Revista de economía mundial 66, 2024, 25-42

ISSN: 1576-0162

DOI: http://dx.doi.org/10.33776/rem.vi66.7699

Subcontracting retreat:

early eStimateS uSing adminiStrative data

RetRaimiento de la subcontRatación:

estimación tempRana usando datos administRativos

Enrique Kato-Vidal

Universidad Autónoma de Querétaro, México

enriquekato@uaq.mx

Paulina Hernández-Mendoza

Secretaría de Desarrollo Sustentable de Querétaro, México

phernandezm@queretaro.gob.mx

Recibido: abril 2023; aceptado: febrero 2024

abStract

Labor ﬂexibility policies have increased temporary and subcontracted

jobs. These types of employment are associated with lower quality jobs. In this

paper, ﬁve-year economic census data and social security administrative data

were collected to estimate, in non-census years, out-of-sample the percentage

of subcontracted jobs. This evidence is relevant for assessing the impact of a

legislation that banned subcontracting. Predictor variables such as time ﬁxed

effects, industry, and the rate of temporary workers were used. The results

show a signiﬁcant decrease in subcontracting, particularly in the secondary

sector. These ﬁndings raise the need for future research on improvements in

post-reform well-being in the labor market.

Keywords: Temporary Work Agencies, Domestic Outsourcing, Out-of-

Sample, Job Quality, INEGI-Mexico.

reSumen

Las políticas de ﬂexibilización laboral han aumentado los empleos

temporales y subcontratados. Esas categorías de empleo están asociadas con

empleos de menor calidad. En este artículo, se recopilaron datos quinquenales

del censo económico y datos administrativos de la seguridad social para

estimar fuera de la muestra, en años no censales, el porcentaje de trabajos

subcontratados. Esta evidencia es relevante para evaluar el impacto de una

legislación que prohibió la subcontratación. En la estimación se utilizaron

efectos temporales y variables predictoras como la industria, el tamaño de

la empresa y la tasa de trabajadores temporales. Los resultados muestran

un descenso importante de la subcontratación, particularmente en el sector

secundario. Estos hallazgos plantean la necesidad de investigación futura

sobre mejoras en el bienestar que haya habido en el mercado laboral.

Palabras clave: Agencias de empleo temporal, subcontratación, fuera de la

muestra, calidad de empleo, INEGI-México.

JEL Classiﬁcation/ Clasiﬁcación JEL: C82, J21, M51.

Revista de economía mundial 66, 2024, 25-42

1. introduction

The growing presence of subcontracted employment has been observed

through ﬁve-year economic censuses in developed and developing countries.

In contrast, employment surveys do not report a growing trend but a stable

presence. This paper seeks to reconcile census information with administrative

data to correct the underestimation by surveys and obtain an intercensal ﬁgures

of domestic outsourcing to monitor this increasingly widespread employment

class. In particular, it seeks to make inferences about subcontracting labor in

2021 and 2023, that is, after the economic crisis caused by COVID-19 and

several years since the last economic census in Mexico. The use of administrative

micro data is increasing in research that analyzes the disruptions caused by the

pandemic in the labor market (Vavra, 2021). Administrative data have also

been used to extrapolate and infer unknown numbers of individuals working

in illegal activities (Baldassarini, Chiariello & Tuoto, 2019). An advantage of

administrative data is that such data allow linking workers with ﬁrms and

overcoming the low response rates for Labor Force Surveys or lack of sample

representativeness (Bosch & Campos-Vazquez, 2014).

Although employment surveys do not adequately quantify the rate of

outsourced jobs (Bernhardt, Spiller & Theodore, 2013), we posit that intercensal

subcontracting ﬁgures can be obtained by combining economic census data

and social security administrative data, given that i) subcontracting labor is

related to observable characteristics, such as industrial structure, ﬁrm size and

proportion of temporary workers and ii) a consistent estimate, accompanied

by robustness tests, can be a useful instrument to compensate for the lack of

information in or underestimations by Labor Force Survey data.

Domestic outsourced employment has been a poorly studied topic,

even though it is a movement traceable to the late 1970s (Weil, 2014) and

recently in a developing economy (Estefan, et al., 2024). Although outsourced

and temporary are related, it is important to note that not all temporary

employment is subcontracted employment. In the past, the high presence

of temporary jobs has been analyzed. Coincidentally, a characteristic of

subcontracting labor is the short duration of their contracts. Temporary jobs

are not a speciﬁc phenomenon of workers with low skills. Many independent

workers have resorted to temporary contracts to complement their income,

and many highly trained professionals have been hired through employment

agencies (MGI, 2016). In a global context, in Europe, half of young people have

28 Enrique Kato-Vidal · Paulina Hernández-Mendoza

a temporary job, and in the United States, the growth of employment from

2007 to 2015 was explained by temporary employment (MGI, 2016).

Thus, as the number of temporary jobs increases, job instability and low

wages increase (Bernhardt, Spiller & Theodore, 2013). Faced with this trend

and to address some of the negative consequences, instruments are required

to assess the depth and scope of subcontracting, in order to grasp an overall

perspective. Empirically, Pollio, et al. (2023) report that a higher fraction of

temporary employees decreases the quality of permanent jobs, refusing the

hypothesis that temporary jobs are a vehicle to a better permanent positions

(Chambel, Lopes & Batista, 2016).

Relying in administrative data has advantages and provides insights on

subcontracting. First, temporary workers, as well as medium and large ﬁrms,

both pillars to subcontracting, are precisely captured, thus correct estimations

can be drawn. Second, data covers and disaggregates ﬁrm size and industry

afﬁliation, this is relevant to analyze the subcontracting composition and

to describe the sources of variation of the overall economy. Finally, due to

ofﬁcial registry, consistent temporal comparisons were possible, this is a major

advantage over survey data and its high non-response rates, especially in

pandemic months.

In April 2021, Mexico approved a strict legislation to regulate subcontracting

and for improving the working conditions (Government of Mexico, 2021;

Estefan, et al., 2024). Follow-up of Bank of México (2021) and the Government

on the issue has used social security records. The timing of the legislation

occurred while the post COVID-19 recovery was still in progress. Our results

found that -due to layoffs- subcontracted labor was at a local minimum, that

was a valuable time frame for ﬁrms to internalize and complied the tighter rules

for subcontracting, before reaching its full capacity. In that setting, possibly the

reform had a lower than expected transition cost.

The remainder of the paper is organized as follows: section two presents

related literature, section three describes the data, section four presents the

methodology used, section ﬁve reports the results, and then the conclusions

are presented.

2 related literature

Independent workers and subcontracted employees could appear similar

due to the short duration of their employment. An independent worker who

supply services or sells goods is characterized (MGI, 2016) by i) a high level

of autonomy, with ﬂexibility to choose quotas and workload; and ii) variable

remuneration, based on contracts or sales. Subcontracted employees, in

contrast, are personnel recruited by ﬁrms from contractors to maintain

separate functions considered noncore. By gender, Kauhanen & Nätti (2015)

found that women are overrepresented in involuntary, temporary or part-time

jobs, and these jobs were associated to the highest risk of unemployment.

Subcontracting retreat: early eStimateS uSing adminiStrative data

reviSta de economía mundial 66, 2024, 25-42

Subcontracting labor and its effects on the lives of workers have been

studied using various sources of information. Bernhardt et al. (2016) analyzed

subcontracting trends in the United States and the effect of subcontracting

on employment quality. As a source of information, they used the Bureau of

Economic Analysis, speciﬁcally the input-output tables, because they broadly

show the national subcontracting trends. One disadvantage of this approach is

that these data are not updated regularly because the information comes from

an economic census; thus, these data should be used with caution (Bernhardt,

et al., 2016: 31-32).

A limited capacity to understand the magnitude of subcontracting labor

and the impact on employment quality prevents the creation of appropriate

policies (Weil, 2014). Therefore, to obtain more detailed information, data

at the industrial level have been consulted (Bernhardt, et al., 2016). The

information reviewed was a) micro data from government surveys of households

and businesses or administrative data; b) datasets of industry trade groups;

c) structured case studies of ﬁrms; and d) interviews with industry experts.

Bernhardt et al. (2016: 35) argue that this information allows a thorough

analysis of subcontracting but that for a more complete understanding, it is

necessary to collect new data that are representative at the national level.

To analyze the growth of temporary employment services, the Current

Employment Survey of the Bureau of Labor Statistics and the National

Association of Temporary and Stafﬁng Services (NATSS) has also been used;

this survey measures employment in temporary services. According to Segal

and Sullivan (1997: 117-118), these data are not too sensitive to measure

temporary employment. A better database is the Current Population Survey

(CPS), a monthly survey of American households that provides explicit

identiﬁers for each household; however, it does not allow matching individuals

with households. CPS also underestimates subcontracting. Segal and Sullivan

(1997: 120) state that this discrepancy may occur because respondents

report the place where they are working as their main employer instead of

reporting their employer as the ﬁrm that provides temporary services to ﬁrms.

In the same vein, Bernhardt, Spiller & Theodore (2013) report the workers’

inability to accurately identify whether their employer was a contractor or not.

In Mexico, to measure the causal impacts of the outsourcing reform Estefan,

et al. (2024) used matched hirer–employee from social security data, coupled

manufacturing survey and census data.

Using administrative social security data from Germany, Goldschmidt and

Schmieder (2017) studied subcontracting in the food, cleaning, security and

logistics sectors. The advantage was the availability of integrated employment

biographies (IEB) data from 1975 to 2009 that contained information regarding

duration of employment, total payment at the end of the period worked, type

of employment and a large number of demographic variables. One limitation

is that there are identiﬁers for establishments but not for ﬁrms. Currently,

there are efforts to systematize and more intensively use administrative data

(D’Angiolini, De Salvo, and Passacantilli, 2016).

30 Enrique Kato-Vidal · Paulina Hernández-Mendoza

In a context in which data do not have a speciﬁc subcontracting variable,

Goldschmidt and Schmieder (2017) applied an on-site outsourcing method

that focused on personnel ﬂow between establishments, i.e., ﬁrms hire a part

of their workforce to work as legally independent subcontractors even though

the employees continue their work in the same physical location. According to

these authors, it is likely that there is underreporting and many cases where

subcontracted workers do not change locations. Additionally, part-time workers

and temporary employment suffer stigma or devaluation (Pedulla & Mueller-

Gastell, 2019). Considering that employment status is related to job quality,

it is not surprising that involuntarily part-time employees have low satisfaction

with working conditions and perform badly on most quality of working life

outcomes (Gevaert, et al., 2023; Dawson, Veliziotis & Hopkins, 2017).

Analyses have been frequently carried out in the United States and Europe,

creating the need to explore what has happened in other parts of the world.

Also, as a consequence of insufﬁcient statistics, the literature on subcontracting

or domestic outsourcing is scarce. Therefore, other types of work are compared.

For example, Lee & Lee (2015) study the wage gap between temporary and

permanent workers. The dichotomy temporary-permanent has also been used

to evaluate the impacts of globalization (Görg & Görlich, 2015). In developing

countries, workers are usually separated into formal and informal. In Brazil the

formality is understood as the employees with a work card and civil servants

(Corseuil & Foguel, 2012). Similarly, in Mexico, the informal workers are those

workers who are not covered by the formal social security program (Bosch &

Campos-Vazquez, 2014). The above mentioned studies can be performed on

labor force or household survey data, however that data do not accurately

measure subcontracted employment; therefore, there is a need to probe the

usefulness of administrative data in understanding subcontracted work.

3 data

In this paper, information from economic censuses on groups of workers is

used: i.e., owners or relatives, paid personnel, and subcontracted personnel.

The ﬁgures from economic censuses are considered reliable because they

are obtained from ﬁeld work in which the informants are responsible for the

ﬁrms. The statistics are released every ﬁve years and are available by industry

and ﬁrm size. Using these data, the rates of subcontracted workers can be

calculated, and these values are associated with the information available in

administrative data, which are collected monthly; therefore, subcontracting

can be monitored in intercensal years. Statistically, the data describe the

same economy, although perceived differently. The Economic Census measure

variables with information provided by most of the ﬁrms, which were visited

during ﬁeldwork. In contrast, administrative data are derived from information

provided by a set of ﬁrms listed by the Social Security Institute. Administrative

data from Mexico have been useful for evaluating employers, temporary

Subcontracting retreat: early eStimateS uSing adminiStrative data

reviSta de economía mundial 66, 2024, 25-42

contracts and employment ﬂexibility (Bosch & Campos-Vazquez, 2014; Kato-

Vidal, 2021).

Figure 1 shows the subcontracting rates across four census years for

microbusinesses and small, medium and large ﬁrms. Figure 1 reveals that

medium and large ﬁrms concentrate the higher shares of subcontracting than

do small ﬁrms and microbusinesses. Furthermore, in the decade from 2003 to

2013, there was a signiﬁcant increase in subcontracting. The increase continued

with lower intensity from 2008 to 2018. A concentration of subcontracting is

observed. Together, three industries accounted for 89% of the subcontracted

workers in Mexico in 2018. Two thirds of the subcontracted workers served

two industries (manufacturing and services). The trade industry also had a high

subcontracting labor.

The estimates obtained were used to predict subcontracting rates in two

years (i.e., 2021 and 2023) in which there is no census information. Table 1 shows

the number of workers in the census years from 2003 to 2018, differentiating

workers who are owners or family members, paid workers and subcontracting

labors (see Figure 2). Subcontracted personnel have increased considerably,

whether compared to total employees or to paid workers. The subcontracting

rates that are reported in the remainder of the paper are calculated with

respect to paid employment. By ﬁrm size, the rates of greatest interest for

the study correspond to medium and large ﬁrms, for which the highest rates

of subcontracting are observed. It is recommended to not truncate the data,

excluding the smaller ﬁrms, especially in developing countries, because those

ﬁrms are associated to a large share of total employment (Li y Rama, 2015).

To predict with greater precision subcontracting rates, ﬁxed effects

were estimated by industry. Given the partial compatibility’ codes between

census information and administrative data, eight of nine administrative

industries were analyzed, the two service subsectors were merged into one.

In the administrative data, there are covariates that are associated with

the subcontracting rate in the literature. For our estimation, we used three

table 1. mexico: employed perSonS by claSS of worker (thouSandS)

2003 2008 2013 2018

Total 16 240 20 117 21 576 27 133

Paid employees 10 552 11 414 12 197 16 254

Owners and relatives 4 298 5 995 5 801 6 193

Subcontracted 1 003 2 342 3 018 4 129

Independent 387 366 560 556

Subcontracted

As a % of total 6% 12% 14% 15%

As a % of paid employees 10% 21% 25% 25%

Note: Independent = fees or commissions without ﬁxed salary. The ﬁgures represent people in all

sectors, including agriculture and other services. Source: economic censuses, INEGI.

32 Enrique Kato-Vidal · Paulina Hernández-Mendoza

covariates: i) rate of women in the labor force, ii) rate of young workers (<25

years of age), and iii) rate of temporary workers. It has been found that women

are more likely to submit for a temporary job (Pedulla & Mueller-Gastell, 2019),

and also women are the majority of workers in temporary and part-time jobs

(Kauhanen & Nätti, 2015).

The second covariate associates non-prime-age workers (younger than

twenty-ﬁve and older than ﬁfty-four) to nonstandard employment relationships

(such as part-time and temporary jobs) (Pedulla & Mueller-Gastell, 2019), and

an intersectionality surges: those on casual contracts are more likely to be

younger, female, single… (Dawson, Veliziotis & Hopkins, 2017). Thirdly, a larger

use of temporary workers might indicate a strategy to reduce the wage bill

and a perceived erosion in the quality of jobs (Pollio, et al., 2023; Kauhanen

& Nätti, 2015).

Table 2 provides the averages (and standard deviations) for our covariates.

Two trends are observed since 2003 through 2023: a) growing participation

by women workers, from 34.4% to 39.5%; and b) an increase in temporary

workers from 10% to a peak of 13% in 2018, onwards a mild reduction is

observed. Regarding the distribution of employment based on ﬁrm size,

medium and large ﬁrms had the highest rates of temporary employment.

table 2. mexico: covariateS by firm Size and year (percent relative to paid employeeS)

Panel a) All ﬁrms

2003 2008 2013 2018 2021 2023

Temporary 9.72 10.99 13.16 13.44 13.14 12.73

s.d. (15.17) (13.09) (10.21) (9.18) (8.28) (8.92)

Women 34.4 35.83 36.55 37.39 38.65 39.54

s.d. (11.12 (11.37) (10.96) (10.84) (10.45) (10.41)

Youth 20.53 17.75 15.81 16.43 14.71 14.28

s.d. (5.49) (4.28) (4.31) (4.23) (4.28) (3.79)

Panel b) 2003-2023

Micro-

businesses

Small

ﬁrms Medium ﬁrms Large

ﬁrms

Temporary 6.99 11.24 14.72 12.74

s.d. (10.0) (12.18) (12.22) (8.54)

Women 39.23 34.89 34.27 39.88

s.d. (12.11) (12.9) (11.1) (8.69)

Youth 10.45 14.73 16.71 17.64

s.d. (3.19) (3.12) (3.67) (5.17)

Note: ‘s.d.’ Standard Deviation. Source: Social security data, IMSS.

Subcontracting retreat: early eStimateS uSing adminiStrative data

reviSta de economía mundial 66, 2024, 25-42

4 empirical Strategy

As explained in the previous section, economic censuses are the instruments

that best measure the subcontracting labor in a country. However, one disadvantage

of this information is the low frequency at which the data are updated, as they are

published only once every ﬁve years. Though, the data are available by ﬁrm size and

economic activity. To ﬁll the data gap, we propose to associate the data provided

by the ﬁrms in economic censuses with the labor information available in the social

security administrative data. The variable of interest is the rate of subcontracted

workers relative to paid employment (θ i,s,t ). The census rate θ serves to estimate

correctly the rate of subcontracted workers obtained from administrative data (θ).

Also providing a ratio, rather than absolute values, facilitate comparison across

time and economies. The data is matched in census year t for each industry-ﬁrm

size pair (i,s).

As such, the following equation is ﬁtted to census data and used to estimate

with administrative data out-of-sample when there is no census information:

Θi = β0 + β1Ti + β2Ii + β3Si + βkXi + εi

figure 1. mexico: Subcontracting labor by firm Size (percent of total Subcontracting labor in

cenSuS yearS)

Note: Microbusinesses, 0-10 employees; small ﬁrms, 11-50 employees;

median ﬁrms, 51-250 employees; and large ﬁrms, 251 and more employees.

Source: economic censuses, INEGI.

34 Enrique Kato-Vidal · Paulina Hernández-Mendoza

where T, I and S are a set of indicator variables to isolate the ﬁxed effects

of the census years T={2003,2008,2013,2018}, industries I={1,…,8} and

ﬁrm size S={1,…,4}, respectively. In the vector Xk, interactions and three

explanatory variables -calculated with administrative data- were included (see

Tables 2 and Figure A1). The last term, ε, corresponds to the random error

term. Estimation was performed using least squares. The estimated coefﬁcients

are reported in Table A1 in the Annex.

The administrative data comprise a sample of workers different from the

sample in the economic census. The problem of representativeness associated

with administrative data samples was noted by Vavra (2021), who recognized

administrative data as a complement, not a substitute, for traditional data

sources. In addition, in our paper, to match the sectoral classiﬁcations used in

the census and by the social security administration, various industries were

excluded; this data exclusion explains the higher rate of subcontracting in the

administrative data than in the economic census in 2018 (30.7% and 25.5%,

Table 1).

It is also possible to compare employment survey data and economic

census data; the 2018 observed rate of subcontracting is lower in the

employment survey data than in the economic census data (18.1% and

25.5%, respectively). Although the available information sources allow us to

figure 2. mexico: claSS of workerS (% total)

Source: Economic censuses, INEGI. Stata Command provided by Cox (2004).

Subcontracting retreat: early eStimateS uSing adminiStrative data

reviSta de economía mundial 66, 2024, 25-42

infer rates close to the census rates, there are degrees of underestimation or

overestimation that must be taken into account. In our case, we are interested

in identifying the direction and amount of change in the subcontracting rate

between 2018, 2021 and 2023, not the absolute subcontracting rate.

Recently, Estefan, et al. (2024) carried out a comprehensive assessment of

Mexican on-site outsourcing. They reported a drastic drop of subcontracting

for 2023, two years after Mexico’s government banned outsourcing and

create a mandatory registry for all specialized contractors (p. 7). We follow

those results to predict the estimates for 2021 and 2023. Particularly, we

assume that 2021 rates maintain historically high values. Later because of the

subcontracting reform, the 2023 rates sharply decline reaching rates as low as

two decades earlier.

5 reSultS

Based on our assumptions, the administrative data indicate that there was

a signiﬁcant reduction in the subcontracting rate in 2023. Using the equation

of the previous section, it was estimated that the rate of subcontracting labor

in Mexico decreased by approximately 13 percentage points (pp) between

2018 and 2023, and less than one pp 2018-2021 (see Table A2). The amount

of this decrease is large and is equivalent to a setback of more than a decade.

The details of the estimate are shown in the Annex (Table A1). Both the dummy

for industry and the dummy for ﬁrm size were signiﬁcant.

To determine why the rate of subcontracting labor decreased, the data

were analyzed by ﬁrm size and by industry. As seen in Table A2, the decrease

in the rate of subcontracted workers was of greater magnitude in medium and

large ﬁrms, that is, ﬁrms with more than 50 workers. The notorious decrease of

15 pp between 2018 and 2023 cannot be explained by changes of covariates,

but due to a government ban on subcontracting. However, in medium-sized

ﬁrms (51 to 250 employees), the estimated decrease of 3.2 pp between 2018

and 2021 is important. This differentiated behavior is explained by the greater

percentage of job losses at medium-sized ﬁrms, both in total and subcontracted

jobs, than at small ﬁrms.

A second approach to explain why a decrease in the percentage of

subcontracting is estimated is to analyze the behavior by industry. As seen in

Table A2, the industries with the greatest reduction in subcontracting were in

the secondary sector (mining, manufacturing and construction), 2018-2023. In

these industries, the relative decrease in subcontracting was bolder. Particularly,

in manufacturing which subcontracting rate dropped below 5% in 2023. To

achieve a greater reduction in subcontracting, a detailed inspection of labor

contracts in the trade sector would be required. Regarding the total economy

from 2018 to 2021, there was a net creation of jobs, which also contributed

to a lower subcontracting rate. That is, there were two different processes that

result in a lower rate of subcontracting: the dismissal of subcontracted workers

and the recovery of employment (not subcontracted).

36 Enrique Kato-Vidal · Paulina Hernández-Mendoza

The results presented were obtained using administrative data that served

to parameterize an equation and facilitate an out-of-sample prediction for

2021 and 2023. As a by-product, Figure A1 show the complex relationship

between the covariates and the subcontracting rate. First, a higher proportion

of temporary workers is associated to a higher percent of subcontracting

labor. Second, a higher rate of women labor is not unambiguously related to

subcontracting: in manufacturing, the highest rate of subcontracting is found

in ﬁrms who employ relatively few women; nevertheless, the opposite is true

in the service sector. Thirdly and last, the subcontracting labor is consistently

higher in large ﬁrms, barely increasing in youth workers. In small ﬁrms, low -and

decreasing rates of- subcontracting are related to a higher percent of youth

workers.

6 concluSionS

In the last two decades, the rate of subcontracting labor has increased

considerably, directly affecting working conditions. Despite this, perhaps due

to the scarcity of available information, there has been little research on the

topic. To reverse subcontracting, in Mexico a law was enacted in 2021 and

the ﬁrst positive outcomes have been observed. To increase visibility and to

monitor subcontracting trend changes, we linked detailed data from economic

censuses and the up-to-date administrative data.

Subcontracting labor is a form of employment that provide ﬁrms with

ﬂexibility, which could be a tool in markets with volatile demand. From the

perspective of workers, subcontracting causes job instability, stigma and a gap

in employment beneﬁts between subcontracted personnel and personnel hired

directly by a ﬁrm. In the long term, statistics show that subcontracting labor

has ceased to be a focal phenomenon and has become a very generalized

practice among industries. In fact, in the last 20 years in Mexico, much of

the expansion of employment occurred through subcontracting, reaching one

subcontracted worker for every four paid workers in 2018. We estimate that

the reform reduced subcontracted staff by half in 2023.

Analytically, a widespread shift in hiring practices was required to attain

this sharp decrease, not only the movement of some covariates. To monitor

the phenomenon of subcontracting in years after the 2018 economic census,

an equation was ﬁtted and applied out-of-sample using administrative data

for 2021 and 2023. The use of administrative data implies risks and involves

a substantial time investment. These costs are worth it because they can

achieve insights that could not be achieved otherwise. In the future, a greater

integration of census statistics, employment surveys and administrative data is

needed so that subcontracting trends can be better appraised.

Subcontracting retreat: early eStimateS uSing adminiStrative data

reviSta de economía mundial 66, 2024, 25-42

referenceS

Bank of Mexico [Banco de México]. (2021). “Dinámica de los Puestos de

Trabajo Aﬁliados al IMSS ante la Reforma a la Subcontratación [Dynamics

of Jobs Afﬁliated to the IMSS before the Subcontracting Reform].” Extracto

del Informe Trimestral, Jul-Sep, Recuadro 4, pp. 49-52. Available at:

https://www.banxico.org.mx/publicaciones-y-prensa/informes-trimestrales/

recuadros/%7BDD53353A-8A8C-9497-E40F-E9B51B89CCE1%7D.pdf

Bernhardt, A., Batt, R., Houseman, & Appelbaum, E. (2016). “Domestic

Outsourcing in the U.S.: A Research Agenda to Assess Trends and Effects on

Job Quality.” Available at SSRN: http://dx.doi.org/10.2139/ssrn.2757254

Bernhardt, A., Spiller, M. W., & Theodore, N. (2013). “Employers Gone Rogue:

Explaining Industry Variation in Violations of Workplace Laws.” ILR Review,

66(4), 808-832. https://www.jstor.org/stable/24369555

Baldassarini, A., Chiariello, V., & Tuoto, T. (2004). “Estimating Criminal

Populations from Administrative Registers.” Statistika- Statistics and

Economy Journal, 99(3), pp. 287-300.

Bosch, M., & Campos-Vazquez, R. M. (2014). “The Trade-Offs of Welfare Policies

in Labor Markets with Informal Jobs: The Case of the” Seguro Popular”

Program in Mexico.” American Economic Journal: Economic Policy, 6(4),

71-99. https://doi.org/10.1257/pol.6.4.71

Chambel, M. J., Lopes, S., & Batista, J. (2016). “The effects of temporary

agency work contract transitions on well-being”. International Archives

of Occupational and Environmental Health, 89, 1215-1228. https://doi.

org/10.1007/s00420-016-1158-y

Corseuil, C. H. L., & Foguel, M. N. (2012). “Economic Expansion and Increase

in Labour Market Formality: A Poaching Approach.” Revista Brasileira de

Economia, 66(2), 207-224. https://bibliotecadigital.fgv.br/ojs/index.php/

rbe/article/view/3870

Cox, N. J. (2004). “Speaking Stata: Graphing Categorical and

Compositional Data.” The Stata Journal, 4(2), pp. 190-215. https://doi.

org/10.1177/1536867X0400400209

D’Angiolini, G., De Salvo, P. & Passacantilli, A. (2016). “ISTAT’s New Strategy

and Tools for Enhancing Statistical Utilization of the Available Administrative

Databases.” Statistika- Statistics and Economy Journal, 2016, 96(3), pp.

65-71.

Dawson, C., Veliziotis, M., & Hopkins, B. (2017). “Temporary employment, job

satisfaction and subjective well-being”. Economic and industrial democracy,

38(1), 69-98. https://doi.org/10.1177/0143831X1455978

Estefan, A., Gerhard, R., Kaboski, J. P., Kondo, I. O., & Qian, W. (2024).

“Outsourcing Policy and Worker Outcomes: Causal Evidence from a Mexican

Ban” (No. w32024). National Bureau of Economic Research. https://www.

nber.org/papers/w32024

Gevaert, J., De Moortel, D., Eiffe, F. F., & Vanroelen, C. (2023). “Does

employment status matter for job quality? A cross-national perspective”.

38 Enrique Kato-Vidal · Paulina Hernández-Mendoza

Work: A Journal of Prevention, Assessment & Rehabilitation, vol. 75, no. 2,

pp. 521-539. https://doi.org/10.3233/WOR-210916

Görg, H., & Görlich, D. (2015). “Offshoring, Wages and Job Security of

Temporary Workers.” Review of World Economics, 151(3), 533-554. https://

doi.org/10.1007/s10290-015-0220-2

Government of Mexico [Gobierno de México]. (2021). “Reforma en materia de

subcontratación laboral [Reform on Labor Subcontracting].” Diario Oﬁcial

de la Federación [Ofﬁcial Gazette], 23/04/2021. México. https://www.dof.

gob.mx/nota_detalle.php?codigo=5616745&fecha=23/04/2021

Goldschmidt, D., & Schmieder, J. F. (2017). “The Rise of Domestic Outsourcing

and the Evolution of the German Wage Structure.” The Quarterly Journal

of Economics, 132(3), pp. 1165-1217. https://doi.org/10.1093/qje/qjx008

Kato-Vidal, E. (2021). “Evaluación de la ﬂexibilización laboral en México [An

Assement of Labor Flexibility in Mexico]”. Problemas del Desarrollo. Revista

Latinoamericana de Economía, 2021, 52. pp. 111-136. DOI: https://doi.

org/10.22201/iiec.20078951e.2021.Especial.69555

Kauhanen, M., & Nätti, J. (2015). “Involuntary temporary and part-time work,

job quality and well-being at work”. Social Indicators Research, 120, 783-

799. https://doi.org/10.1007/s11205-014-0617-7

Lee, H., & Lee, J. (2015). “The Impact of Offshoring on Temporary Workers:

Evidence on Wages from South Korea.” Review of World Economics, 151(3),

555-587. https://doi.org/10.1007/s10290-015-0215-z

Li, Y., & Rama, M. (2015). “Firm Dynamics, Productivity Growth, and Job

Creation in Developing Countries: The Role of Micro- and Small Enterprises.”

The World Bank Research Observer, 30(1), 3-38. https://doi.org/10.1093/

wbro/lkv002

McKinsey Global Institute -MGI-. (2016). Independent Work: Choice, Necessity,

and the Gig Economy. McKinsey & Company. https://www.mckinsey.com/

featured-insights/employment-and-growth/independent-work-choice-

necessity-and-the-gig-economy

Pollio, C., Landini, F., Prodi, E. & Arrighetti, A. (2023). “Does Temporary

Employment undermine the Quality of Permanent Jobs?” SSRN: http://

dx.doi.org/10.2139/ssrn.4441533

Pedulla, D. S., & Mueller-Gastell, K. (2019). “Nonstandard work and the job

search process: Application pools, search methods, and perceived job

quality”. RSF: The Russell Sage Foundation Journal of the Social Sciences,

5(4), 130-158. DOI: https://doi.org/10.7758/RSF.2019.5.4.05

Segal, L. M., & Sullivan, D. G. (1997). “The Growth of Temporary Services

Work.” Journal of Economic Perspectives, 11(2), 117-136. https://doi.

org/10.1257/jep.11.2.117

StataCorp. (2017). Stata User´s Guide. Release 15. Stata Press. Available at:

https://www.stata.com/manuals15/rmargins.pdf

Vavra J. (2021). “Tracking the Pandemic in Real Time: Administrative Micro Data

in Business Cycles Enters the Spotlight.” Journal of Economic Perspectives,

35(3), pp. 47-66. https://doi.org/10.1257/jep.35.3.47

Subcontracting retreat: early eStimateS uSing adminiStrative data

reviSta de economía mundial 66, 2024, 25-42

Weil, D. (2014). The Fissured Workplace: Why Work Became So Bad for So

Many and What Can Be Done to Improve It. Harvard University Press

https://www.jstor.org/stable/j.ctt6wppdw

data SourceS

Instituto Mexicano Del Seguro Social [Mexican Social Security Institute] -IMSS-

. (2021). Puestos de trabajo registrados por los patrones [Job Positions

Registered by Employers] [Data]. Mexico: IMSS http://datos.imss.gob.mx/

Instituto Nacional De Estadística y Geografía [National Institute of Statistics]

-INEGI-. (2019). Censos Económicos [Economic Census] 2008-2018 [Data].

Mexico: INEGI https://inegi.org.mx/app/saic/

Instituto Nacional De Estadística y Geografía [National Institute of Statistics]

-INEGI-. (2021). Encuesta Nacional de Ocupación y Empleo [National

Labor Force Survey] (2008-2021) [Data]. Mexico: INEGI https://www.inegi.

org.mx/programas/enoe/15ymas/

40 Enrique Kato-Vidal · Paulina Hernández-Mendoza

table a1. eStimateS of Subcontracting rateS (cenSuS and adminiStrative data)

 Baseline  Preferred

 [1] [2] [3]

Time trend 

2008 (versus 2003) 10.214** 12.398** 7.326º

2013 14.617** 17.548** 5.485

2018 18.815** 21.572** 9.153

Firm size

Small ﬁrms (versus Microbusinesses) 6.882 -3.184 -1.045

Medium ﬁrms 21.319** 5.115 -32.438*

Large ﬁrms 29.478** 14.320** -4.502

Industry ﬁxed effects No Yes** Yes**

Covariates (% workers) 

Youth No 1.151* -1.638º

Women No -0.674 1.052

Temporary No 1.252* 2.748**

Temporary2No -0.015* -0.031**

Interactions 

Firm size * % Youth No No Yes**

Industry * % Women No No Yes**

Intercept -12.403* -14.655 -16.262

F (prob) 27.3 (0.000) 24.1 (0.000) 114.0 (0.000)

R MSE 15.129 15.06 14.7

Adjusted R-squared 0.54 0.54 0.57

AIC 1071.87 1074.05 1075.74

Note: The symbols º, * and ** denote signiﬁcance levels of 10%, 5% and 1%, respectively. Standard

errors are grouped by pair of industry x ﬁrm size. N = 128 (32 pairs and four census years).

annex

Subcontracting retreat: early eStimateS uSing adminiStrative data

reviSta de economía mundial 66, 2024, 25-42

table a2. eStimateS: obServed and predicted Subcontracting labor (adminiStrative data)

Subcontracting rate

(relative to paid employees)

2003 2008 2013 2018 Estimate Difference

Act. Est Act. Est Act. Est Act. Est 2021 2023 2018-

2021

2018-

2023

Total (%) 6.9 6.9 17.1 17.1 21.5 21.5 25.7 25.7 25.0 13.0 -0.7 -12.7

0 Agricultural 0 2.7 16.2 13.0 19.9 14.0 15.6 22.1 25.3 14.1 3.2 -8.0

1 Mining 10.2 10.0 16.6 17.7 24.7 23.7 30.6 30.8 27.9 13.6 -2.8 -17.2

3 Manufacturing 6 2.1 12.8 12.0 16.5 18.1 17.2 20.3 17.4 3.5 -2.9 -16.8

4 Construction 5.4 1.9 11.3 15.8 17.1 14.6 16.7 18.2 17.4 7.1 -0.9 -11.1

5 Electricity, gas & water 4.7 -- 3.2 4.6 3.3 20.3 33.4 24.0 26.2 16.1 2.1 -7.9

6 Trade 12.3 33.8 47.8 43.9 57.7 48.0 58.2 50.2 49.7 38.6 -0.5 -11.6

7 Transport & communication 6.6 4.3 11.3 13.0 15.2 14.7 15.2 16.3 12.1 -- -4.2 --

8 Services 9.9 4.7 17.7 16.9 17.6 18.7 18.8 23.7 24.0 15.1 0.3 -8.6

Firm size (%)

Microbusinesses 2 -- 3.5 4.7 4 6.2 4.1 9.1 10.5 -- 1.4 --

Small 5.7 -- 10.3 10.0 11.9 13.8 13.2 18.2 19.1 7.4 0.8 -10.8

Medium 9.9 14.6 23.6 24.5 30.3 27.7 35.0 32.1 28.8 16.9 -3.2 -15.1

Large 9.9 20.4 31 29.3 39.9 38.3 50.6 43.4 41.6 28.0 -1.9 -15.5

Note: ‘Act.’ and ‘Est.’ denote Actual and Estimates, respectively. ‘--‘ indicates a value non distinct from zero The estimated coefﬁcients are reported in Table

A1. Predicted values were calculated using margins and the option out-of-sample (StataCorp, 2017). The columns 2018-2021 and 2018-2023 reports the

percentage points (pp) of difference between estimates. Source: Economic census, administrative data and own estimates.

42 Enrique Kato-Vidal · Paulina Hernández-Mendoza

figure a1. covariateS effectS on Subcontracting rate (%)

Panel a)

Temporary workers

Panel b) Women

Panel c) Youth

Note: Selected effects are shown. These estimates were predicted using results from Table A1. The

effects were increasing or decreasing depending not only on the covariate but on the sector and ﬁrm

size. The x-axis extreme values represent the interquartile range.