Sharing Economic Rents with Workers? Evidence from Matched Employer-Employee Data in Vietnam^†

I. Introduction

There has been a resurgence of interest in understanding how economic rents are shared between workers and firms (Card et al., 2018; Bell et al., 2024). This has been partly driven by a growing body of evidence indicating the persistence of imperfect competition in the labour market, whereby firms play a significant role setting individual wages (Manning, 2021). Furthermore, greater accessibility to linked employer-employee datasets has contributed to a revival of empirical studies, making it possible to investigate rent sharing between workers and firms in greater depth (Guiso et al. 2005; Guertzgen 2009; Juhn et al., 2018).

We contribute to this literature by examining rent sharing by small and medium enterprises (SMEs) in a dynamic emerging economy, in this case Vietnam (McCaig and Pavcnik, 2018). In a companion paper, Yamashita and Ha (2025) examined rent sharing for firms engaged with global value chains (GVC). This paper expands the scope of this work by extending the coverage of data and the period. Our dataset provides a rare opportunity to investigate the extent of rent sharing among SMEs using a rich source of information pertaining to firms and employee attributes.

We estimate the degree of individual wage regression with value added per employee as an indicator of economic rent, together with firm-level and individual-level wage predictors. To account for the unobserved dimensions of individual wage variations, we apply panel firm fixed effects as well as worker fixed effects. While less desirable, we use lagged value added as the primary source in our identification strategy.

Our empirical analysis is guided by the following three intertwined theoretical perspectives (Blanchflower et al., 1996; Bell et al., 2024). The first is rent sharing mediated by the bargaining of a labour union (Breda, 2015). This has been tested in several countries, including Vietnam. However, in the complex system of the Vietnamese economy, a trade union was found to have limited bargaining power to capture part of economic rent (Bach et al. 2021). The second approach recognises market power in a monopsonistic model. This assumes that firms face an upward-sloping supply curve within which favourable demand shocks translate into a rise in rents, wages and employment (see recent work on this approach by Lamadon et al., 2022). The third class of theoretical frameworks discerns rent sharing through the lenses of incentive pay. In particular, a small body of evidence suggests that workers’ wages are somehow shielded from unfavourable fluctuations in demand shocks (Guiso et al., 2005; Juhn et al., 2018). Guiso et al. (2005) systematically tested the notion of the role of the firm as an insurance provider against risks using a matched employer-employee dataset. The key insight is that firms provide insurance to workers, offering them flat compensation against idiosyncratic shocks in economic rents. This suggests that firms do not necessarily engage in rent sharing, even in good times.

Our analysis suggests that wage elasticity is in the range of 0.08-0.10, which is aligned with other elasticity estimates between 0.05 and 0.20, as surveyed by Card et al. (2018). This finding suggests that small firms even with limited market power would share economic rents with employees. A further analysis reveals that rent sharing is more pronounced if a worker is related to owners or has a non-production job. We find no statistical difference in rent sharing between exporters and non-exporters. This represents a sharp contrast to existing studies that find pass-through from rents to wages in the case of exporters (Verhoogen, 2008; Bustos, 2011; Frías et al., 2022; Krishna et al., 2014; Macis and Schivardi, 2016). The differences in the methods and the working sample (ours is SMEs, whereas the others focus on larger exporting firms) are attributable to this contrast in findings. More importantly, once we introduce worker fixed effects, which absorb time-invariant characteristics (such as unobserved innate skills and abilities), rent sharing is no longer observed. Corroborating with the above findings, rent sharing in our analysis is largely driven by workers’ (observed and unobserved) attributes. Hence, we cannot robustly claim that our findings are causal. Nevertheless, our results highlight an important heterogeneous aspect of rent sharing and offer new evidence that small firms with limited market power share economic rents even in the context of a less robust institutional environment to support workers.

Finding rent sharing for small firms offers a contrasting view that wages are given because firms are too small to set wages. A positive shock in productivity or demand is conventionally assumed to be absorbed via an employment adjustment rather than reflected in a wage adjustment in a competitive market (Bach et al., 2021). A further analysis reveals that this observed rent sharing is largely driven by workers’ attributes (in particular, the relationships among owners and those with non-production jobs). This defines the importance of modelling a richer dimension of workers’ attributes when determining the degree of rent sharing.

II. Empirical Strategy

Our empirical framework follows studies on conventional rent sharing by regressing individual workers’ wages on worker and firm attributes using matched employer-employee data (Guiso et al., 2005; Macis and Schivardi, 2016; Juhn et al., 2018; Verhoogen, 2008; Bell et al., 2024). We consider the following wage regression:

where the subscripts denote the following: i for workers, j for firms, and t for year. Both year and firm fixed effects (δ_t, δ_j) are controlled.

In this formulation, the variable of interest R is economic rent, represented by value added (VA) per employee. Our interest is to estimate wage elasticity with respect to changes in economic rent, R . Conceptually, value added is an ideal measure rather than profit because it represents the exact rent available after subtracting costs from revenues (Juhn et al., 2018).

A vector of employee attributes (gender, age, job tenures, level of education and on-the-job training) is subsumed in X . The inclusion of a conventional set of demographic and educational variables for workers controls the return of human capital investment in the wage equation.

The immediate concern in Equation (1) is the endogeneity issue between wages (w) and rents (R). The attainment of an effective instrumental variable (IV) remains a contentious issue in the literature, especially in the context of the employer-employee datasets (Card et al., 2018; Bell et al., 2024). In the ideal context, one should be able to identify firm-specific shocks that translate into wages mediated by rents. However, such a valid instrument has seldom been forthcoming. One promising approach in this endeavour is to use the economic value of innovations (e.g., captured by the excess return in the stock market following a patent grant) as a rent generation instrument (Van Reenen, 1996). While effective, this method by design cannot be implemented in the current study, as the data only cover SMEs, and none of them are listed on the stock market.

Against this backdrop, we proceeded to estimate Equation (1) under the assumption that shocks take time to be reflected on economic rent impacting wages, as in the case of multiyear contracts. In practice, we use the lagged value of value added (VA) to exert causal effects, influencing the current wages (to be explained further in the data section below).

III. SME data1

The data used for this study were extracted from the Survey of Small and Medium Enterprises (SME) of manufacturing industries in Vietnam, as jointly conducted and administered by the Central Institute for Economic Management, the University of Copenhagen and the United Nations University World Institute for Development Economics Research (WIDER) as a biennial interval survey, starting in 2005 and ending in 2015.2

SME data have two modules. The first is the enterprise survey, containing information pertaining to firms’ characteristics, including sector affiliation, year of creation, the average number of employees, sales, capital investment, wage bills, and the value of exports and imports. This module is panel data with a unique firm identifier. Each wave of the survey covered approximately 2,500 SMEs in ten provinces, which are spread across three regions of Vietnam, i.e., the north (Ha Noi, Ha Tay, Phu Tho, and Hai Phong), south (Ho Chi Minh, Long An, and Khanh Hoa) and central (Nghe An, Quang Nam, and Lam Dong) regions. The surveyed enterprises include households, informal firms, private firms, co-operatives, and limited liability firms, which are represented in each province (Trifković, 2017). Regarding the sampling design, a representative sample of registered household and non-household firms in manufacturing was drawn from the 2002 Establishment Census and the Industrial Survey 2004-2006 of the General Statistics Office of Vietnam (GSO) under a stratified sampling procedure.

The second module constitutes the employee survey from around 400-500 firms, randomly selected from the enterprise module.3 This employee module includes information about each worker’s attributes, consisting of weekly or monthly wages, annual gross earnings (including irregular payments such as overtime, shift work, and bonuses), gender, marital status, educational attainment, prior experience, tenure in the current enterprise, and the category of six different occupations (managers, professionals, service, office, sales, and production workers). This information is available for between one and seven employees.

This study uses both enterprise and employee modules from the survey years of 2007, 2009, 2011, 2013, and 2015. Several notes are in order in the data processing stage. First, the main wage data at an individual level consist of monthly wage recorded at the time of the survey year (usually in June). The survey data also provide working hours. Hence, in principle, it is possible to construct hourly wages. However, there is little variation in the monthly working hours reported in the original survey data. We decided to use the monthly wage rates. It should also be noted that individual wage data correspond to the survey year (s), while firm-level economic accounts (including value added) correspond to the mean values of s-2 and s-1. For example, the 2015 survey included (monthly) individual wages corresponding to the survey year (i.e., 2015) and economic accounts corresponding to 2014 and 2013. For simplicity, we denote the lagged firm-level economic account variables as t-1. Third, the employee data can only be a panel for the adjacent surveys (i.e., for the periods 2011-2013 and 2013-2015).

A. Descriptive Statistics

Table 1 details the sample size with column (1) indicating the number of firms, which has consistently been around 400 since the 2007 survey. Column (2) lists the number of employees, which serves as the sample size for our individual wage regression estimations.

TABLE 1

NUMBER OF FIRMS AND EMPLOYEES IN THE MATCHED EMPLOYER-EMPLOYEE DATA

Notes: Based on SME data.

Table 2 illustrates the distribution of firms by 2-digit industry. Notably, fabricated metal, food, and wood product industries comprise approximately half of the total number of firms. This concentration across industries illustrates that the sampled firms usually operate in relatively more labour-intensive industries.

TABLE 2

NUMBER OF FIRMS ACCORDING TO THE 2-DIGIT VIETNAM STANDARD INDUSTRY CLASSIFICATION

Notes: Based on SME data, Vietnam.

Table 3 presents the basic statistics of the firm-level variables. We document the summary statistics according to four different firm types, referred to here as all, family, exporter, and non-exporter types. Family firms operate as household businesses, presumably primarily under family member control. Exporting firms are defined if firms record any positive export sales at least one year during the sample period, and non-exporting firms are all other firms without positive export sales.

TABLE 3

SUMMARY STATISTICS FIRM-LEVEL VARIABLES

Notes: Based on SME data. Size refers to the number of (full-time) employees. Profit is profit per employee (million dong). Revenue is total revenue per employee (million dong). Remuneration is total wage bills per employee (million Dong). Value added is value added per employee (million Dong). All accounting values are converted into their corresponding real value according to the 2005 CPI as the base year. All refers to the full sample; family is ownership identified as a household business; exporter refers to an exporting firm (if positive export sales are recorded for at least one survey year); and non-exporter refers to all other firms not classified as exporters.

The average size (measured by the number of employees) is 20 (with a median of 8). In the World Bank’s definition, this sample consists of small firms with to 50 employees, whereas microenterprises are firms with up to ten employees and medium enterprises have up to 300 employees.

As expected, family firms are on average the smallest group of firms with average employment of six workers. Exporting firms, in contrast, have average employment of 66. At the same time, if one considers value added per employee as an indicator of firm-level productivity, exporters are the most efficient group of firms, as documented well in the literature (Bernard and Jensen, 1999).

Table 4 likewise splits the sample according to firm type and shows employee-level variables for each. Individual (monthly) wages in family firms are relatively low compared to those in other firm types. Perhaps this is a reflection of the low-wage industries (often, informal sectors) in which family firms operate (Steer and Sen, 2010). Wages for exporter firms stand as the highest in comparison. This is partly explained by having a relatively more educated and presumably more skilled workforce (on average, 74% of workers at least with a high school diploma). This contrasts with the average 42% of workers with a high school diploma in family firms.

TABLE 4

SUMMARY STATISTICS, EMPLOYEE-LEVEL VARIABLES

Notes: Based on SME data. Gender is a dummy taking a value of one if the worker is male. Education is a dummy taking a value of one if a worker has at least a high school diploma. Age is the age of a worker. Training is a dummy variable that takes a value of one if on-the-job training is provided. Tenure is the length of service in years at the current firm. Wage (in 1,000 Vietnamese Dong) is the real monthly wage recorded in June.

IV. Results

A. Baseline Results

Table 5 presents the benchmark results. We only report the variable of value added per employee (a proxy for firm-level economic rent). In column (1), the wage equation is estimated only with value added. In column (2), the model includes both firm controls (the number of employees and the capital and labour ratio) as well as worker-level controls (gender dummy, education dummy, age, job training dummy, job tenure, and real (monthly) wage rate, together with year and firm fixed effects.

TABLE 5

ESTIMATION OF THE RENT-SHARING MODEL FOR THE SURVEY PERIOD 2007-2015

Notes: All models are estimated by ordinary least squares with standard errors clustered according to the firm type in parentheses. The dependent variable is the log of the real wage rate at the individual worker level. Value Added (VA) is the real value added per employee with a one-period lag (t-1) and a two period-lag as t-2. VA^a is defined by subtracting total wage bills from value added. Firm controls include relevant firm-level variables, the log of the number of employees and the log of capital and labour ratio. Employee controls include worker-level variables, in this case gender, education age, training and tenure (see Table 4 for definitions of these variables). Year and firm fixed effects are included in all regressions. *** denotes 1% significance; ** denotes 5% significance.

The estimated coefficient for rent sharing in column (1) is 0.082, showing statistical significance. This implies that on average a 10% increase in economic rent (value added per employee) is associated with a 0.8% increase in the wage rate of an individual employee. This benchmark estimate conforms within the estimates of other studies in the range of 0.05 and 0.20 with regard to the rent-sharing elasticity found in the literature (Card et al., 2018).

The inclusion of firm and employee attributes increases wage elasticity to 0.1 (column 2). In the subsequent regressions, we include a two-period lag of value added without (column 3) and with firm and workers controls (column 4). While further lags are statistically insignificant, a one-period lagged value added sustains as a positive sign with statistical significance.

In columns (5) and (6), we redefined value added because it typically includes total wage bills. This could exacerbate endogeneity by creating linkages between value added and residual wage in Equation (1). To mitigate this concern, we subtract total wage bills from value added and use it as a rent variable. As expected, the redefined value added shows lower wage elasticity (0.03 in column 5 and 0.04 in column 6). However, the estimates preserve statistical significance with the standard test.

Overall, the benchmark estimates in Table 5 show that there is evidence of rent sharing among small firms after controlling for firm and worker-level controls. In the further analysis below, we explore several dimensions of heterogeneity in firms’ and workers’ attributes.

B. Heterogeneity

In Table 6, we provide the rent-sharing estimate by introducing a dummy variable to reflect a set consisting of firm and worker heterogeneity. An interaction term between rents and various indicators of firm ownership and worker attributes should indicate heterogeneity in rent sharing. In all regressions, we have included firm and worker control variables together with firm and year fixed effects.

TABLE 6

RENT SHARING BY TYPE OF ENTERPRISE AND WORKER FOR THE PERIOD OF 2007-2015

Notes: All models are estimated by ordinary least squares with standard errors clustered according to the firm type in parentheses. The dependent variable is the log of real wage rates at the individual worker level; Value Added (VA) is the real value added per employee with a one-period lag (t-1). Firm controls include relevant firm-level variables, the log of the number of employees and the log of capital and labour ratio; Employee controls include worker-level variables, specifically gender, education age, training and tenure (see Table 4 for definitions of these variable). Family firms represent the legal status of these firms as household establishments/businesses. Corporate firms include private firms (sole proprietorships), partnerships, collectives/cooperatives, limited liability companies, joint-stock companies with state capital, joint-stock companies without state capital, joint ventures with foreign capital, state enterprises (central), and state enterprises (local). Related to owners is a dummy variable coded as one if a worker is an owner, related to an owner living in the same household, or related to an owner but not living in the same household. Exporter refers to an exporting firm (if positive export sales are recorded for at least one survey year). Production worker is a dummy variable coded as one if a worker is engaged with production work. Contract is a dummy variable coded as one if a worker has a formal wage contract signed. Union is a dummy variable coded as one if a worker is a member of a trade union. Year and firm fixed effects are included in all regressions. *** denotes 1% significance; ** denotes 5% significance.

Most notably, in column (3), the results indicate that if a worker is related to the owner (i.e., is a family member), wage elasticity is greater compared to those of workers without such family connections. Similarly, rent sharing appears to be prevalent depending on the type of job. The estimates in column (5) suggest that wage elasticity is greater for non-production workers (such as sales, services and managerial jobs) than for production workers.

While our analysis reveals that exporters pay higher wage levels on average (Table 2), the responsiveness of wages to changes in firm performance (wage elasticity) is statistically indistinguishable between exporting and non-exporting firms (column 4). This finding diverges from previous studies that have demonstrated strong rent-sharing by exporting firms. These studies typically leverage quasi-natural experiments, such as plausibly exogenous shocks in the form of, for instance, exchange rate devaluations and trade liberalisation (Verhoogen, 2008; Bustos, 2011; Frías et al., 2022; Krishna et al., 2014; Macis and Schivardi 2016). Furthermore, our research focuses on small and medium-sized enterprises (SMEs), whereas the cited studies predominantly examine larger exporting firms (e.g., those with at least 50 employees), which may be more productive and skill-intensive. Such larger exporters may be in a better position to benefit from positive demand shocks. These methodological and sample differences may limit our ability to detect significant rent-sharing effects.

In columns (6) and (7), we have introduced an interaction term with the role of worker unions and the presence of formal wage contracts. It could be the case that having a formal contractual agreement between employers and employees limits any further fluctuations in wages and economic rent. This may restrict any deviation from the set wage negotiated between employees and employers. Specifically, the employee module of the SME survey asks if workers receive any formal contracts in a yes-or-no binary survey question.4 We also split the sample according to the coverage of labour unions at the worker level (column 7). In neither case did we find that having a wage contract or union membership facilitates rent sharing with employees, possibly because having such mediation instruments could shield workers from idiosyncratic shocks while preserving wage stability. A lack of rent sharing for union members echoes the finding of Bach et al. (2021) in their study that used the same SME dataset used here. In any case, rent sharing occurs regardless of the institutional setting and whether wage bargaining is facilitated by trade unions or not (Card et al., 2018).

C. Stayer Sample

In the last segment of the analysis, we explore the panel structure of the employee dataset, allowing us to control for unobserved worker heterogeneity. By focusing on employees who are observed within the same firm over the survey period, we can reduce potential bias related to unobserved skill differences and innate abilities. Therefore, we limit our analysis to employees who appear in multiple surveys. However, as noted in the data section, one issue that arises when using the panel component of employee data is that we can only identify repeatedly surveyed employees if they appeared only in adjacent surveys (e.g., it is only possible to identify the panel of workers either in 2011-13 or 2013-15). In practice, wage regression (Equation 1) now includes worker-fixed effects, 𝛿_𝑖, which would absorb all time-invariant components among workers.

Our empirical results for the stayer sample are presented in Table 7, which is divided into Panel A (2011-13) and Panel B (2013-15). Column (1) reports the estimated wage regression, leveraging the panel structure of workers but omitting worker fixed effects. In Panel B, we find a statistically significant positive effect of rent sharing. However, when we include worker fixed effects, this effect disappears (column 2 of Panel B). This indicates that worker fixed effects account for the primary source of individual wage variation. The restricted sample of stayers, while closer to the ideal dataset for causal evidence, necessarily reduces the number of observations, potentially lowering the statistical power. Consequently, we refrain from strongly claiming a causal interpretation of our findings with regard to rent sharing.

TABLE 7

RENT SHARING FOR THE STAYER SAMPLE

Notes: The dependent variable is the log of the real wage rate at the individual worker level; Value Added (VA) is the real value added per employee with a one-period lag (t-1). Firm controls include relevant firm-level variables, the log of the number of employees and the log of capital and labour ratio. Employee controls include worker-level variables, specifically gender, education age, training and tenure (see Table 4 for definitions of these variables).

V. Conclusion

This paper examined the extent of rent sharing within the context of one of the most dynamic emerging economies, Vietnam. Empowering a matched employer-employee dataset, we estimated wage elasticity with respect to economic rents. The benchmark estimates indicate a rent-sharing coefficient of 0.08-0.1, falling within the range of the elasticities documented in prior studies (Card et al., 2018). Furthermore, our analysis revealed that the observed rent sharing is more pronounced if a worker is related to an owner or has a non-production occupation. While we do not show any statistical difference in rent sharing between exporters and non-exporters, our study suggests the presence of rent sharing even for small firms with limited market power. However, we caution against claiming our results are causal evidence because rent sharing is no longer observed as soon as we introduce worker fixed effects. Despite this limitation, our study demonstrates the presence of rent sharing by small firms in the context of weaker institutional settings. There is a need for further research on this topic, especially that which uses a richer dataset source capable of capturing several dimensions of worker and firm attributes and their effects on rent sharing.