewed as the older workforce. We consider the
minimum wage increase in Kamchatka. Based on official statistics following econometric models for the two age groups:
underlying our data, it is unlikely that this new wage floor was driven
by favorable macroeconomic conditions. The increase occurred in June 𝑦𝑌 = 𝛽1𝑌 𝑇 + 𝛽2𝑌 𝑃 𝑜𝑠𝑡 + 𝛽3𝑌 𝑇 × 𝑃 𝑜𝑠𝑡 + 𝑣𝑌 , (3)
2012. However, in 2011, Kamchatka ranked 63rd out of 83 regions in
terms of real GDP growth. In the same year, the region ranked 58th 𝑦𝑂 = 𝛽1𝑂 𝑇 + 𝛽2𝑂 𝑃 𝑜𝑠𝑡 + 𝛽3𝑂 𝑇 × 𝑃 𝑜𝑠𝑡 + 𝑣𝑂 , (4)
for unemployment and 49th for the consumer price index. Thus, the
large minimum wage increase was not preceded by booming economic where 𝑦 is one of two binary outcomes: (i) unemployment, which is
growth, exceptionally low unemployment, or a rapid rise in prices. equal to one if the individual is unemployed (0 for the employed), and
Moreover, regions typically update the local minimum wage in the first (ii) informal employment, which is equal to one if the individual is
quarter of the year to account for local macroeconomic factors, such as employed informally (0 for the formally employed). 𝑇 is an indicator
rising consumer prices. This was not the case for Kamchatka in 2012, of the treated region (Kamchatka), and 𝑃 𝑜𝑠𝑡 is a dummy indicating the
where the new wage floor occurred in June. post-treatment period (June 2012 and onward). Equations (3) and (4)
One probable force behind the minimum wage increase might be are canonical difference-in-differences specifications. If the OLS esti-
the unexpected appointment of a new governor in Kamchatka in Febru- mate of 𝛽3𝑌 and 𝛽3𝑂 capture the policy impact, then the existence of the
ary 2011, Vladimir Ilyukhin. His appointment followed the voluntary difference in unemployment responses to minimum wages across the
resignation of the former governor, Alexey Kuzmitsky, who had gov- two age groups implies the rejection of the following null hypothesis:
erned the region from July 2007 until February 2011, nearly a year
less than the typical five-year term. We were unable to retrieve the 𝐻0 = 𝛽3𝑌 − 𝛽3𝑂 = 0 (5)
official reason for the former governor’s resignation from the Kremlin’s
website archives. However, the most important concern is whether We test the above hypothesis by combining Equations (3) and (4)
the appointment of the new regional administration in early 2011 as follows:
led to employers adjusting their workforce in anticipation of the new 𝑦 =𝛿1 𝑇 + 𝛿2 𝑃 𝑜𝑠𝑡 + 𝛿3 𝑌 𝑜𝑢𝑛𝑔 + 𝛿4 𝑇 × 𝑃 𝑜𝑠𝑡 + 𝛿5 𝑇 × 𝑌 𝑜𝑢𝑛𝑔 +
minimum wage policy—or some other policy affecting the local labor (6)
𝛿6 𝑌 𝑜𝑢𝑛𝑔 × 𝑃 𝑜𝑠𝑡 + 𝛿7 𝑇 × 𝑃 𝑜𝑠𝑡 × 𝑌 𝑜𝑢𝑛𝑔 + 𝑣,
market. Fortunately, this can be tested with our data: as we show in the
results section, neither unemployment nor informality rates responded where the new binary variable 𝑌 𝑜𝑢𝑛𝑔 takes on the value of 1 if the
to the ‘‘placebo’’ minimum wage assigned to 2011, the year of the new individual’s age belongs to the closed interval [15, 29], and it is equal to
governor’s appointment. This finding lends support to the possibility 0 otherwise. The parameter of interest in Eq. (6) is 𝛿7 , which is known
that there was no precautionary workforce adjustment. as the difference-in-difference-in-differences estimator. Testing that 𝛿7
We utilize a subsample of workers from the two regions mentioned in Eq. (6) is equal to zero is algebraically identical to testing the null
above that is available in RLFS 2010–2013. Considering the past ob- hypothesis given by Eq. (5). Under the assumption that both regions
servations that the unemployment response to minimum wages varies would have followed parallel unemployment trends in the absence of
by workforce age group, we divide our sample into the following two the new minimum wage’s introduction, 𝛿7 from Eq. (6) identifies the
policy impact on the workforce aged below 30.
6
We also consider a variant of Eq. (6) with a rich set of individual
It is worth noting that in September of the same year, Kamchatka raised
controls (age, age squared, marital status dummy, gender, educational
its minimum wage by 40%. In an auxiliary investigation, we tested whether
this second minimum wage hike brought different employment effects but
attainment, urban area dummy, number of children under 18 years
failed to find such evidence. These findings suggest that it is the first—the in the household), region-quarter fixed effects, and the region-specific
June 2012—minimum wage increase that brought employment effects. These quartic month-year trend. Later in the text, we show that the estimated
auxiliary results are available upon request from the authors. impact is not sensitive to the inclusion of the proposed set of controls.

6
G.K. Ten and S. Wang Economic Modelling 148 (2025) 107086

3.3. Industry-specific employment responses to the minimum wage policy Table 2

Unemployment effects of the minimum wages in Russia.
(1) (2) (3) (4) (5)
Given that the variation of estimated employment effects of the min-
Overall 15–19 20–29 30+ Informality
imum wage in the literature, it is natural to consider the characteristics
Panel A: Second stage
of a given labor market that policymakers should pay close attention Log (MW) 0.001 0.040+ 0.053** 0.007 0.124**
to when introducing a new wage floor. In this paper, we argue that (0.012) (0.022) (0.012) (0.012) (0.016)
the substitutability of capital for labor — which likely varies across Panel B: First stage for Log (MW)
countries and within a given country across industries — is one labor- Bodman index 0.172**
market characteristic that helps predict whether raising the minimum (0.020)
wage will result in layoffs. This insight comes from the theoretical 𝑁 321
works by Allen (1938) and Hamermesh (1993), which show that la- Notes: The table reports the unemployment effects of the minimum wage estimated by
bor demand’s wage elasticity depends upon the substitution elasticity 2SLS. 𝑁 = 321. The unit of observation is region-year. The controls comprise the log of
between capital and labor. More specifically, assuming that (i) a given real GDP per capita, CPI, the log of the working-age population, region fixed effects, and
year fixed effects. Robust standard errors clustered by region are shown in parentheses;
industry employs capital (𝐾) and labor (𝐿) to produce output (𝑌 ); (ii)
+, *, and ** denote significance at the 10%, 5%, and 1% levels, respectively.
the inputs are transformed into the output by a constant elasticity of
substitution production function; and (iii) in equilibrium, the market
for the final good clears, Allen (1938) shows that the wage elasticity of
𝐿 ) can be expressed as follows: which are identical to those employed in Eq. (1).8 𝜅𝑗×𝑖 and 𝜃𝑡 denote
labor demand (𝜖𝑤
the region×industry and year fixed effects, respectively, and 𝑢𝑗𝑖𝑡 is the
𝐿 𝑤𝐿 𝑌 𝑟𝐾 error term that we cluster at the region level. We estimate Equation
𝜖𝑤 = 𝜖 − 𝜎, (7)
𝑌𝑝 𝑝 𝑌𝑝 (8) by 2SLS using the Bodman index and its interaction with 𝜓𝑗 as
where 𝑤 and 𝑟 denote the price of labor and capital, respectively; 𝑝 is instruments for the minimum wage variable and its interaction with
the market price of the good 𝑌 ; 𝜖𝑝𝑌 is the price elasticity of demand; 𝜓𝑗 , respectively. Our focus is on the estimated value of 𝛼 + 𝜆𝑗 , which
and 𝜎 is the elasticity of substitution between capital and labor. informs about the elasticity of youth/informal employment with respect
As can be seen from Eq. (7), the magnitude of employment loss to the regional minimum wage in a given industry.
triggered by raising wages depends upon two parameters. First, a Next, we estimate the industry-specific substitutability between cap-
higher price elasticity of demand for the final good (𝜖𝑝𝑌 ) results in a ital and labor in the following two ways. The first approach is to use log
larger loss of workers. Second, the substitutability between capital and capital investments as the outcome in Eq. (8); in this case, the estimated
labor — characterized by the value of 𝜎 — matters. As the elasticity 𝛼 + 𝜆𝑗 informs about a given industry’s ability to substitute capital for
of substitution converges to zero (𝜎 → 0), the production function ap- labor. For example, a negative value 𝛼 + 𝜆𝑗 would imply that employers
proaches the case of perfect complements, in which the cost-minimizing in the industry in question reduce capital investments in response to the
demand for labor and capital is invariant to their relative prices. In the minimum wage increase, which likely indicates the complementarity
opposite case (i.e., 𝜎 → ∞), the production function converges to the between capital and labor. On the other hand, a positive value would
case of perfect substitutes, in which demand for labor and capital is indicate that capital and labor can substitute for each other in a given
very sensitive to their relative prices. industry.
Another approach we adopt is to follow Arrow et al. (1961), who
Inspired by this insight from Allen (1938), we show empirically
use the joint variation of the value-added per worker and the real wages
that employment losses induced by higher minimum wages are indeed
to identify the substitution elasticity in each industry. Specifically, we
concentrated in industries capable of substituting capital. To do so, we
fit the following model:
first estimate the industry-specific employment response to minimum
wage increases by fitting the following equation: log 𝑉𝑖𝑡𝑗 = 𝜎 𝑗 log 𝑊 𝐴𝐺𝐸𝑖𝑡𝑗 + 𝑋𝑖𝑡′ 𝜌 + 𝜁𝑡 + 𝑡𝑐𝑖 + 𝜔𝑗𝑖𝑡 (9)
∑
𝐽
The outcome variable, log 𝑉𝑖𝑡𝑗 ,
is the natural log of value added per
log 𝑌𝑗𝑖𝑡 = 𝛼 log 𝑀𝑊𝑖𝑡 + 𝜆𝑗 log 𝑀𝑊𝑖𝑡 × 𝜓𝑗 + 𝑋𝑖𝑡′ 𝜂 + 𝜅𝑗×𝑖 + 𝜃𝑡 + 𝑢𝑗𝑖𝑡 , (8)
𝑗=1
worker in industry 𝑗, region 𝑖, and year 𝑡.
After obtaining a set of estimates for industry-specific (i) employ-
where 𝑗, 𝑖, and 𝑡 index industry, region, and year, respectively.7 The ment responses from Eq. (8) and (ii) elasticities of substitution from Eq.
outcome variables (log 𝑌𝑗𝑖𝑡 ) are the following: (i) the natural log of (9), we assess whether the two sets of point estimates are negatively
the number of workers aged below 30; and (ii) the natural log of correlated.
the number of workers employed informally. When constructing the
employment outcome variables, we proceed as follows. First, we obtain 4. Unemployment effects of the Russian minimum wage policy:
the total labor force size from the Russian Federal State Statistics Results
Service. Next, using the labor force survey described in Section 2.2,
we calculate the share of the labor force aged below 30 and the share 4.1. Evidence from the Russian regions
of informally employed workers in each industry×region×year cell.
Finally, with the two pieces of information in hand, we estimate the Table 2 presents the instrumental variable estimates of Eq. (1).
total number of workers aged below 30 and those employed informally. Panel A displays the second-stage results, while Panel B shows the
The main regressors are the natural log of the regional minimum first-stage regression of the log real minimum wage on the excluded
wage log 𝑀𝑊𝑖𝑡 and its interaction with the indicator of industry 𝑗 instrument, the Bodman index, and covariates. We begin by discussing
(log 𝑀𝑊𝑖𝑡 ×𝜓𝑗 ); the term 𝑋𝑖𝑡′ is a vector of region-level control variables the results in Panel A.
Column (1) of Panel A confirms the findings of recent literature,
indicating that the estimated minimum wage coefficient is statistically
7
In our analysis, we consider the following industries: (i) Agriculture, insignificant and numerically close to zero in the overall unemployment
Forestry, and Fishing; (ii) Mining; (iii) Building and Construction; (iv) Man- regression. Columns (2) and (3) suggest that the young workforce is
ufacturing; (v) Transportation and Communication; (vi) Electricity, Gas, and
Water; (vii) Wholesale and Retail; (viii) Finance and Insurance; (ix) Real
8
Estate; (x) Education; (xi) Health; (xii) Hotels and Restaurants; and (xiii) Other We control for the regional CPI, the natural log of the regional GDP per
Services. capita, and the natural log of the working-age population.

7
G.K. Ten and S. Wang Economic Modelling 148 (2025) 107086

particularly vulnerable, as a one percent increase in the real minimum results for unemployment, and columns (4)–(6) report the results for
wage results in a 0.04 and 0.053 percentage point increase in the informal employment.
unemployment rates among those aged 15–19 and 20–29, respectively. Notably, the point estimate of the triple interaction term (𝑇 ×
Column (4) implies that the older workforce — those aged 30 years 𝑃 𝑜𝑠𝑡 × 𝑌 𝑜𝑢𝑛𝑔) remains robust to the inclusion of controls. For instance,
and above — is unlikely to be affected by the minimum wage policy. columns (1)–(3) suggest that the minimum wage hike increased the
Column (5) shows an increase in the informality rate in response to the probability of unemployment by approximately 3 percentage points
minimum wage hike, suggesting that the minimum wage policy binds for individuals under the age of 30. Columns (4)–(6) show that the
employers. Faced with higher labor costs, companies may hire workers probability of informal employment among young workers increased
off-the-books to replace formal workers. by 4 percentage points following the minimum wage increase.
As discussed in the methodology section, we can identify the impact In summary, the results in Table 5 indicate that the introduction
of interest if the Bodman index affects the regional minimum wage of the minimum wage policy is associated with a 0.03 (0.05) increase
while being excluded from the unemployment equation. As Panel B in the probability of being unemployed (employed informally) among
shows, the Bodman index positively affects the regional minimum young workers. The coefficients of the 𝑇 × 𝑃 𝑜𝑠𝑡 term in columns
wage, and the impact is precisely estimated. Therefore, we must justify (1)–(3) suggest that the minimum wage policy is unlikely to affect
the assumption that the minimum wage is the only channel through unemployment among older workers, consistent with prior research
which our proposed instrument affects unemployment. and our previous findings in this section. However, columns (4)–(6)
Our first strategy is to check whether our instrument affects unem- suggest that informality responds positively to the minimum wage even
ployment through other channels, such as health, migration, agricul- among more experienced employees.
tural productivity, energy costs, and labor force participation. While The validity of the results reported in Table 5 relies on the as-
this list is largely speculative, we believe it includes the most plausible sumption that employment trends in both regions would have followed
pathways through which the Bodman index may affect regional unem- parallel paths in the absence of the policy intervention. To test this
ployment. For example, one might argue that a worse perception of assumption, we exploit the fact that both regions adhered to the state’s
winter severity correlates with workers’ health, influencing their deci- minimum wage floor in 2010 and 2011. We define a placebo post-
sion to participate in the labor force. There is also the possibility that treatment variable, which is set to 1 for the year 2011 and 0 for
workers prefer locations with milder winter conditions. Additionally, 2010. We then substitute the term 𝑃 𝑜𝑠𝑡 with this placebo treatment
fluctuations in the Bodman index may affect agricultural output and in Eq. (6) and estimate the resulting model, with the results reported
electricity costs. Finally, economic activity could be related to winter in Table 6. The findings suggest that the two regions do not exhibit
severity for reasons unrelated to workers’ health, such as productivity statistically significant differences in unemployment trends during the
losses or caregiving for sick dependents (Han et al., 2023). pre-treatment period. This result further supports the validity of the
We therefore consider the following regional characteristics: (i) proposed quasi-experimental design.
log population-to-doctors ratio, (ii) log number of registered diseases Our findings from the Kamchatka region are largely consistent with
per 1,000 people, (iii) net migration inflow (per 10,000 people), (iv) the benchmark result presented in Section 4.1. The regional panel data
agricultural output normalized by regional GDP, (v) log of electricity estimate suggests that a one-percentage-point increase in the minimum
production (billions of kWh), (vi) average household utility expenses wage raises the local unemployment rate among workers aged below 30
normalized by average household income, and (vii) labor force partici- by 0.05 percentage points. The Kamchatka case study indicates that a
pation rate. We check whether these regional characteristics respond to 63% increase in Kamchatka’s minimum wage led to a three-percentage-
changes in the Bodman index. Table B1 in Online Appendix B presents point higher probability of unemployment among workers aged below
the results, showing that none of the variables respond to changes in 30. In other words, a one-percent increase in Kamchatka’s minimum
the Bodman index. This exercise largely excludes the possibility that the wage raised the share of unemployed youth by 0.05 percentage points,
Bodman index affects the regional unemployment rate through health, which is consistent with the baseline estimate from the regional panel
migration, agricultural productivity, electricity costs, or the regional data.
labor force participation rate. This strengthens the assumption that
the regional minimum wage is the only channel through which our 4.3. Industry-specific employment response to the minimum wage policy
instrument affects the local unemployment rate.
Our second approach to assessing the validity of our instrument Columns (1) and (2) of Table 7 present estimates of industry-
is a direct statistical test of the exclusion restriction, which requires specific responses of the number of young workers and the number of
at least two instruments for the minimum wage variable. We use an workers employed informally, respectively, to changes in the minimum
additional instrumental variable, the mean of the minimum wages in wage. The results indicate that, in most industries, an increase in the
geographically bordering regions. Table 3 presents the results. Panel minimum wage leads to a reduction in the number of workers aged
A shows that the inclusion of the additional instrument does not below 30 and an increase in informal employment, with the exceptions
significantly alter the baseline point estimates. Panel B displays the of Education and Health. The substitutability between formally and
first-stage regression result. The coefficient of the additional instru- informally employed workers becomes more apparent when we plot
ment, 𝐿𝑜𝑔(𝑎𝑣𝑒𝑟𝑎𝑔𝑒 𝑀𝑊 𝑎𝑚𝑜𝑛𝑔 𝑛𝑒𝑖𝑔ℎ𝑏𝑜𝑟𝑠), is positive and statistically the coefficients from column (1) of Table 7 against those from column
significant, confirming the spillover effect from neighboring regions. (2) in Fig. 3. Industries that displace fewer young workers in response
Moreover, the 𝐹 statistic of the excluded instruments supports the to the minimum wage tend to hire more workers informally.
strength of the first-stage relationship. The last row of Table 3 reports Column (3) of Table 7 presents the estimated response of capital
the 𝑝-values of the Sargan-Hansen 𝐽 test. None of the 𝑝-values fall investments to changes in the minimum wage, while Column (4) reports
below the 10% significance level, further supporting our belief that the the estimated elasticity of substitution between capital and labor for
Bodman index is excluded from Eq. (1). each sector. Due to the lack of wage data for the financial sector,
the corresponding cell is left empty. We then link the industry-specific
4.2. Unemployment response in Kamchatka employment responses to minimum wages (as reported in column 1 of
Table 7) to the results in columns 3 and 4, and illustrate them in the
In this subsection, we present the findings from the natural experi- two panels of Fig. 4.
ment in the Kamchatka region. Table 4 provides summary statistics for Panel (a) of Fig. 4 plots the point estimates from column (1) against
all variables used in the analysis, while Table 5 presents the regression those from column (4) of Table 7. The resulting downward-sloping
results both with and without controls. Columns (1)–(3) report the line illustrates the estimated cross-industry average employment loss

8
G.K. Ten and S. Wang Economic Modelling 148 (2025) 107086

Table 3
Testing exclusion restriction.
(1) (2) (3) (4) (5)
Overall 15–19 20–29 30+ Informality
Panel A: Second stage
Log (MW) −0.000 0.039+ 0.052** 0.007 0.126**
(0.012) (0.021) (0.012) (0.012) (0.016)
Panel B: First stage for Log (MW)
Bodman index 0.175**
(0.021)
Log (average MW among neighbors) 0.192+
(0.104)
𝑁 321
Sanderson-Windmeijer 𝐹 -test 34.50
𝑝-value of Sargan-Hansen 𝐽 -test 0.35 0.72 0.46 0.78 0.17

Notes: The table reports the unemployment effects of the minimum wage estimated by 2SLS. The unit of observation
is region-year. Controls comprise the log of real GDP per capita, CPI, the log of the working-age population, region
fixed effects, and year fixed effects. Robust standard errors clustered by region are reported in parentheses; +, *,
and ** denote significance at the 10%, 5%, and 1% levels, respectively.

Table 4
Summary statistics for Kamchatka case study.
(1) (2)
Mean
(Standard deviation)
Panel A: Outcome variables
Unemployed 0.061
(0.240)
Informally employed 0.117
(0.321)
Panel B: Control variables
Grad School 0.002 0.002
(0.047) (0.047)
College 0.268 0.278
(0.443) (0.448)
Mid professional 0.237 0.244
(0.425) (0.430)
Primary professional 0.195 0.196
(0.396) (0.397)
High School 0.234 0.222
(0.423) (0.416)
Middle School 0.059 0.052 Fig. 3. Substitution between young and informally employed workers.
(0.236) (0.223) Notes: The unit of observation is industry. The 𝑦-axis (and 𝑥-axis) shows the estimated
Male 0.497 0.494 effect of the minimum wage on the number of workers employed informally (workers
(0.500) (0.500) aged below 30).
Youth 0.205 0.194
(0.404) (0.395)
Age 41.44 41.76
(12.33) (12.19) investment responses to minimum wage increases (Hau et al., 2020;
Number of children aged < 18 years old 0.571 0.571 Baek et al., 2021). Second, our results indicate that informal recruit-
(0.829) (0.825) ment is another means of adjusting to higher labor costs: Employers
Married 0.625 0.639 are more likely to hire workers informally when facing an increased
(0.484) (0.480)
Urban 0.588 0.601
minimum wage.
(0.492) (0.490) These results raise questions about the effectiveness of the Russian
𝑁 36,400 34,170
minimum wage policy as a tool for alleviating poverty. Job losses
among young workers and the informalization of the labor market due
Notes: This table presents the means and standard deviations of key variables used in
to minimum wage increases may limit the policy’s ability to effectively
the case study of the Kamchatka region.
assist the poor.

4.4. Who are informally employed?

moderated by the elasticity of substitution between capital and labor.
As shown in the figure, a higher elasticity of substitution is associated A substantial portion of our analysis is dedicated to investigating
with a larger (in magnitude) employment loss in a given industry, the implications of the Russian minimum wage policy for informal
consistent with theoretical predictions. Panel (b) of Fig. 4 plots the employment. The key feature of informal employment relevant to our
point estimates from column (1) against those from column (3) of analysis is its invisibility to legislators, meaning the minimum wage
Table 7. The pattern is as follows: industries that are more capable does not bind for such workers. In this section, we provide descriptive
of increasing capital investments in response to the minimum wage characteristics of informally employed workers using the SIPP data
increase tend to reduce their demand for young workers, consistent described in Section 2.9 We first examine the proportion of individuals
with the capital-labor substitution hypothesis discussed in Section 3.3.
The patterns in Figs. 5 and 6 suggest that industries have at least
two ways to adapt to the minimum wage shock. First, if the produc- 9
Although the Labor Force Survey might seem to be a natural candidate
tion technology allows for substituting capital for labor, the industry for such an investigation, it does not contain information on earnings, which
responds accordingly. This finding aligns with the literature on capital is why we chose the SIPP data for this part of the analysis.

9
G.K. Ten and S. Wang Economic Modelling 148 (2025) 107086

Table 5
Employment effects in Kamchatka.
(1) (2) (3) (4) (5) (6)
Unemployed Informally employed
Post −0.011* −0.014** −0.011 0.006 0.004 −0.004
(0.004) (0.004) (0.011) (0.007) (0.007) (0.018)
Treated 0.015** 0.016** 0.051** 0.070** 0.072** −0.006
(0.005) (0.005) (0.018) (0.008) (0.008) (0.028)
Post × Treated −0.002 0.002 0.011 0.057** 0.061** 0.063*
(0.007) (0.007) (0.016) (0.012) (0.012) (0.030)
Youth 0.052** −0.038** −0.038** 0.017+ −0.025* −0.024*
(0.007) (0.010) (0.010) (0.009) (0.012) (0.012)
Post × Youth −0.011 −0.013 −0.013 −0.025+ −0.027* −0.027*
(0.009) (0.009) (0.009) (0.012) (0.012) (0.012)
Treated × Youth 0.015 0.029** 0.029** 0.015 0.029* 0.028*
(0.011) (0.010) (0.010) (0.013) (0.013) (0.013)
Post ×Treated × Youth 0.032* 0.029* 0.028* 0.042* 0.040* 0.040*
(0.014) (0.013) (0.013) (0.018) (0.018) (0.018)
Controls NO YES YES NO YES YES
Region-Quarters FE NO NO YES NO NO YES
Flexible Trends NO NO YES NO NO YES
𝑁 36,400 34,170

Notes: The outcome variables are: the unemployment dummy (equal to 1 if unemployed, 0 if employed) in Columns
(1)–(3); and the informal employment dummy (equal to 1 if employed informally, 0 if employed formally) in Columns
(4)–(6). Controls comprise age, age squared, marital status (dummy), educational attainment (a set of five dummies),
male (dummy), urban residence (dummy), and the number of children under 18 years. ‘‘Flexible Trends’’ refers to
region-specific quartic time (month of the year) trends. Robust standard errors, clustered by 𝑟𝑒𝑔𝑖𝑜𝑛 × 𝑚𝑜𝑛𝑡ℎ, are in
parentheses; +, *, and ** denote significance at the 10%, 5%, and 1% levels, respectively.

Table 6
Placebo test.
(1) (2) (3) (4) (5) (6)
Unemployed Informally employed
Placebo ×Treated ×Youth −0.009 −0.018 −0.020 0.017 0.012 0.013
(0.023) (0.022) (0.022) (0.028) (0.028) (0.029)
Controls NO YES YES NO YES YES
Region-Quarters FE NO NO YES NO NO YES
Flexible Trends NO NO YES NO NO YES
𝑁 18,274 17,078

Notes: The outcome variables are: unemployment dummy (equal to 1 if unemployed, 0 if employed) in Columns
(1)–(3); and informal employment dummy (equal to 1 if employed informally, 0 if employed formally) in Columns
(4)–(6). The ‘‘Placebo’’ dummy is assigned to the period before 2012, i.e., prior to the year of the new minimum
wage’s introduction in Kamchatka. Specifically, it is equal to 1 for the year 2011 and 0 for the year 2010. Controls
comprise age, age squared, marital status (dummy), educational attainment (a set of five dummies), male (dummy),
urban residence (dummy), and number of children under 18 years. ‘‘Flexible Trends’’ refers to region-specific quartic
time (month of the year) trends. Robust standard errors, clustered by 𝑟𝑒𝑔𝑖𝑜𝑛 × 𝑚𝑜𝑛𝑡ℎ, are in parentheses; +, *, and
** denote significance at the 10%, 5%, and 1% levels, respectively.

employed in the informal sector across various age and education control for region-year fixed effects, 𝜃𝑟×𝑡 , absorbing region-level shocks,
groups, and then explore the association between informal employment including the minimum wage. 𝑢𝑖𝑟𝑡 is the error term clustered at the
status and annual earned income. region level. We fit Equation (10) using the sampling weights available
To study the share of informal workers in various age and education in the SIPP data.
groups, we fit the following specification: The estimates of 𝛼𝑎 (𝜂𝑒 ) capture the change in the proportion of
individuals employed informally relative to the omitted group, which
∑
9 ∑
5
comprises individuals aged 15 to 19 years (without any secondary
′
𝐼𝑛𝑓 𝑜𝑟𝑚𝑎𝑙𝑖𝑟𝑡 = 𝛼𝑎 𝑎𝑔𝑒 𝑔𝑟𝑜𝑢𝑝𝑖𝑟𝑡 + 𝜂𝑒 𝑒𝑑𝑢𝑐𝑖𝑟𝑡 + 𝑋𝑖𝑟𝑡 𝛽 + 𝜃𝑟×𝑡 + 𝑢𝑖𝑟𝑡 . (10)
education). We report two sets of parameter estimates: (i) Unadjusted,
𝑎=1 𝑒=1
The outcome variable is an indicator of informal employment for obtained without any controls, and (ii) those obtained after including
individual 𝑖 observed in region 𝑟 in year 𝑡 (1 for employed informally the full set of controls.
Fig. 5 presents the results by age group. We observe that, relative to
and 0 for formally employed). The term 𝑎𝑔𝑒 𝑔𝑟𝑜𝑢𝑝𝑖𝑟𝑡 is an indicator
individuals aged 15 to 19 years, each older age group has a lower pro-
for one of nine age groups: [20, 25), [25, 30), . . . , [55, 60), and 60+.10
portion of informally employed individuals. This result holds regardless
The omitted group consists of workers aged 15 to 19 years. Similarly,
of the deployment of controls. Moreover, an overall negative trend is
𝑒𝑑𝑢𝑐𝑖𝑟𝑡 is an indicator for one of five education groups: college (or
evident as we move to older worker groups, indicating that, conditional
higher), incomplete college, secondary professional, secondary general, on having a job, the likelihood of being employed informally decreases
and incomplete secondary. The omitted group comprises individuals with age.
without any secondary education. We also control for a vector of Fig. 6 displays the proportion of employed individuals with informal
′ , including marital status, urban-area dummy, gender,
covariates, 𝑋𝑖𝑟𝑡 jobs by education group. It is clear that higher educational attainment
and an indicator for having at least one child under 18. Additionally, we is negatively associated with the likelihood of having an informal job
(as opposed to formal employment). The point estimates show little
difference between the unadjusted and fully saturated models.
10
The notation [𝑎, 𝑏) represents a half-open interval, indicating a range of In summary, informally employed workers are, on average, (i)
values from 𝑎 (inclusive) to 𝑏 (exclusive). younger and (ii) less educated compared to their formally employed

10
G.K. Ten and S. Wang Economic Modelling 148 (2025) 107086

Table 7
Implications of the minimum wages for different industries.
Log (Youth) Log (Informal) Log (Investments) 𝜎̂
(1) (2) (3) (4)
Finance 0.521* −1.101* 0.148**
(0.249) (0.457) (0.050)
Agriculture & Fishery −1.212*** 0.155 −0.170** 0.309*
(0.287) (0.446) (0.064) (0.133)
Mining −1.027 0.332 −0.044 0.717*
(0.675) (0.809) (0.071) (0.284)
Manufacturing −0.744** 1.072* −0.170** 0.574***
(0.232) (0.433) (0.054) (0.149)
Electricity, Gas & Water −0.724* 0.375 −0.124* 0.233
(0.297) (0.762) (0.051) (0.397)
Building and Construction −0.525* 1.000* −0.229*** 0.387***
(0.267) (0.441) (0.066) (0.101)
Wholesale and Retail −0.613* 0.981* −0.158** 0.094
(0.240) (0.418) (0.056) (0.088)
Hotels & Restaurants −0.186 2.453*** −0.215* −0.093
(0.343) (0.533) (0.096) (0.257)
Transportation & Communication −0.597** 0.861* −0.181*** −0.030
(0.219) (0.386) (0.054) (0.270)
Real Estate −0.490+ 1.372* −0.187*** 0.328
(0.279) (0.617) (0.051) (0.224)
Education −0.863*** −0.126 −0.010 0.274
(0.238) (0.489) (0.050) (0.193)
Health −0.844*** −0.278 −0.052 0.723***
(0.240) (0.558) (0.055) (0.120)
Other −0.615** 1.279** −0.247*** 0.367*
(0.230) (0.435) (0.056) (0.177)
𝑁 3,131 2,797 2,955

Notes: Every regression controls for the log of real GDP per capita, CPI, the log of the working-age population, region fixed
effects, and year fixed effects. Robust standard errors clustered by region are in parentheses; +, *, and ** denote significance
at the 10%, 5%, and 1% levels, respectively.

counterparts. We now proceed with a descriptive examination of the Table 8

Implications of informal employment for earned income.
implications of informal employment for annual earnings by fitting the
(1) (2)
following model:
Informal employment −0.568** −0.337**
′ (0.046) (0.011)
𝐿𝑜𝑔(𝐴𝑛𝑛𝑢𝑎𝑙 𝑒𝑎𝑟𝑛𝑖𝑛𝑔𝑠𝑖𝑟𝑡 ) = 𝜏𝐼𝑛𝑓 𝑜𝑟𝑚𝑎𝑙𝑖𝑟𝑡 + 𝑋𝑖𝑟𝑡 𝜙 + 𝜗𝑟×𝑡 + 𝑣𝑖𝑟𝑡 . (11)
Controls NO YES
𝑁 308,341 308,341
The outcome variable is the natural log of annual earned income
Notes: This table reports the difference in the natural log of annual earned income
for individual 𝑖 observed in region 𝑟 in year 𝑡. The main regressor is between informally and formally employed workers. Controls comprise age group fixed
an indicator of informal employment, 𝐼𝑛𝑓 𝑜𝑟𝑚𝑎𝑙𝑖𝑟𝑡 (1 for informal, 0 for effects, educational attainment fixed effects, marital status, an urban-area dummy,
formal). We control for a vector of covariates, 𝑋𝑖𝑟𝑡 ′ , which comprises gender, an indicator of having at least one child aged below 18, and region-year fixed
effects. Robust standard errors clustered by region are in parentheses; +, *, and **
fixed effects for each age and education group as in Eq. (11), as well denote significance at the 10%, 5%, and 1% levels, respectively.
as marital status, an urban-area dummy, gender, and an indicator for
having at least one child under 18. We also control for region-year
fixed effects, 𝜃𝑟×𝑡 , thereby absorbing region-level shocks and deflating 5. Does the Russian minimum wage policy help the poor?
for the annual income variable. 𝑣𝑖𝑟𝑡 is the error term clustered at the
region level. We use sampling weights when running the regression. The minimum-wage policy is designed to assist individuals at the
The estimate of 𝜏 represents the difference in real annual earned income lower end of the income distribution. However, according to our find-
between informally and formally employed workers. ings, in Russia, this policy leads to job losses among workers under
Table 8 reports the results. Column (1) shows the simple regression 30 years of age and an increase in the proportion of workers employed
result, while column (2) presents the estimate obtained after including informally. Given these findings, it is natural to question whether the
the full set of controls. Results in both columns suggest that having an policy meaningfully aids the poorest.
informal job (versus formal employment) is associated with a reduction To examine the implications of the minimum wage policy for family
income, we adopt an approach deployed in Dube (2019). Specifically,
in annual earned income of approximately 28.6% (= [𝑒−0.337 − 1] × 100)
Dube estimates the household income effects of the policy by fitting the
to 43.3% (= [𝑒−0.568 − 1] × 100).
following specification:
Taken together, the results from this subsection paint the following
picture: compared to formally employed workers, those with informal ∑
3
𝐼𝑐𝑖𝑟𝑡 = 𝛼𝑐𝑘 log(𝑀𝑊𝑟(𝑡−𝑘) ) + 𝑋𝑖𝑟𝑡 𝛹𝑐 + 𝛩 + 𝛾𝑟 𝑡 + 𝜖𝑐𝑖𝑟𝑡 (12)
jobs are younger, less educated, and earn lower income. These findings 𝑘=−1
suggest that informal employment represents a form of precarious The outcome variable is an indicator of the 𝑖th individual observed
work, providing a buffer against full unemployment for individuals in region 𝑟 at time 𝑡 in a family whose value of equalized income
with sufficiently low reservation wages, provided that local labor de- divided by the federal poverty threshold falls below 𝑐. For example, the
mand supports it. Therefore, it is perhaps unsurprising that the increase observation with 𝐼1𝑖𝑟𝑡 = 1 indicates that the individual’s family is below
in unemployment due to the minimum wage observed in our earlier the official poverty threshold. Similarly, 𝐼0.5𝑖𝑟𝑡 = 1 indicates that the
analysis is accompanied by a rise in informal employment. equalized income of the 𝑖th individual’s household is below 0.5 times

11
G.K. Ten and S. Wang Economic Modelling 148 (2025) 107086

Fig. 4. Capital-labor substitution.

Notes: In both panels, the unit of observation is industry, and the 𝑦-axes show the estimated effect of the minimum wage on the number of workers aged below 30. In Panel (a),
the 𝑥-axis shows the estimated elasticity of substitution, while in Panel (b), the 𝑥-axis shows the estimated effect of the minimum wage on capital investments.

the official poverty rate. Following the baseline analysis in Dube’s work, are unavailable for the recessionary years. Finally, our dataset lacks
we consider the values of 𝑐 between 0.5 and 1.75 in increments of 0.25. information on the ethnicity of respondents. Besides the exclusion of
In Dube (2019), the set of key variables of interest includes con- the above information from the set of control variables, our set of
temporaneous, one-year leading, and three-year lagged values of the controls is analogous to that employed in Dube (2019). Additionally,
real minimum wage. The model also incorporates individual-level char- we narrow the focus of our analysis to the working-age population,
acteristics (quartic age, gender, race, education, family size, number resulting in a final sample of 373,555 individuals.
of children, and marital status) and state-level macroeconomic indi- Using the set of estimated coefficients of the minimum wage vari-
cators (unemployment rate, income per capita, and EITC supplement) ables from Eq. (12), we calculate the long-run elasticity of the pro-
on its right-hand side. The term 𝛩 includes the region fixed effect, portion of individuals falling below the income-to-needs cutoff, 𝑐, with
division-year fixed effect, and the region fixed effect interacted with respect to the minimum wage by estimating the following expression:
the indicators of recessionary years (2007–2009); 𝛾𝑠 𝑡 represents the ∑3
𝛼𝑐𝑘
region-specific linear trend, and 𝜖𝑐𝑖𝑠𝑡 denotes the error term. 𝐸𝑙𝑎𝑠𝑡𝑖𝑐𝑖𝑡𝑦 = 𝑘=0 ,
Our replication exercise deviates from the original estimation strat- 𝜇𝑐
egy of Dube (2019) in three ways. First, Russia does not have an where 𝜇𝑐 is the share of the population below 𝑐, which we can estimate
equivalent of the EITC program. Second, microdata on households using our data.

12
G.K. Ten and S. Wang Economic Modelling 148 (2025) 107086

Table 9
Long-run minimum wage elasticities for share of individuals with family income below
multiples of federal poverty threshold.
Multiple of the poverty threshold United States Russia
(1) (2) (3)
0.50 −0.455+ −0.048
(0.247) (0.058)
0.75 −0.461* −0.071
(0.186) (0.043)
1.00 −0.446** −0.070*
(0.137) (0.033)
1.25 −0.294* −0.036
(0.103) (0.024)
1.50 −0.156* −0.025
(0.093) (0.021)
1.75 −0.167* −0.022
(0.084) (0.018)
𝑁 4,662,781 373,555

Notes: This table reports the long-run (3 years) elasticities of the share with family
income under various multiples of the federal poverty threshold with respect to the
Fig. 5. Proportion of employed individuals with informal jobs, by age group. minimum wage. The multiples of the federal poverty threshold are listed in Column
Notes: The unit of observation is the individual. 𝑁 = 323,290. The 𝑦-axis shows the (1). Column (2) shows the U.S. estimates obtained from Dube (2019), and Column (3)
estimated difference in the proportion of employed individuals with informal jobs presents the authors’ estimates obtained from a nearly analogous empirical exercise
between the age group specified on the 𝑥-axis and those aged 15 to 18 years. Circles using Russian data. Robust standard errors, clustered by state (Column 2) or region
represent estimates obtained from the model without control variables, while diamonds (Column 3), are reported in parentheses. +, *, and ** denote significance at the 10%,
represent those obtained from the model with a full set of controls. The controls 5%, and 1% levels, respectively.
comprise educational attainment, marital status, an urban-area dummy, gender, an
indicator of having at least one child under 18, and region-year fixed effects. Vertical
bars represent 95% confidence intervals calculated using standard errors clustered by
region. design: a one-percentage-point increase in the nominal minimum wage
in Kamchatka raised the share of unemployed youth by 0.05 percentage
points. This finding also passes a set of robustness checks.
Next, we investigate whether the estimated employment effects of
the minimum wage policy differ across industries. We show that indus-
Table 9 presents the estimated elasticities for different income
tries with a greater capacity to substitute capital for labor are more
cutoffs. Column (1) shows multiples of the federal poverty threshold.
Column (2) reports baseline results retrieved from the original U.S. likely to experience steeper losses in youth employment. However,
study by Dube (2019).11 Column (3) presents the results obtained using when substitution is not feasible due to the underlying production
our data. Contrasting the results reported in columns (2)–(3), we ob- technology, industries tend to employ more workers informally when
serve that poverty in Russia is less sensitive to changes in the minimum the minimum wage increases. Consequently, we infer that differences
wage. For instance, in the United States, the estimated reduction in in employers’ ability to substitute labor with capital and formal workers
the proportions of individuals falling below 50% and 100% of the with informal ones — factors that vary across countries and time peri-
federal poverty threshold in response to a one-unit change in 𝐿𝑜𝑔(𝑀𝑊 ) ods — help explain the variation in estimated unemployment effects of
is −0.455 and −0.446, respectively. However, in Russia, the estimated minimum wages across studies.
elasticities are −0.048 and −0.070. The numerical values of estimated Lastly, we assess the effectiveness of Russia’s minimum wage policy
elasticities decrease in both countries as the income-to-needs cutoff, 𝑐, as a poverty alleviation tool. Consistent with observed job losses among
increases. youth and rising informal employment, we find its impact on work-
The results in this section indicate the relatively limited effec- ers’ income to be minimal. These findings cast doubt on the policy’s
tiveness of the Russian minimum wage as a poverty-alleviating tool effectiveness in supporting low-income families, its primary objective.
compared to the same policy in the United States. Job losses among It is important to acknowledge some limitations of our analysis that
young workers and the informalization of the Russian labor market, may open avenues for future research. First, our study focuses on a
both induced by minimum wage increases, are two natural explanations short time frame, covering the years 2009–2013. This period follows
for this finding. the global financial crisis of 2008 and precedes the economic downturn
triggered by the annexation of Crimea in 2014. Future research could
6. Conclusion examine the implications of the minimum wage using data spanning a
longer period while isolating the effects of a higher wage floor from
This paper assesses the employment effects of the minimum wage broader macroeconomic shocks. These shocks — driven by import
policy in Russia. We first investigate whether regional unemployment restrictions and domestic monetary policy — could also be leveraged
rates respond to the rising wage floor. Our results suggest that overall to explore possible heterogeneity in the minimum wage effect under
unemployment rates remain unchanged, while young workers (those different macroeconomic conditions.
aged under 30) are affected. The point estimates indicate that a one- Next, a recent advancement in the difference-in-differences and two-
percentage-point increase in the real minimum wage raises unemploy- way fixed effects literature has highlighted the fragility of estimated
ment rates among teenagers (15–19 years old) and young workers minimum wage effects in the presence of staggered policy adoption
(20–29 years old) by 0.046 and 0.056 percentage points, respectively. and heterogeneity in its effects (Dube and Lindner, 2024). Relatedly,
We also provide evidence of a redistribution of workers toward the in- Callaway and Sant’Anna (2021) discuss a family of estimators de-
formal sector resulting from the implementation of a higher wage floor. signed to address these concerns. The authors also demonstrate an
Our estimates are robust to alternative model specifications. We also application of one of their proposed methods by aggregating multi-
show that our findings hold when we employ an alternative empirical ple individual cases of minimum wage increases into a generalized
event-study research design using U.S. data. Specifically, they select
a time window, 2001–2007, during which the federal minimum wage
11
Refer to Table 3 in Dube (2019). remained unchanged. With this ‘‘clean’’ control group in hand, they

13
G.K. Ten and S. Wang Economic Modelling 148 (2025) 107086

Fig. 6. Proportion of employed individuals with informal jobs, by education group.

Notes: The unit of observation is the individual. 𝑁 = 323,290. The 𝑦-axis shows the estimated difference in the proportion of employed individuals with informal jobs between the
education group specified on the 𝑥-axis and those without any secondary education. Circles represent estimates obtained from the model without control variables, while diamonds
represent those obtained from the model with a full set of controls. The controls comprise age group, marital status, an urban-area dummy, gender, an indicator of having at least
one child under 18, and region-year fixed effects. Vertical bars represent 95% confidence intervals calculated using standard errors clustered by region.

estimate employment effects by leveraging variation in the minimum college wage premium over time (e.g., Fortin, 2006; Boero et al., 2025)
wage stemming from states that adopted higher wage floors at different may also prompt an investigation into the role of the minimum wage
points within this period. Notably, Cengiz et al. (2019) adopt a similar — and the forces moderating its impact — in shaping wage gaps among
approach to constructing a ‘‘clean’’ control group in their analysis. workers with different levels of education. Lastly, a nascent literature
These advancements may prompt efforts to construct a ‘‘clean’’ explores the effect of raising the wage floor on the non-profit sector
control group in Russia, where the minimum wage remains unchanged (e.g., Balsam et al., 2023; Meer and Tajali, 2023), making it natural to
for a sufficiently long period. The primary challenge, however, is the inquire whether the capital-labor substitution hypothesis investigated
high frequency of minimum wage adjustments in Russia compared in our paper can be tested in firms whose main objective function
to the U.S. Specifically, as Dube and Lindner (2024) pointedly note, differs from profit maximization.
local wage-floor increases in the U.S. between 1980 and 2019 largely
occurred in three waves: Once in the 1980s and twice in the 2000s. Declaration of competing interest
Combined with long periods of respite from federal minimum wage
increases, this pattern allows for various comparisons of employment
The authors Gi Khan Ten and Shun Wang declare no interest con-
trends between treated states and a ‘‘clean’’ control group. Meanwhile,
flict.
the Russian government strives to raise the federal wage floor on an
annual basis, making it particularly challenging to identify a control
group where the minimum wage remains unchanged for a sufficiently Appendix A. Supplementary data
long period.
At a broader level, of course, further research is required to obtain Data and code to replicate these findings are available online at the
a bigger picture of the ramifications of the minimum wage increase following link: https://doi.org/10.1016/j.econmod.2025.107086.
on the economy, as well as the role of firm-level technology and the
prevalence of informality in the labor market. For example, a strand Data availability
of literature examines the implications of the minimum wage for non-
labor outcomes, such as food consumption choices (e.g., Clark et al., I have shared the link to my data stored in Mendeley.
2020), health (e.g., Liu et al., 2024; Neumark, 2024), and the gender
pay gap (e.g., Majchrowska and Strawiński, 2018). Future research may Minimum Wage and Unemployment in Russia: A New Look on an Old
investigate the potential heterogeneity of these effects depending on the Construct (Original data) (Mendeley Data)
substitutability of production inputs and the extent of noncompliance
with the policy in question. Additionally, there is a rich literature
on skill-biased technological change and its implications for the wage
References
distribution (e.g., Card and DiNardo, 2002; Mallick and Sousa, 2017),
which opens the door to empirical explorations of the effects of skill- Allegretto, S., Dube, A., Reich, M., 2011. Do minimum wages really reduce teen
biased technological change on the magnitude of the minimum wage employment? Accounting for heterogeneity and selectivity in state panel data. Ind.
elasticity of labor demand. The ongoing discussion of the evolution of Relations 50 (2), 205–240.

14
G.K. Ten and S. Wang Economic Modelling 148 (2025) 107086

Allegretto, S., Dube, A., Reich, M., Zipperer, B., 2017. Credible research designs for Giuliano, L., 2013. Minimum wage effects on employment, substitution, and the teenage
minimum wage studies: A response to neumark, salas, and wascher. Ind. & Labor labor supply: Evidence from personnel data. J. Labor Econ. 31 (1), 155–194.
Relations Rev. 70 (3), 559–592. Gustafson, M., Kotter, J.D., 2018. Minimum wage and corporate policy. Working Paper.
Allen, R.G.D., 1938. Mathematical Analysis for Economists, Vol. 8, Macmillan. Haepp, T., Lin, C., 2017. How does the minimum wage affect firm investments in fixed
Arrow, K., Chenery, H., Minhas, B., Solow, R., 1961. Capital-labor substitution and and human capital? Evidence from China. Rev. Dev. Econ. 21 (4), 1057–1080.
economic efficiency. Rev. Econ. Stat. 43 (3), 225–250. Ham, A., 2018. The consequences of legal minimum wages in honduras. World Dev.
Azar, J., Huet-Vaughn, E., Marinescu, I., Taska, B., Von Wachter, T., 2024. Minimum 102, 135–157.
wage employment effects and labour market concentration. Rev. Econ. Stud. 91 Hamermesh, D., 1993. Labor Demand. Princeton University Press.
(4), 1843–1883. Han, A., Kim, T., Ten, G.K., Wang, S., 2023. Air pollution and gender imbalance in
Baek, J., Lee, C., Park, W., 2021. The impact of the minimum wage on the char- labor supply responses: Evidence from South Korea. Econ. Model. 124, 106290.
acteristics of new establishments: Evidence from South Korea. Labour Econ. 72, Harasztosi, P., Lindner, A., 2019. Who pays for the minimum wage? Am. Econ. Rev.
102059. 109 (8), 2693–2727.
Balsam, S., Mao, C.X., Xu, M., Zhang, Y., 2023. The effect of state minimum wage Hau, H., Huang, Y., Wang, G., 2020. Firm response to competitive shocks: Evidence
increases on nonprofit organizations. Nonprofit Volunt. Sect. Q. 52 (6), 1722–1752. from China’s minimum wage policy. Rev. Econ. Stud. 87 (6), 2639–2671.
Bassanini, A., Bovini, G., Caroli, E., Ferrando, J.C., Cingano, F., Falco, P., Felgueroso, F., Katz, L., Krueger, A., 1992. The effect of the minimum wage on the fast-food industry.
Jansen, M., Martins, P.S., Melo, A., Oberfichtner, M., Popp, M., 2024. Labor market Ind. Labor Relations Rev. 46 (1), 6–21.
concentration. Wages Job Secur. Eur. J. Hum. Resour.. Liu, D., Robone, S., Turati, G., 2024. The effect of minimum wages on self-reported
Bodman, G., 1916. Petrographische studien tiber engine antarktische gesteine, Wissens. physical and mental health in China. Econ. Model. 141, 106865.
Ergebnisse der Schwedischen Siidpolar-Exped. 1901–1903. Geol. Und Paloontologie Lordan, G., Neumark, D., 2018. People versus machines: The impact of minimum wage
3 (15), 58–59. on automatable jobs. Labour Econ. 52, 40–53.
Boero, G., Nathwani, T., Naylor, R., Smith, J., 2025. The college wage premium in the Majchrowska, A., Strawiński, P., 2018. Impact of minimum wage increase on gender
UK: decline and fall? Oxf. Econ. Pap. 77 (1), 1–18. wage gap: Case of Poland. Conomic Model. 70, 174–185.
Burkhauser, R.V., Couch, K.A., Wittenburg, D.C., 2000. A Reassessment of the new Mallick, S.K., Sousa, R.M., 2017. The skill premium effect of technological change: New
economics of the minimum wage literature with monthly data from the Current evidence from United States manufacturing. Int. Labour Rev. 156 (1), 113–131.
Population Survey. J. Labor Econ. 18 (4), 653–680. Meer, J., Tajali, H., 2023. Effects of the minimum wage on the nonprofit sector. Oxf.
Callaway, B., Sant’Anna, P.H., 2021. Difference-in-differences with multiple time Econ. Pap. 75 (4), 1012–1032.
periods. J. Econometrics 225 (2), 200–230. Meer, J., West, J., 2016. Effects of the minimum wage on employment dynamics. J.
Card, D., DiNardo, J.E., 2002. Skill-biased technological change and rising wage Hum. Resour. 51, 500–522.
inequality: Some problems and puzzles. J. Labor Econ. 20 (4), 733–783. Neumark, D., 2024. The effects of minimum wages on (almost) everything? A review
Cengiz, D., Dube, A., Lindner, A., Zipperer, B., 2019. The effect of minimum wages on of recent evidence on health and related behaviors. Labour 38 (1), 1–65.
low-wage jobs. Q. J. Econ. 134 (3), 1405–1454. Neumark, D., Salas, J.M.I., Wascher, W., 2014. Revisiting the minimum wage-
Chen, Y., 2019. What do establishments do when wages increase? Evidence from employment debate: Throwing out the baby with the bathwater? Ind. Labor
minimum wages in the United States. Working Paper. Relations Rev. 67 (3), 608–648.
Clark, K.L., Pohl, R.V., Thomas, R.C., 2020. Minimum wages and healthy diet. Contemp. Neumark, D., Wascher, W., 1992. Employment effects of minimum and subminimum
Econ. Policy 38 (3), 546–560. wages: Panel data on state minimum wage laws. Ind. Labor Relations Rev. 46 (1),
Clemens, J., Wither, M., 2019. The minimum wage and the Great Recession: Evidence 55–81.
of effects on the employment and income trajectories of low-skilled workers. J. Neumark, D., Wascher, W., 2010. Minimum Wages and Employment. The MIT Press.
Public Econ. 174, 53–67. Pérez, J.P., 2020. The minimum wage in formal and informal sectors: Evidence from
Comola, M., Mello, L.D., 2011. How does decentralized minimum wage setting affect an inflation shock. World Dev. 133, 104999.
employment and informality? The case of Indonesia. Rev. Income Wealth 57 (1), Porta, R.La., Shleifer, A., 2008. The unofficial economy and economic development.
79–99. Brookings Pap. Econ. Act. 39 (2), 275–363.
Draca, M., Machin, S., Van Reenen, J., 2011. Minimum wages and firm profitability. Portugal, P., Cardoso, A., 2006. Disentangling the minimum wage puzzle: An analysis
Am. Econ. Journal: Appl. Econ. 3 (1), 129–151. of worker accessions and separations. J. Eur. Econ. Assoc. 4 (5), 988–1013.
Dube, A., 2019. Minimum wages and the distribution of family incomes. Am. Econ. Rybczynski, K., Sen, A., 2018. Employment effects of the minimum wage: Panel data
Journal: Appl. Econ. 11 (4), 268–304. evidence from Canadian provinces. Contemp. Econ. Policy 36, 116–135.
Dube, A., Lester, T.W., Reich, M., 2010. Minimum wage effects across state borders: Sabia, J., Burkhauser, R., Hansen, B., 2012. Are the effects of minimum wage increases
Estimates using contiguous counties. Rev. Econ. Stat. 92 (4), 945–964. always small? New evidence from a case study of New York State. Ind. Labor
Dube, A., Lindner, A., 2024. Minimum wages in the 21st century. Handb. Labor Econ. Relations Rev. 65 (2), 350–376.
5, 261–383. Wong, S.A., 2019. Minimum wage impacts on wages and hours worked of low-income
Fortin, N.M., 2006. Higher-education policies and the college wage premium: workers in Ecuador. World Dev. 116, 77–99.
Cross-state evidence from the 1990s. Am. Econ. Rev. 96 (4), 959–987.

15