Industry Switching in Developing Countries

Abstract

Disentangling and defining the contribution of firm turnover to sector-level productivity growth is an important challenge in development economics research¹ that is also relevant to policy making. This primary motivation for this study was reinforced when we discovered that some contributions to the literature (e.g., Aw et al. 2001) categorize all firms leaving a particular sector as exit firms even if these firms do not actually close but instead switch to production in a different sector. Similarly, new entrants in a given sector regularly include both genuinely new firms and firms that previously operated in other sectors. Some studies of firm dynamics (e.g., Bernard et al. 2006) have followed establishments over time, treating sector switchers as incumbents. The classification of firms that change their main sector of production potentially affects the understanding of the reallocation processes that enhance productivity.

This thesis is substantiated by the literature on firm capabilities, including Sutton (2005) and Bernard et al. (2010), which suggests that the average productivity of switchers is likely to differ from that of real entrants/exits if the switching firms have underlying capabilities (e.g., know-how and working practices held within the firm setup) that affect the firm's level of productivity. Switchers have already incurred sunk costs and gained knowledge related to their capabilities when they initially established production. For a given postentry/switching productivity draw (in the spirit of learning models on the evolution of industry), this situation suggests that the average productivity level of switchers exceeds that of new entrants. If switching firms exhibit specific characteristics and are motivated differently than real entrants/exits, then these dynamics must be understood when assessing changes in economic policy or the external environment to determine their effect on productivity.²

Accordingly, rather than examining the drivers of productivity growth within firms, this study describes the dispersion of productivity across manufacturing firms in Vietnam (using census data for the period 2001–2008), with a special focus on disentangling the contribution of firm turnover to sector-level productivity growth. An industry switcher is defined as a firm whose main year t product and main year t − 1 product belong to different four-digit industries. Although this definition may classify marginal changes in the product mix as industry switching, the data show that most industry switchers are one-product firms. Consequently, this characterization of industry switching corresponds to extreme changes in production.

This study aims to analyze the processes by which firms decide to enter, exit, or switch sectors, to make predictions about the expected relative productivity differences between firms, and to establish the contribution to productivity growth of firm turnover and sector switching (as distinguished from the “real” entry and exit of firms). Given that one of our core objectives is to diagnose the extent to which differences exist between entry, exit, and switching firms, we consider the association between firm turnover and the observed characteristics of firms and sectors as well as the association with exogenous shocks, such as regulatory changes and trade liberalization.³ Finally, factors correlated with switching behavior are explored empirically and compared to factors related to entry and exit in Vietnam.

Vietnam is a populous Southeast Asian economy that has experienced rapid economic growth since 1986, when the comprehensive economic reform process known as Doi Moi was initiated. During the past decade, Vietnam has been one of the fastest-growing economies in the world, with GDP growing at an annual rate of 7.3 percent from 2000 to 2009 (World Bank 2011). The Doi Moi process has included wide-ranging reforms of enterprise, commercial, and investment laws, especially since 2000. These reforms have been coupled with extensive trade liberalization, accession to the WTO in 2007, and significant inflows of foreign direct investment. These features offer further motivation for understanding industry evolution in a dynamic transition economy.

Section I below defines sector switching and specifies three testable hypotheses. Section II presents the empirical approach, and section III describes the data. Empirical results follow in section IV, and section V concludes that correctly assessing the respective contributions of entry, exit, and switching firms to productivity growth has important analytical and policy implications.

I. Defining Industry Switching

Industry switching can be defined at various levels of sector aggregation. Switching is defined in this study at the four-digit International Standard Industrial Classification (ISIC) level, and the conceptual differences between switchers, entrants, exits, and incumbents are illustrated in table 1.⁴ Differences in the classification of firms are considered from the perspective of a firm that manufactures furniture (ISIC 3610) and switches in or out of sawmilling and planing of wood (ISIC 2010). In previous studies, switchers-in (captured in row 3) were classified as either new entrants (row 1), as in Aw et al. (2001), or as incumbents (row 5), as in Eslava et al. (2004). Similarly, switchers-out (captured in row 4) have been labeled exits (row 2) or incumbents (row 5).⁵ Next, we focus on the differences in the way the literature has treated sector switchers.⁶

The above definition of industry switching has implications for multiproduct firms. The data contain information on whether firms produce more than one product (six-digit) but do not specify what products these firms produce. Thus, a multiproduct firm producing two products will be regarded as a sector switcher when its percentage rates of production marginally change from 49 percent of product 1 and 51 percent of product 2 to 51 percent of product 1 and 49 percent of product 2. Therefore, producers of more than one product are controlled for in the empirical analysis, and multiproduct firms are excluded as a robustness check. The data clearly show that switching in Vietnam is not limited to changes in the product mix. In fact, the majority of switching, as defined in this study, is associated with single-product firms.

Several empirical papers in the literature (see Bartelsman and Doms (2000) and Syverson (2011) for reviews) have shown that enterprise turnover contributes significantly to sector-level productivity growth, and various models seek to explain why productivity should enhance reallocation. Vintage models of industry dynamics suggest that, on average, new entrants are more productive because their productivity is enhanced by the latest technology. It is assumed that firm productivity remains constant over time unless it is affected by a random shock. Therefore, it is expected that the productivity of entrants will dominate that of incumbents and switchers, and firm exit occurs when productivity relative to entrants drops below a certain threshold. Eslava et al. (2004) and Foster et al. (2008) provide recent empirical evidence supporting these vintage model predictions.

Learning models on the evolution of industry yield similar insights regarding the mean productivity differences between entrants and exits, but they differ from vintage models with respect to the expected relative productivity between incumbents and entrants. For a given distribution function summarizing the heterogeneity in productivity among firms that are strictly decreasing in current productivity, Hopenhayn (1992) shows that the productivity distribution of incumbents stochastically dominates the productivity distribution of entrants. Learning models have received empirical support in several papers, including Bartelsman and Doms (2000).

The seminal literature on the evolution of industry and industry dynamics does not consider the possibility that firms may choose to reallocate resources into new sectors to maximize profits or to avoid exiting the industry. At first glance, the importance of this possibility may not be clear. Changes to the main area of production may require a firm to undergo a learning period to gain knowledge of new production processes, and this change may be so fundamental that it may be similar to closing the business and opening a new firm in another sector. However, recent literature on the importance of firm/owner entrepreneurial capabilities (see Sutton [2005] for initial thoughts along these lines) suggests that productivity differences between switchers and entrants/exits should be expected. Given that switching firms have already incurred sunk costs when they initially established production (and have acquired, in the process, knowledge of different business procedures, access to public utilities and general market conditions), the sunk costs of switching sectors are arguably lower than the sunk costs facing new firms entering a sector.⁷ Moreover, the variance of the postswitch productivity draw is expected to be smaller than that of entrants into the same sector due to the business experience gained from previous production (i.e., knowledge of the firm's underlying capabilities).

The above considerations lead to the following three testable hypotheses regarding productivity differences between entrants, exits, and incumbents, on the one hand, and switchers, on the other:

• Switchers-in versus Entrants

  The aggregate sunk costs for switching firms will be below those of entrants. Switchers-in will have greater knowledge of both their productivity potential (entrepreneurial capabilities) and general market conditions and will therefore be likely to have a higher level of productivity than entrants.

• Switchers-out versus Exits

  Switchers-out will be more productive than exits. Switchers-out would also have exited if their expectations about future profitability (which depends on productivity) were equal to or less than that of exit firms. Switching depends on the observed characteristics of one sector relative to those of other sectors because expectations about future market conditions play an important role in the decision to switch and/or exit.

• Switchers versus Incumbents

  Switchers will be less productive than incumbents. They face a sunk cost of switching, and their information regarding their productivity and existing market conditions is related to their former sector rather than the new sector.

II. Empirical Approach

This section first outlines the method used to measure productivity and then describes the empirical model detailing the firm-, sector-, and industry-specific factors included in the analysis.

Productivity Measurement

Following Aw et al. (2001), we used an index number approach to estimate total factor productivity for firms in each manufacturing subsector.⁸ Productivity is measured relative to the mean level of productivity in a given sector and year. To analyze productivity over time, this productivity differential is linked to changes in the reference levels of productivity from year to year.

The sector-specific total factor productivity (TFP) index is given in equation (1). The productivity of a firm is compared, in any given time period, relative to the average productivity of the sector.⁹

(1)

where Y_ijt measures the output of firm i in sector j in year t, X_mjit is the amount of input m used by the firm, and s_mjit is the expenditure of the firm on input m as a share of the total expenditure.

Firm-specific productivity scores are used to compute a measure of productivity for each subsector in each year in equation (2).

(2)

where ⁠.

Equation (3) shows how this weighted productivity measure can be decomposed into the average unweighted productivity level of each subsector and a term that captures how the allocation of resources contributes to productivity (Olley and Pakes 1996).

(3)

This measure is used to consider the proportional contribution of switchers to aggregate productivity growth and to compare productivity between incumbents, entrants, exits, and switchers.

Empirical Model

Productivity is critical to firm decision making, but many other firm characteristics and sector-specific characteristics can be expected to play a role in switching behavior. Equations (4a) and (4b) show the empirical models of the decisions to switch into a sector and out of a sector, respectively.

(4a)

(4b)

where ω_it is the productivity of firm i in period t, k_it /l_it is the capital-labor ratio, log(l_it) is the log number of employees, SOE_it is a dummy indicator for whether the firm is state owned, FOE_it is a dummy indicator for whether the firm is foreign owned, and multiprod_it, is a dummy indicator for whether the firm produces more than one product. The subscript j in equation (4a) refers to the new sector entered, and the subscript m in equation (4b) refers to the old sector exited. For example, ω_jt is the average productivity level in sector j that the firm switches into in period t, and ω_mt is the average productivity level in sector m that the firm switches out of in period t. Other sector-specific variables include the following: k_jt /l_jt refers to the average capital-labor ratio; log(l_jt) is the log average size of firms; CR_jt is the concentration ratio; FR_jt is the concentration of foreign-owned firms; SR_jt is the concentration of state-owned firms; TR_jt is the tariff rate; and EX_jt is the level of exports. Additionally, γ_i represents firm-specific fixed effects, θ_j and θ_m are sector-specific fixed effects, τ_t represents time dummies, and ϑ_p represents province fixed effects.

The decisions to enter, exit, or switch sectors are modeled in a similar fashion. For entry, as in equation (4a), the characteristics of the new sector are included, and for exit, as in equation (4b), the characteristics of the former sector are used. The inclusion of firm-specific fixed effects eliminates any time invariant unobserved heterogeneity influencing firms' decisions. As such, the identification of the effect of the firm-specific effects included in the model comes from the within-firm variation in firm characteristics. It is possible that some time-varying factors remain unobserved, so the (i, t) indexed variables are endogenous to firm behavior and therefore to the entry, exit, and switching outcomes. Consequently, no claim is made that results are causal.¹⁰ Before presenting the data in section III, the motivation for the choice of firm-, sector-, and industry-specific variables included in the model is explained along with expectations regarding their influence.

Firm-Specific Factors

First, a firm's productivity level is a critical factor in decisions regarding whether to stay in production, switch sectors, or exit production altogether.¹¹ Firms evaluate their expected current and future profits on the basis of their own observed productivity level.

Second, capital accumulation is a key mechanism for increasing profitability. Thus, capital-intensive firms should increase profitability over time and should be less likely to exit and more likely to enter as an industry evolves. Bernard et al. (2006) have found that a firm's capital-labor ratio is an important determinant of decisions to switch sectors in the United States. In Vietnam, where labor-intensive firms arguably have a comparative advantage, labor-intensive firms are unlikely to have a higher probability of exit.

Third, the link between firm size and probability of survival has long been considered important.¹² This link may affect a firm's decision to switch sectors because larger firms may find it more difficult to retrain employees. To capture the effect of this factor, firm size, measured by number of employees, is included.

Fourth, arguably, ownership structure influences firm decision making in Vietnam, even after firm productivity, capital accumulation, and size are controlled for. The political hierarchy in the management structures of Vietnamese state-owned enterprises (SOEs) is likely to limit intersector dynamics; consequently, SOEs are unlikely to switch sectors. In contrast, the ongoing reform/privatization process has led to the dismantling of many SOEs, so one might expect a positive association between state ownership and firm closure. Foreign-owned enterprises (FOEs), or enterprises with some foreign participation, are also expected to be more “locked into” specific sectors because of legal constraints.¹³

Fifth, Bernard et al. (2010) and Goldberg et al. (2010) explain reallocations that enhance productivity through changes in product mix. Thus, multiproduct firms will be in a better position to churn products (i.e., dropping inefficient products to produce more efficient products) in response to changes in the economic environment and thus may be less likely to switch sectors.

Sector-Specific Factors

We will now consider a range of sector-specific measures for establishing the competitiveness of a sector relative to other sectors in the industry.

First, higher levels of average productivity in a sector make it more difficult for firms to compete. Because of difficulties in comparing productivity across sectors, we also consider the dispersion in productivity distribution a unit-neutral measure of productivity. It is easier to survive in sectors with a larger dispersion in productivity because low productivity levels are more likely to be tolerated. Accordingly, it is expected that firms are less likely to leave, and more likely to enter, sectors with a wider dispersion of productivity distribution.

Second, the average capital intensity of the sector, as measured by the capital-labor ratio, may be influential in determining a sector's competitiveness. Audretsch (1991) found that firm survival is much less likely when there is a high capital-labor ratio, but Bernard et al. (2006) found that firms in labor-intensive sectors switched into more capital-intensive sectors when exposed to competitive pressures from imports.

Third, the average size of firms within a sector may affect switching decisions. It is more difficult for entering and switching firms to compete in a sector where the size of the average firm is large, given the economies of scale already enjoyed by incumbents. Consequently, it is expected that firms are more likely to enter or switch into sectors where firms are smaller, on average.

Fourth, the sector concentration ratio (CR), defined as the ratio of the accumulated revenue of a sector's four largest firms to total revenue in the sector, is likely to be influential because it is a proxy for sector competition. Siegfried and Evans (1994) document that a high CR may strengthen collusion efforts among incumbent firms and increase the likelihood that firms will attempt to prevent entry and maintain higher expected profits. In contrast, Audretsch (1991) shows that a high CR helps the survival rates of new entrants in the short run. On balance, it is likely that a high CR reduces incentives to move out of a given sector and is indicative of barriers to entry.

Industry-Specific Factors

Changes in exogenous conditions may “shock” enterprises, leading to different productivity outcomes. Trade liberalization, in the form of a tariff reduction, may lead to low-productivity firms exiting sectors or switching to sectors that remain protected. Because trade liberalization may lead to opportunities in new export markets, however, it is hypothesized that more productive firms will switch to sectors where these opportunities emerge.

In the manufacturing industry in Vietnam, it is critical to explore the impact of the privatization of SOEs and the entry of FOEs into the market. Where an entire industry is undergoing deregulation, a significant amount of productivity-enhancing reallocation is likely to take place. The dominance of state enterprises (SR), as measured by the share of total sector output controlled by SOEs, is likely to play a role in exit and switching decisions. Preferential treatment of SOEs makes it difficult for non-SOEs to compete and may force efficient non-SOE firms to exit (or to decide not to enter) industries with high concentrations of SOEs.

In contrast, during the transition from a planned to a market economy, the SOE share of material inputs bought at market conditions may increase the attractiveness of industries with high concentrations of SOEs for smaller, private enterprises acting as producers of intermediates for SOEs, as suggested by Jefferson and Rawski (1994) in their study of China. An added dimension is that deregulating SOEs may increase competition as a result of the decline in the level of protection and barriers to entry, thereby inducing firms to switch.

Similar arguments apply when considering the dominance of foreign enterprises (FR) in a sector, or the FOE share of total sector output. Aitkin and Harrison (1999) emphasize that preferential treatment of foreign-owned firms may distort competition, forcing out equally efficient domestically owned counterparts. However, governments grant special treatment to FOEs to promote technology transfer, so FR may create a basis for domestically owned firms to produce intermediate inputs, as is the case with SOEs. Whether FR is positively or negatively related to sector switching and firm exit depends on which of these contrasting effects dominate.

III. Data

The data originate from the 2001–2008 Enterprise Surveys collected annually by the General Statistics Office of Vietnam. These surveys include all enterprises with 30 employees or more registered in Vietnam, a representative sample of smaller firms,¹⁴ and all firms whose main activity is in the manufacturing sector.¹⁵ Therefore, the sample includes approximately 50,000 firms.

The extent of diversification within the manufacturing sector is substantial, as illustrated by the number of four-digit subsectors within each two-digit sector (table 2). There are no data available to examine product diversification at the six-digit level, as in Bernard et al. (2010) and Goldberg et al. (2010), but we know that between 5 and 19 percent of manufacturing firms produce more than one product. For these firms, diversification across a product mix is a potential source of productivity growth that cannot be captured in this analysis. However, changes in the product mix that make one activity relatively more important than another when defined at aggregated levels are captured by the definition of switching. The factors that determine the decision of multiproduct firms to change their product mix are likely to differ from those that determine the decision of single-product firms to completely switch sectors, making this an important control variable in the analysis.

The manufacturing sector in Vietnam is characterized by significant enterprise dynamics (table 2). This characterization is consistent with much of the existing literature on firm dynamics that finds a positive correlation between exit and entry rates at the sector level (Dunne et al. 1988, 1989; Disney et al. 2003; Roberts and Tybout 1996). Following Bernard et al. (2006), our focus here is exclusively on four-digit switching, which ranges from 6 to 35 percent.¹⁶ Moreover, the majority of this switching is attributable to single-product firms.

In comparison, Bernard et al. (2006) find that approximately 8 percent of U.S. manufacturing firms switched activities during five-year periods between 1977 and 1997. The fact that switching is common in Vietnam is indicative of an evolving industrial sector where new opportunities are emerging as a result of deregulation, trade liberalization, and on-going structural transformation. The extent of observed switching behavior adds further weight to the argument that the effects of switching on productivity should be separated from the effects of standard exit/entry. A better understanding of the forces responsible for switching may be helpful in designing effective economic policy in developing countries.

IV. Empirical Results

Our empirical results are grouped under the following four headings: productivity growth, switching, firm entry and exit, and robustness checks.

Productivity Growth

Productivity is first estimated for each subsector of the manufacturing industry. Output is the total revenue of the firm deflated by the two-digit sector-level GDP deflator. Inputs are composed of the following elements: (i) labor, measured as the total number of persons employed at the end of the year; (ii) capital, measured as the total assets of the firm at the end of the year deflated by a capital price series; and (iii) other costs of production deflated by the two-digit sector-level producer price index. The cost of labor is the firm's reported wage bill deflated by the producer price index, and the cost of capital is charged at the average annual commercial bank lending rate in each year plus an estimated depreciation rate of 2 percent per annum.¹⁷

Each sector's trend in productivity from 2001 to 2008 is computed using the index number approach outlined in section II.¹⁸ In all sectors in all years, the covariance between output and productivity is positive, indicating that more productive firms account for a larger share of output (table 3). The size of the covariance term does not change significantly over time, suggesting that the main source of productivity growth can be traced to changes in the productivity level of firms rather than to increases in the concentration of output in more productive firms.¹⁹ This does not mean, however, that reallocations of output within and across sectors do not contribute to productivity growth.

To determine whether switching is economically important, the proportional contribution of switchers to aggregate productivity in the sector that they switch out of is compared with their proportional contribution to the sector that they switch into (table 4). Moreover, productivity differences between incumbents, entrants, exits, and switchers are described in a manner that follows Aw et al. (2001) (table 5). This description facilitates the testing of the hypotheses outlined in section I.

Switching firms generally contribute significantly more to productivity in the sector that they switch into than in the sector that they left (table 4). This pattern holds for both weighted and unweighted TFP, with the only exception occurring between 2001 and 2002. Although the magnitude of the gains is not substantial, these gains are achieved in the year immediately following the switch, when firms have undoubtedly incurred some costs in switching activities.²⁰

A firm fixed-effects regression for switching firms was conducted for the year of the switch, where firm TFP is regressed on dummy indicators controlling for time, industry, province, and sector-level productivity. These results, which are not reported here, confirm the findings presented in table 4 and suggest that there is a break in a positive direction after the switch.

The comparison of productivity performance between switching and other firms is not straightforward. Relying on productivity levels is not appropriate because, in the construction of the productivity index, firms are positioned relative to the average within the sector, making comparisons across sectors impossible to interpret. The same situation arises when we attempt to interpret growth rates in productivity levels for switching firms.

To overcome this obstacle, the productivity performance of each firm is ranked within each sector in each year. Simple t-tests are then performed of the differences between incumbent, switching, entry, and exit firms in terms of their productivity ranking. Incumbent firms rank higher in the productivity distribution than entry and exit firms on both measures (table 5). Similarly, exit firms rank higher than entry firms. These results are not surprising because entry firms may incur high sunk costs that negatively affect their productivity performance in the year of entry.²¹

The results also show that switchers-in rank higher in terms of productivity than entry firms in the sector into which they enter, and switchers-out rank higher than exit firms in the sector that they switch from. This result suggests that switching firms are indeed a separate and important source of productivity growth, which is consistent with the predictions in section I above and with the emerging literature on firm capabilities, which we have referenced. Finally, incumbents have a higher productivity ranking than switchers in the sectors that they switch into and out of, which is also consistent with our expectations.²²

Given the difficulties in comparing productivity levels across sectors and, consequently, of comparing productivity in switching firms before and after they switch sectors, it is not possible to conclude definitively that switching firms become more productive as a result of moving from one sector to another. Two observations may be made, however. First, switching firms contribute more to overall productivity in the sector that they switch into than in the sector that they switch from, suggesting that switching leads to a reallocation of resources that enhances productivity (table 4). Second, switching firms have higher productivity levels and rank higher in the productivity distribution than entry and exit firms in the same sector (table 5).

Switching

The first set of switching results relates to the decisions of firms to switch into and out of particular sectors (table 6). The former identifies sector-specific pull factors because sector characteristics in the year after the switch are included. The latter identifies sector-specific push factors because the sector characteristics in the year before the switch are included. In other words, although we are not estimating a dynamic model, the key focus is to identify the push and pull factors through the timing of their inclusion in the model. Switching and sector-specific variables are defined at the four-digit industry classification level, and sector-level variables are computed separately for each firm i by excluding information on firm i in the computation of the sector-level aggregates. These definitions and computations ensure that the individual characteristics of the switchers do not drive the sector-specific effects.