Scholarly article on topic 'Growth of Incumbent Firms and Entrepreneurship in Vietnam'

Growth of Incumbent Firms and Entrepreneurship in Vietnam Academic research paper on "Social and economic geography"

0
0
Share paper
Academic journal
Growth and Change
OECD Field of science
Keywords
{""}

Academic research paper on topic "Growth of Incumbent Firms and Entrepreneurship in Vietnam"

growth and change

Growth and Change

Vol. 43 No. 4 (December 2012), pp. 638-666

Growth of Incumbent Firms and Entrepreneurship in Vietnam

ENRICO SANTARELLI AND HIEN THU TRAN

ABSTRACT This paper analyzes the relationship between the performance of incumbent firms and the net entry of new firms by combining various theoretical views of entrepreneurship. Different regression models to treat dynamics and endogeneity issues are applied to test the research hypothesis with regional micro-data for 61 Vietnamese provinces from 2000 to 2008. The main finding is that net entry is associated with the performance of incumbent firms and the overall performance of the economy. Incumbents' growth and gross domestic product growth induce changes in the existing production system and stimulate the creation of an economic environment more favourable to new firm formation. Consistent with the hypotheses put forward within the "knowledge spillover," the "error-correction," and other approaches, incumbents may generate new entrepreneurial opportunities not only for themselves but also for the whole society.

Introduction

This paper explores the relationship between performance of incumbent firms and entrepreneurship in Vietnam. From a theoretical viewpoint, an analytical integrated framework on the dynamic relationships among incumbent firms, entrepreneurship, and firm entry is set up, based on two main approaches emerged within the fields of entrepreneurship and development studies. From an empirical viewpoint, we use firm-level data from the General Statistics Office (GSO) of Vietnam dealing with 61 provinces over the 2000-2008 period.

Entrepreneurship studies have put forward two main theories for explaining the relationship between performance of incumbent firms and new firm formation: the "knowledge spillover" and the "error-correction" hypotheses.

According to the knowledge spillover theory of entrepreneurship (Acs et al. 2009), incumbent firms invest in activities leading to new knowledge creation with the purpose of generating innovations. However, knowledge and ideas generated by

Enrico Santarelli is a professor of Economics in the Department of Economics, University of Bologna, Bologna, Italy. His e-mail address is: enrico.santarelli@unibo.it. Hien Thu Tran is a lecturer of Entrepreneurship in the Centre of Commerce and Management, RMIT International University Vietnam, Hanoi, Vietnam. Her e-mail address is: hien.tran@rmit.edu.vn. The authors extend many thanks to the editor and anonymous referees for their excellent comments and suggestions.

Submitted July 2011; revised January 2012; accepted March 2012. © 2012 Wiley Periodicals, Inc

such investment are not fully encapsulated and captured by incumbents themselves, with a certain amount of knowledge that may leak outside and possibly be captured by outsiders. An even stronger relationship between incumbents and newcomers is found when dealing with "entry by spin-offs" (Helfat and Lieberman 2002; Klepper 2010; Klepper and Sleeper 2005). Spin-off firms inherit a sort of genetic imprinting from their parent firm, which shapes their subsequent pattern of behaviour. Thus, spin-offs have incentives to develop product variants that their parents would not find profitable (Helfat and Lieberman 2002).

The error-correction or "alertness and opportunity recognition" hypothesis (Kirzner 1997) argues that incumbent firms will commit errors at various rounds of the process of exploiting their profit opportunities, due to their knowledge limitations. Some errors may be recognized and corrected by incumbent firms themselves, but many other errors remain unknown to them and are therefore likely to become the source of opportunities for nascent entrepreneurs. Once new entrepreneurs pursue these opportunities, they contribute to the correction of errors generated by incumbent firms in the previous round. In this process, incumbent firms are not capable of preventing new entrepreneurs from recognizing and acting to correct their errors.

The paper is structured as follows: The second section discusses how the issues of firm growth and of firm entry and survival have been analyzed in the relevant literature. The third section introduces the main research hypotheses. The fourth section describes variables and econometric strategy. The fifth section reports the empirical results. The paper ends with some concluding remarks.

Incumbent Firms, Entrepreneurial Opportunities, and Entry

Penrose (1959: xii) depicts the firm as a collection of physical and human resources whose services are made productive by a "coherent administrative organisation." As long as resources are used productively, the firm will continue to grow and, therefore, accumulate resources. Additional accumulation of productive resources widens the firm's productive opportunities, by increasing the possibilities of deploying resources in more productive ways. However, although a firm's productive possibilities always expand with number and variety of available resources, the pool of its productive opportunities does not necessarily expand equivalently: context and uniqueness of the firm's administrative organisation limits its possibility to seize all productive opportunities.

The concept of "entrepreneurial opportunity" described by Shane (2003: 18) as "a situation in which a person can create a new means-ends framework for recombining resources that the entrepreneur believes will yield a profit" is a subset

of Penrose's "productive opportunities." As the firm continues to grow, it generates a larger number of productive resources that expand the available resource pool and, therefore, the firm's potential to recombine resources at higher values. However, because of the limited flexibility of resources, especially human capital, only a fraction of these possibilities are exploited by the firm. New business founders therefore compensate for this shortage of human capital by in turn performing the entrepreneurial tasks of discovering and exploiting business opportunities (Shane and Venkataraman 2000).

According to the "knowledge spillover theory of entrepreneurship" (Acs et al. 2009; Audretsch 1995; Audretsch, Keilbach, and Lehmann 2005), which can be characterized as "Schumpeterian" for its focus on sources and determinants of innovative entrepreneurship, the most important advantage of nascent entrepreneurs is that they do not need to invest into new knowledge as incumbent firms have to (Audretsch, Keilbach, and Lehmann 2005). Nascent entrepreneurs can enjoy the free lunch because of the appearance of the knowledge spillover within incumbent firms and other knowledge-creating institutions. Here, the knowledge spillover is the gap between new knowledge created by a given organisation and new knowledge commercially exploited by such organisation. Because of the basic conditions inherent in new knowledge, like high uncertainty, asymmetries, and transaction costs, the management team of incumbent firms has to leave away many new ideas which other individuals or agents evaluate as worth to pursue (ibid.: 75-76). As Acs et al. (2009) state, the divergence in valuation of knowledge across economic agents and within the decision-making process of incumbent firms can induce agents to start new firms as a mechanism to appropriate the (expected) value of their knowledge. Further, and consistent with Shane (2003), nascent entrepreneurs can freely exploit the technological and managerial experience and knowledge accumulated while working as paid employees and take advantage of their current customer and supplier linkages without initial investments as well. They have to decide when and how to enter the market, taking into account the current environmental conditions (Lévesque and Shepherd 2004).

The empirical evidence supports the knowledge spillover theory of entrepre-neurship in the case of advanced countries. In particular, Audretsch (1995) and Caves (1998) find that industries with a greater investment in new knowledge exhibit higher start-up rates and vice versa, whereas Audretsch and Feldman (1996) and Audretsch and Lehmann (2005) show that the greater the local knowledge stock in a region, the richer the pool of entrepreneurial opportunities and the higher the level of absorptive capacity for knowledge of that region. The case of a transition or developing country, such as Vietnam, is likely to be significantly different from that of a more or less advanced market economy. For this reason,

an additional explanation subsumable under the knowledge spillover theory but originally designed for some of the Asian and Latin-American's most successful economies during the last decades should be taken into account. This can be defined as the "imitation" hypothesis (Hausmann and Rodrik 2003), submitting that those entrepreneurs who make the "right investment decision," even in labour-intensive or natural-resource-based productions, can orient the investment of other entrepreneurs. This imitation effect is due to the fact that the initial entrepreneur making the "discovery" is able to appropriate only a fraction of the social value that this discovery generates, while signalling to other potential entrepreneurs the existence of opportunities. As a consequence, "learning what can be produced" is a type of entrepreneurship that may foster growth in transition/developing countries, even though the same hypothesis might be relevant also for advanced ones.1

As described by Klepper and Sleeper (2005) in their study on entry in the laser industry from its start through 1994, spin-off firms inherit knowledge from their parent firms that shapes their nature at birth: individuals switching from wage employment to entrepreneurship learn what and how they can produce by inheriting knowledge and abilities from the firm with which they were previously employed. Using a Hotelling-like model of sequential entry, Klepper and Sleeper allow spin-offs to enter product niches in which different variants of an industry's product are supplied to informed buyers. In turn, buyers are assumed to purchase one unit of the variant that maximizes their utility. Consistent with the law of one price, all sellers charge the same price for their variant of the product, so that buyers are not price constrained in their decision to purchase the variant closest to their tastes. Buyers' preferred points are uniformly distributed along the Hotelling line, and each seller chooses a point on the line corresponding to their product variant. Thus, each seller invests in the specific know-how needed to produce and market her variant, with the success of the investment conditioning the unit cost of production and the gross profit per unit of output. As a consequence, the most successful investment enables the firm to earn sufficient gross profit to repay its cost with the smallest market share. The resulting equilibrium for the model is one in which firms locate at points along the Hotelling line that are spaced equally apart. Findings by Klepper and Sleeper support the view of spin-offs being mainly tied to the experiences of incumbent producers rather than the prospects for new producers: As long as demand conditions are not unfavourable, they enter when their parents generate the information they need to exploit.

In identifying where new entrepreneurs come from, one has to take account the fact that, consistent with the occupational choice model of entrepreneurship, a large fraction of new business founders are individuals who switch from working for a wage to self-employment. As it is confirmed by well-established empirical

evidence, employees are significantly more likely to quit small rather than large firms to found new ventures. Johnson and Cathcart (1979) found that the propensity to be a new firm founder is higher for individuals previously employed in small firms, arguing that small firms' employees get a better preparation to start a new business. The transition into self-employment would be easier for employees of small firms because of the wide range of tasks accomplished during their job experience, the close relationship with the manager(s), and the knowledge of the types of market that could be served by a new business. This finding led other authors to conclude that small firms are the most effective incubators of new entrepreneurial ventures (for a survey, cf. Santarelli and Sterlacchini 1994). However, as recently shown by Parker (2009), one may assume that economic agents differ in terms of their attitudes to risk, with the less risk-averse individuals rationally selecting into entrepreneurship while the more risk averse work for them as employees (Kihlstrom and Laffont 1979).

An implication of Kihlstrom and Laffont's (1979) model is that the owners of large firms effectively "insure" their workers by offering them a smoothed wage. Here, a self-selection mechanism is at work, whereby individuals with particular characteristics are more likely to choose to work for small firms and to engage in entrepreneurship at different stages of their lives (Parker 2009). By following this line of reasoning, Parker suggests that while small firms may generate many benefits for the broader economy, they do not stimulate new venture creation through the active shaping of tomorrow's entrepreneurs. Nor do they provide such unattractive workplaces that their employees choose entrepreneurship as an escape route. Instead, it seems that people self-select into both entrepreneurship and employment in small firms at different stages of their lives. Empirical evidence for Britain (1991-2003) supports this hypothesis giving results that turn out to be robust to the inclusion of controls for measured and unobserved worker characteristics.

Main Hypotheses

The approaches summarized in the previous section have mostly focused on the actual mechanisms that transmit knowledge. For this reason, based on the above discussions we sketch out an integrated framework to indicate the dynamic interactions among endowment of entrepreneurial factors, opportunity seeking and exploiting, incumbent firm growth, and new firm formation.

When entrepreneurs of incumbent firms discover opportunities and want to realize them, they use accumulated resources to create new means-ends frameworks. The obvious outcome of this process is the growth of incumbent firms that is in turn the most easily observable measure of overall industry performance by would-be entrepreneurs, since most of them have a dependent job in the same

firms/industry. Besides, most new firms are very small and it is thus reasonable to assume that the potential founders of new firms will be particularly affected by the performance of existing small firms. As a consequence, accelerated growth in incumbent firms can be detected as a clear signal of industry growth and can foster the establishment of new small firms within the same industry. By observing incumbents' growth, nascent entrepreneurs perceive the availability of market opportunities, therefore deciding to start up. Further, the growth of incumbents increases the stock of productive resources available for the society as a whole, in this way relaxing the capital constraints faced by nascent entrepreneurs.

H1 The growth of incumbent firms located in the region generates localized externalities and spillover/ imitation effects.

Accelerated growth by incumbent firms implies that the number of firms having less excess profit tends to lower. As a consequence, the number of firms that have to leave the industry tends to decrease.2 In combination, net entry displays a strong positive correlation with the growth of incumbent firms in the same period. In this connection, the positive correlation between the performance of spin-offs and parent firms highlighted by Klepper (2010) suggests that some kind of learning mechanism is in operation, with better firms having more knowledge to be learned by their employees who are in turn likely to found spin-offs exploiting knowledge they learned from their parents.

H2 The rate of new firm formation in a region has a strong positive correlation with the growth of sales of incumbent firms in that region.

Variables and Econometric Strategy

Data description. We use panel regional-level data from the GSO for 61 provinces in Vietnam from 2000 to 2008. All firms reaching a certain size threshold or desiring to adopt a formal ownership form (partnership, limited liability, corporation, etc.) are requested to register into the National Enterprise Database that is managed and aggregated annually at provincial level by GSO.

Along with many merits, the database has some limitations. First, it does not contain data for the entrepreneurial activities of small households, who are not required to officially register. Second, and the most important, it does not allow singling out firm registrations and cancellations: Thus, we can only calculate the net number of new entries as difference between total number of firms included in a given year and the same value in previous year. Third, it does not deal with the issue of mergers and acquisitions. Finally, from 2004 new provinces were created through separations from the existing ones. This has increased the number of

Vietnamese provinces to 64. For simplification, in our analysis the values related to provinces that were founded after 2004 are added to the provinces from which they were separated.

During the period 2000-2008, net entry determined an increase in the total number of firms by about 16,000 per year, mostly in the private sector. Over the period 2000-2003, net entry increased slightly, from 9,392 in 2000 to 11,228 in 2003, and to 19,747 in 2004. The period 2005-2008 marked a sharp rise of new firms from 21,632 net entries in 2005 to 49,918 ones in 2008. In general, the increase in the number of firms in the period 2005-2008 was more than twice the one in the period 2000-2004.

Comparing the percentage share of enterprises by economic sector in 2000 and in 2008, one observes that the majority of additions occurred in the construction and real estate trading sectors. From the beginning to the end of the period, the share of enterprises increased from 3 to 11 percent in the construction industry and from 10 to 14 percent in real estate trading. In general, there has been an increase in the share of enterprises in the infrastructure service sector and a corresponding decrease in agriculture and forestry and in manufacturing.

Considering the number of enterprises per 1,000 persons in each province, it emerges that provinces in Red River and Mekong Delta are the preferred locations. Hanoi, the capital city, and Ho Chi Minh City, the biggest commercial and cultural city, have the highest firm density: On average, 6 firms per 1,000 persons, whereas mountainous and rural provinces such as Ha Giang, Son La, and Tuyen Quang are generally not the most favoured location choice for entrepreneurs. There is a large divergence between the share of firms per 1,000 inhabitants in the 6 densely populated and highly developed provinces (Khanh Hoa, Hai Phong, Binh Duong, Da Nang, Ho Chi Minh, and Hanoi) and the remaining provinces. The majority of provinces has less than 1 firm per 1,000 persons (55/61 provinces) whereas the other 6 provinces have from 2 to 6 firms per 1,000 persons.

In commenting these figures, one has to take into account that after 2000, with the adoption of the 1999 Enterprise Act many things changed in the institutional arrangements regulating free enterprise in Vietnam. These include the simplification of administrative procedures for starting a new firm, the elimination of minimum capital requirements, and the adoption of new forms of enterprises. One cannot neglect the impact that this deregulation process exerted on the overall phenomenon of new firm formation in the country. In fact, during the period 2000-2008, net entry determined an increase in the total number of firms by about 16,000 per year, mostly in the private sector.

Operationalisation of variables. As far as the dependent variable is concerned, we are aware that different measures of new firm entry may produce

strikingly different results in empirical analyses. In particular, two alternative approaches can be adopted to compare start-up rates across regional markets (Santarelli, Carree, and Verheul 2009): the ecological and the labour market approach.

The ecological approach standardizes the number of entrants relative to the number of firms in existence to investigate the amount of start-up activity relative to the size of the existing population of businesses. A measurement bias could occur due to regional heterogeneity in mean establishment size (MES), i.e., average number of employees per establishment that overstates start-up rates in regions where MES is higher and understates them in relatively low MES regions (see Audretsch and Fritsch 1994).

The labour market approach standardizes the number of new firms with respect to the size of the workforce. It extends the concept of entrepreneurial choice proposed by Evans and Jovanovic (1989), according to which all firms are the result of individual actions. Each person in the labour pool is considered as a nascent entrepreneur, and has the potentiality to set up his own business (Audretsch and Fritsch 1994).

We believe that the labour market approach is the most adequate way of analyzing start-up processes and much better suited than the ecological approach. As only data on the number of firms in existence at the end of each year are available, a net measure of the difference between the total number of new entries and the total number of existing firms will be used in our analysis, rather than the gross entry rate. Our measure proxies the success of regions in retaining new firms once they have been created, therefore capturing the potential long-term impact of new firms on the local economy (Hart and Gudgin 1994).

As our main independent variable, we adopt annual growth of revenues of incumbent firms in a region during the period 2000-2008 to study the relationship between their sales performance and the patterns of net entry in that region. Empirical evidence in advanced countries based on observation of the dynamics of both the survival rate and the hazard rate during the follow-up period (see, e.g., Audretsch, Santarelli, and Vivarelli 1999) provides support to the widely used definition of incumbents as those firms operating for more than 6 years (Audretsch 1995). However, one has to consider that in a very dynamic business environment of a transitional economy as Vietnam, the rules of games are continuously changing and the number of new entries is increasing over time. Thus, in this paper, incumbent firms are defined as existing firms aged more than 3 years.

Several control variables are also used. They can be categorized into two different sets. The first set includes other sources of entrepreneurial opportunities in addition to those created by incumbent firms. The second refers to other

motivations to establish new firms rather than those inspired by entrepreneurial opportunities. Each source is represented by one or two specific variables that either have been employed in previous research or reflected the unique regional factors of Vietnam.

Appendix A presents the construction and descriptive statistics of the dependent and independent variables. Appendix B shows the pooled pair-wise correlation matrix of respective variables. Since variables are aggregated data at provincial level, by nature of the construction intercorrelations among them are quite high and significant. For instance, governmental investments are higher in provinces that are richer in terms of their endowment of entrepreneurial and innovative spirit; MES is certainly higher in provinces with a larger share of population working in the private sector; technological resources are generally located in urban areas.

Control variables: Regional entrepreneurial climate. Small firms are often considered the seedbeds for future entrepreneurs, as their employees display a considerably higher propensity to start their business than those in older firms (Beesley and Hamilton 1984; Wagner 2004). From the management perspective, employees in small firms have relatively good possibility of direct contact with business founders who may serve as their role models (Reynolds 1994). From an economic viewpoint, since employment in small firms is often less secure and well paid than it is in large firms, individuals working for small firms are more prone to entrepreneurship than their more risk-averse peers in large firms (Storey 1982).3 Two proxies will be adopted to reflect the entrepreneurial spirit of a particular region: the proportion of micro-sized firms in the total existing firms (1 year lagged);4 and the share of enterprises' employment in the total regional labour force (1 year lagged).5 A number of studies have found a positive and statistically significant relation between proportion of small firms and start-up rate (Fotopou-los and Spence 1999; Guesnier 1994; Hart and Gudgin 1994; Keeble and Walker 1994; Piergiovanni and Santarelli 1995). However, while Keeble and Walker suggest that this effect is only limited to the manufacturing sectors whereas the service sectors reflect the importance of large firms, Audretsch and Fritsch (1994) could not find the predominance of small firms in manufacturing sectors due to the relevance of economies of scale in such activities, and Garofoli (1994) found that this relationship does not hold for Italy due to its unique structural characteristics.

Control variables: Entrepreneurial demand. Expanding markets and demand for goods and services are both an incentive for incumbent firms to extend their production activities and major drivers of firm births. Regional gross domestic income6 per capita (1-year lagged gross domestic product [GDP] per capita at competitive price of 1994) is used here as the indicator for the level of demand and

welfare. Previous studies such as Audretsch and Fritsch (1994), Reynolds (1994), Davidsson, Lindmark, and Olofsson (1994), and Armington and Acs (2002) find out that entry tends to be higher in regions where gross value added per person is higher. However, Kangasharju (2000) and Sutaria and Hicks (2004) do not find any influence of income per capita on new manufacturing firm births, concluding that personal wealth is a relatively minor driver of the decision to start up a new firm.

Control variables: Regional structural factors. Urbanisation/Agglomeration. Regions with a high population density may have higher start-up rates than rural areas due to better access to large and differentiated markets for production factors such as capital, labour, and services. Moreover, agglomeration economies may favour firms' access to the knowledge spillovers of both academic institutions and other firms located in the region. Krugman (1991) offers three reasons for the concentration of firms in agglomerated locations: 1) pooled market for high-skilled labours, 2) non-pecuniary transactions, or production of non-tradable specialized inputs, and 3) informational spillovers. However, sunk costs of starting a business (wages, rent for office space, etc.) are usually higher in a high-density agglomeration than in rural areas. On the other hand, although agglomerations provide a large local output market, a larger number of local suppliers may result in a more intense competitive environment. Two indicators will be adopted to investigate the agglomeration effects on new firm births: population density (1 year lagged) and the share of urban population in the total regional population (1 year lagged). Positive and significant effects of population density on start-up rates can be found in Guesnier (1994), Audretsch and Fritsch (1994), Keeble and Walker (1994), Armington and Acs (2002), and Brixy and Grotz (2007). Whereas the urban incubator theory, according to which urban areas have advantages as incubators for new firms, is supported for the case of UK (Keeble and Walker 1994), it is not supported for the case of Ireland (Hart and Gudgin 1994).

Market structure. The proxy is 1-year lagged MES, defined as the mean number of employees per establishment. It is measured as the ratio of the total number of employees working in enterprises of all ownership types over the number of firms in the region. Its coefficient has been hypothesized to be negatively related to regional entry rate since larger average establishment size indicates greater dominance by large firms in the market, as well as greater entry barriers for small start-ups. However, while Armington and Acs (2002) report a negative impact of MES on new firm formation, Audretsch and Fritsch (1994) find no evidence of such effect, and Sutaria and Hicks (2004) even find a positive relation between MES and the region's entry rate due to larger firms outsourcing to smaller neighbouring firms' specialty goods and services.

Education background/Innovativeness. A large number of studies emphasize the crucial role of knowledge and ideas as a stimulating source for new business entry (Acs et al. 2009; Agarwal et al. 2004; Klepper and Sleeper 2005; Shane 2000; Shepherd and DeTienne 2005). There is no doubt that innovative start-ups are more likely to occur in regions that are characterized by a high level of knowledge and innovative activities. The regional share of technical and research and development (R&D) personnel in the total labour force is therefore used as a measure of these activities.

Institutional factors. These are among the major drivers of entrepreneur's decisions to establish new firms (Verheul, Carree, and Santarelli 2009). Unfortunately, we do not have reliable indicators to reflect this factor at the provincial level. Therefore, in this paper, we use only 1-year lagged public investment7 as an indicator for the attractiveness of a regional economic environment and the presence of public infrastructures. By enhancing a region's attractiveness, public investment may also attract new firms, which further contribute to regional growth. In Vietnam, contingent on annually proposed macroeconomic strategies, the government adjusts its public investments into each province accordingly. To account for differences in the economic size of provinces (large provinces receive more state support than small ones), annual public investment per person at working age will be used as a proxy for regional economic development.

Control variables: Other factors. Income effect. For the case of Vietnam, 1-year lagged average compensation per month of employees working in SMEs (small and medium-sized enterprises) is used as a proxy for their opportunity cost as nascent entrepreneurs to actually start up their own businesses. We have discussed above the higher likelihood for employees in small firms to become self-employed. It is plausible that the opportunity cost of their start-up decision is represented by their salaries. The higher the salaries they receive, the less likely they will split off to set up their own establishments.

Unemployment effect. Regional unemployment may affect the level of start-up activity in contradictory ways (Santarelli, Carree, and Verheul 2009). On one hand, unemployed persons face low opportunity costs when setting up their own businesses with no other prospects for employment ("necessity entrepreneurship"). Hence, a high level of unemployment may force individuals to become self-employed workers, especially if residential mobility is unattractive (Guesnier 1994; Reynolds 1994; Wang 2006). On the other hand, high unemployment rates are generally seen as signs of quantitative and structural problems on the labour market (Armington and Acs 2002; Fritsch 1992; Storey 1994). High regional unemployment rates may indicate slow growth, relatively low demand, and correspondingly bad prospects for a successful start-up, thereby dampening incentives for new firms

to locate within the region. Moreover, unemployed persons may have little capital of their own and/or limited access to external finance sources. In fact, empirical evidence reflects these two conflicting forces. While Wagner and Sternberg (2004) suggest that unemployed individuals have a higher propensity to be nascent entrepreneurs than people in employment, Gaygisiz and Koksal (2003) and Sutaria and Hicks (2004) imply a negative significant impact of the unemployment rate on new firm entry. However, in most of the empirical studies, the impact of the unemployment rate on new business entry has been found to be weakly significant or insignificant (Armington and Acs 2002; Brixy and Grotz 2007; Keeble and Walker 1994; Reynolds, Storey, and Westhead 1994). Data on 1-year lagged urban unemployment rate in regions (the average number of unemployed in a year divided by this year's labour force) will be included to investigate its effect.

As the samples are exhaustive geographical regions of a country whereby the economic situation in a region is likely influenced by the one of nearby regions, we need to isolate such spatial autocorrelation:

Spatial autocorrelation. Spatial autocorrelation can cause the standard deviation of estimated coefficients to be underestimated (Brixy and Grotz 2007). On one side, the effect of factors influencing new firm entry may not be limited to a region but may spill over adjacent regions. Thus, the mean of regional start-up rates in the regions neighbouring each region is included as a measure of the spillover effect. It is expected to have a positive effect on the dependent variable due to possible spatial interdependence. On the other side, unobserved factors not fully reflected in the explanatory variables of a region but influencing neighbouring regions might also matter. In this case, the mean of residuals of neighbouring regions will be included (Fritsch and Falck 2007).8

Model development and estimation methods

We propose two model treatments: a static model where all the independent variables are 1 year lagged; and a dynamic model where the lagged value of the dependent variable is also included to isolate the effect of potential performance shock that may impede or stimulate new entries. For each model, we estimate two different specifications: the first one treats growth of incumbent firms as exogenous, whereas the second measures diversification endogenously to take into account regional-specific characteristics that may both stimulate firm entries and foster firm growth.

However, since the period considered for the empirical analysis was marked by a rather stable and fast economic growth trend all over the country, the influence of the economic cycle on firm entries at the aggregate level might be trivial.

Indeed, the insignificant effect of the lagged dependent variable in the dynamic model justifies this conjecture, and indicates the superiority of the static model.

Although endogeneity bias is commonly confronted in cross-sectional studies, it is less frequently considered as a factor hindering economic analysis in the case of panel data estimations like those performed in the present study. This is partially due to the conception that fixed-effects estimation will eliminate most forms of unobserved heterogeneity (Verbeek and Nijman 1992). In this connection, Vella (1998) claims that certain forms of heterogeneity will not be eliminated with either panel fixed effects (FE) or random effects (RE) models. The Durbin-Wu-Hausman test will be adopted to check whether endogeneity is likely to bias our estimation.

The static model. Incumbent firm growth is measured exogenously with regional-level characteristics: In addition, the error components model is adopted to introduce the regional and time effects in the error terms. Thus, spatial and temporal heterogeneities are incorporated into the model by its variance. Assuming that the primary predictors selected for this study would take approximately 1 year to influence the process of new firm formation, all explanatory variables are taken 1 year lagged:

yu =a + xh,+ yzut-1 + v, + eit (1)

(i = 1, 2,..., n; t = 1, 2,..., T)

where x,t-1 is the explanatory variable; zu-1 is the vector of control variables; vi + Eit is the residual, in which v, is the regional-specific residual; it differs between regions but, for any particular unit, its value is constant; E,t is the usual error term with the following assumptions:

Eit is uncorrelated with Vi for all i and t.

E(E,t) = 0 ,

a2e i = i', t = t' (H3a)

H3: E(s1,s1',')= .

0 otherwise ( H3b) H4: slt ~ N(0, aS)

- Tests for violations of assumptions:

+ Heteroskedasticity (H3a): Two diagnostic tests, Breusch-Pagan's and White's test, are employed to check for the presence of heteroskedasticity (Table 1). It was indeed confirmed by both tests. Thus, estimation with ordinary least squares (OLS) is rejected, and the alternative estimation technique capable of correcting for heteroskedastic errors is a "robust" regression method with standard errors corrected for heteroskedasticity by White's method.

Table 1. Test for Heteroskedasticity.

Breusch-Pagan test chi2(1) = 1375.33

p > chi2 = 0.000 White test chi2(90) = 442.7472

p > chi2 = 0.000

Table 2. Wooldridge Test for AR(1) Serial Correlation.

Serial correlation test start-up rate

Wooldridge first-order F(1,60) = 0.877

serial correlation test p-value = 0.3528

AR (1), autoregressive of order 1.

+ Serial correlation in time-series data (H3b): The Wooldridge (2002) test for first-order autocorrelation in panel data is insignificant even at 10 percent level, which indicates the absence of first-order serial correlation (Table 2). Serially correlated errors will give biased estimators by increasing variances of estimated coefficients. In this case, we can feel secured that net start-up rate as the dependent variable satisfies the assumption of no serial correlation.

- Estimation methods: Given the panel structure of the data, fixed-effects or random-effects regression model can be used. According to Balestra and Krish-nakumar (2008), as the nature of the sample is closed and exhaustive, and the type of inference is with respect to effects that are in the sample, fixed-effects regression is the natural candidate. Estimators are obtained from fixed-effects regression of equation (1) as follows: 1 1

Define regional-specific means by yi = ~ X,=i yu and x = ~ Xt=1x", and the deviations from these means as y* = yi - yJT and X* = Xt - lTx'. Applying the within transformation, model (1) becomes

y* = X*# + e*(i = 1,2,..., n)

The transformation eliminates totally unobserved regional-specific effects Vi when n is fixed (clearly the case) and T goes to infinity, which results in efficient estimator

Table 3. Test for Endogeneity.

Durbin-Wu-Hausman test F(1,415) = 3.61

p-value = 0.0580

$wg = X (X*' X*)-1 X*' y* i=1

- Incumbent firm growth measured endogenously with regional-level characteristics: Causality may run in both directions—from incumbent firm growth to new firm formation (as we mentioned from the above theoretical framework) or vice versa, i.e., it may be more efficient for larger firms to outsource to new neighbouring firms some of their functions and are able to focus on their core production capabilities, which further enhance their growth. Thus, incumbent firm growth is likely to be correlated with controlled observable regional characteristics and unobserved characteristics absorbed in error terms.

The Durbin-Wu-Hausman test indicates the presence of endogeneity of incumbent firm growth although the presence is statistically significant at only 10 percent level (Table 3). The test begins with the reduced form regression in which the assumed-endogenous incumbent firm growth is the dependent variable and all other observed exogenous regional-level characteristics are independent ones. Then, residuals predicted from this regression are added into the structural form equation (1). The endogeneity problem is determined based on the significance of the residual coefficient.

In case of endogeneity problems, instrumental variable (IV) two-stage least squares (2SLS) estimation is often adopted. However, since heteroskedasticity is present, we will apply the generalized method of moments (GMM) technique, which is more efficient than the 2SLS (Baum and Schaffer 2003). Following Arellano-Bond (Arellano and Bond 1991), the instrument for the endogenous incumbent firm growth is its one-period lagged values. This makes the endogenous predetermined and, hence, not correlated with the error term in equation (1). The results from GMM estimation applied for the static model are presented in Table 4.

The dynamic model. In equation (2), the 1-year lagged dependent variable is included:

yu = a+xltfi + yzut-1 + v, + eit (2)

(i = 1, 2,..., n; t = 1, 2,..., T)

Table 4. Determinants of New Business Formation.

Regional net entry rate- —Hetero skedasticity- adjusted models

Static model Dynamic model

)':: = a + Xi,,-i/i + X-/./-1 + V,- + t,-, yu = ocyu-i + xi,t-\ß + yzi,t-\ + Vi +

FE GMM GMM exogenous GMM endogenous

(1) (2) (3) (4) (5) (6) (7) (8)

CD 73 O

O c s 00 m

Tl 73 m z m c

73 </> I

Intercept -0.172 (0.093) -0.189* (0.096)

Net entry rate (t-1) na. n.a.

Growth of incumbent firms (t-1) 0.019* (0.009) 0.019* (0.009)

Entrepreneurial climate

Share of micro-sized firms (t-1) -0.028(0.081) -0.026(0.082)

Share of enterprises'employment (t-1) na. 0.281 (0.543)

Entrepreneurial demand Growth of GDP per capita (t-1)

Agglomeration

Growth ofpopulation density (t-1) 0.0339(0.0748) 0.031(0.081)

Share of urban population (t-1) 0.504(0.305) 0.499(0.312)

Market structure

MES (t-1) 0.0009* (0.0004) 0.0009* (0.0004) Market innovativeness

Share of technical/R&D personnel (t-1) 6.655* (3.224) 6.575* (3.396) Income effect

Growth of monthly compensation per na. 0.0495* (0.026) employee (t-1)

-0.0397 (0.026) -0.114** (0.042) 0.0116 (0.013) 0.0112 (0.0145) 0.0148 (0.009) 0.0128 (0.0114)

-0.129 (0.115) -0.147 (0.118) -0.212 (0.123)

-0.232 (0.13)

0.0184* (0.009) 0.0175* (0.009) 0.014* (0.0

0.014* (0.0'

0.0226 (0.014) 0.024 (0.015)

-0.015 (0.03) n.a.

-0.009 (0.033) 0.179 (0.111) 0.004 (0.109) 0.043 (0.114) 0.0218 (0.113)

0.203 (0.407)

0.6699* (0.302)

0.987* (0.428)

0.067** (0.0035) 0.058** (0.005) 0.067** (0.0032) 0.054** (0.0067) 0.068** (0.006) 0.0597** (0.0084) 0.065** (0.006) 0.0599** (0.0081) -0.225 (0.155) -0.194 (0.138) -0.015 (0.054) 0.0006 (0.055) -0.022 (0.049) -0.0125 (0.0493)

0.211** (0.082) 0.217** (0.C

0.243 (0.422) 0.299 (0.433) 0.177 (0.387)

0.229 (0.385)

0.00005 (0.0003) 0.00004 (0.0002) 0.0033** (0.001) 0.0032** (0.0011) 0.0033** (0.0011) 0.0029* (0.001) 9.775** (1.64) 10.088** (1.972) 8.125* (3.577) 7.837* (3.628) 7.674* (3.285) 7.13* (3.33)

0.0692* (0.0318)

0.042 (0.034)

0.034 (0.037)

Table 4. (continued)

CD 73 O

D m o m <: □0 m 73

Regional net entry rate—Heteroskedasticity-adjusted models

Static model Dynamic model

)':: = a + xi,-i/i + yu = ayu -1 + l/>' +

Kf.i-l + V,- + fci, Pf.i-i + Vi + fci,

FE GMM GMM exogenous GMM endogenous

(1) (2) (3) (4) (5) (6) (7) (8)

na. -0.0002 (0.005) n.a. 0.009 (0.006) na. 0.0041 (0.0084) n.a. 0.0074 (0.0103)

0.022** (O.OOS) 0.01SS* (0.00S6 ) 0.022** (0.00S) 0.0257** (0.009) 0.0187* (0.0103) 0.0138 (0.0104) 0.0156 (0.0107) 0.009 (0.011)

0.496** (0.109) 0.492** (0.109) 0.287** (0.075) 0.284** (0.076) 0.549* (0.254) 0.551* (0.262) 0.537* (0.218) 0.543* (0.231)

209.21** 157.56** 184.52** 139.80** na. na. n.a. na.

n.a. n.a. 0.096 0.183

n.a. n.a. n.a. na. -3.13** -3.39** -2.39* -2.38*

na. n.a. n.a. na. 0.09 -0.07 -0.42 -0.57

Unemployment effect Unemployment rate (t-1)

Regional economic environment Public investment per capita (t-1)

Spatial autocorrelation control Spatial spillover effects .F-value

Hansen J statistic

Arellano-Bond test for autocovariance of order 1

Arellano-Bond test for autocovariance of

order 2 Observations

* p< .05; ** p< .01. Standard errors are in parentheses.

FE, fixed effects; GDP, gross domestic product; GMM, generalized method of moments; MES, mean establishment size; n.a., not applicable; R&D, research and development.

Several econometric problems may arise from estimating equation (2):

1. The diversification index zit is assumed to be endogenous.

2. Time-invariant unobserved firm characteristics (fixed effects) Vi may be correlated with diversification index zü and control explanatory variables Xü.

3. The panel data set has a short time dimension (T = 8) and a larger number of regions (n = 61). Thus, the presence of the lagged dependent variable yit-i would give rise to autocorrelation since it is correlated with fixed effects. It is therefore also treated as endogenous variable.

OLS and within-group estimators of a, g b are all inconsistent. To solve problem 1 and problem 2, one would usually use fixed-effects IV estimation (2SLS), depending on the availability and validity of exogenous instruments. Therefore, we used the Arellano-Bond (Arrellano and Bond 1991) difference GMM estimator first proposed by Holtz-Eakin, Newey, and Rosen (1988). Lagged levels of the endogenous regressor zit are used as instruments, which rise over time. This makes the endogenous variable predetermined and, hence, not correlated with the error term in equation (2). To cope with problem 2 (fixed effects), the difference GMM uses first differences to transform equation (2) into:

Ayit = aAylt-1 + yAzlt + fiAXit + Aeit (3)

Transforming the regressors by first differencing, the fixed regional-specific effect is removed, because it does not vary with time. Finally, to cope with problem 3, the Arellano-Bond estimator was designed for small-T and large-N panels. For the endogenous lagged dependent variable, the first-differenced lagged dependent variable is instrumented with its past levels.

Model Estimation: Results and Interpretation

Table 4 presents the estimation results for both the static and the dynamic model. In Appendix C, we report the results from estimations in which incumbent firms are defined as "existing firms aged more than 3 years."

The static model adopts fixed-effects regression when incumbent firm growth is assumed as exogenous, and GMM technique when it is assumed as endogenous. The dynamic model applies differenced GMM estimation with two treatments: exogenous and endogenous incumbent firm growth.

Overall, the static model with GMM treatment is the most preferable estimation procedure, and for this reason we will base our comments on the results

forthcoming from this specification. Several factors contribute to make the static model with GMM treatment preferable to any other specification:

1. The data set has panel structure with a short time dimension and a large number of observations.

2. Heteroskedasticity tests indicate the presence of heteroskedasticity in the data set.

3. The Wooldridge test for first-order serial correlation shows the absence of serial correlation.

4. Durbin-Wu-Hausman test for endogeneity suggests the presence of endoge-neity of "incumbent firm growth."

In general, the growth of incumbent firms in a particular region does have a stimulating effect on firm formation activities. The evidence is stronger when the sign of the associated regression coefficient is constant and the coefficient estimate is statistically significant across all model specifications (Levine and Renelt 1992). This finding strongly supports both hypothesis 1 (H1) and hypothesis 2 (H2) put forward in the third section, and is to a large extent consistent with the empirical findings by Santarelli and Sterlacchini (1994) in relation to Italy, therefore showing that not necessarily does the relationship between incumbent firms' growth and new firm formation follow different patterns in advanced and transition/developing economies.

In fact, incumbents' growth, which is the performance indicator most easily observed by would-be entrepreneurs, also provided that most of them have a dependent job in the same industry. The growth of incumbent firms motivates competition among nascent entrepreneurs to "seize" and transform opportunities arising in a fast-growing industry into new firms.

The parameters reduce their significance in both exogenous and endogenous treatments of the dynamic model. However, the insignificant 1-year lagged "incumbent firm growth" variable justifies the superiority of the findings from the static model with endogeneity treatment.

With respect to the control variables, five predictors—growth of GDP per capita, share of urban population, share of technical personnel, growth of monthly compensation per employee in SMEs, and governmental investment per capita—are found to have statistically significant effects on the dependent variable. These findings as a whole, and in particular those for the growth of GDP and the monthly compensation variables, are consistent with a simple take on new firm formation, namely that increases in wealth and a favourable economic climate lead to increases in new firm formation. In other words, enterprise creation may represent a response

to, rather than a cause of, wealth creation. In this connection, it can be argued that there is a multiplier trickle-down effect that benefits both employees and business owners. Accordingly, when an exogenous increase in demand occurs, this is only in part satisfied by existing incumbent firms—whose performance improves—and partly by an increase in the stock of firms. This suggests that a macroeconomic mechanism has to be considered, whereby greater economic growth simultaneously lifts incumbent revenues and stimulates new entry.

However, the findings with the "share of technical personnel" variable provide a still preliminary but worth of further investigation support to the knowledge spillover theory of entrepreneurship put forward by Acs et al. (2009). Should they be confirmed by further studies dealing with transition/developing countries, one might argue that this theory provides a valuable framework for the explanation of the relationship between innovation and entrepreneurship in both advanced and emerging economies.

The indicator used as a proxy for the entrepreneurial climate and the (low) level of barriers to entry in a region, i.e., our measure of small firm presence, does not seem to influence entry. This is quite surprising for the case of an emerging economy like Vietnam in which we should expect that the prevalence of micro-sized firms in the market is conducive to new firm formation. A number of comparable studies support the stimulating effect of the entrepreneurial climate on firm entries (Fotopoulos and Spence 1999; Guesnier 1994; Hart and Gudgin 1994; Keeble and Walker 1994). Two reasons can be submitted to explain this apparently puzzling finding: either nascent entrepreneurs themselves are not motivated to participate into a market of intense competition among newly established firms or "the revolving door" effect of the market is so efficient that a significant number of new entries in the previous year will immediately cause an equivalent number of exits among both unprofitable incumbents and "bad entries" (Santarelli and Vivarelli 2007), which may subsequently produce negative net entry rates in the current year. The fixed-effects regression when the share of micro-sized firms of the current year is used instead of the 1-year lagged one indeed justifies the latter explanation. The share of micro-sized firms has a positive relationship with net entry in the same year: the greater the dominance of micro-firms, the larger the number of new entries in the same year, which is able to more than compensate exits.

But, consistent also with the "creative destruction" mechanism proposed by Schumpeter (1934), this might lead to a negative relationship with net entry in the next year: Entry of a great number of new firms creates a highly competitive and turbulent market such that an equivalently large number of firms, either "bad entries" or stagnating incumbents, are forced to exit. The positive, though

statistically insignificant, effect of the share of enterprises' employment in the total regional labour supply, on the other hand, indicates that the dominance of a strong private sector in the market is favourable to start-up activities. This finding partially confirms the "incubator theory," assuming that people employed in smaller firms are more prone to set up a business of their own. It is likely that working in smaller firms allows employees to have deeper and broader insights into how to run a firm, while working in larger firms enables them to be more specialized. Since nearly 95 percent of enterprises in Vietnam are household businesses with household proprietorship (Rand andTarp 2007), a large fraction of the individuals who are employed in the private sector are self-employed. Thus, it is fair to say that a higher share of the labour force employed in the private sector reflects a higher entrepreneurial spirit of the region.

The positive and significant coefficient of the growth of the regional GDP per capita variable indicates that net entry is higher where gross value added per person is higher. The effect is quite strong no matter what methodological treatment is applied.

For MES, consistent with Sutaria and Hicks (2004), the estimation shows a positive and significant relationship with net entry. The dominance of larger firms will give abundant outsourcing opportunities to smaller neighbouring firms in producing specialty goods and services, and hence, generally stimulates firm entries. However, this result may differ across industries, since entry barrier industries characterized by economies of scale and high labour or capital inten-siveness are more severe than those in modern high-tech industries.

For a potential entrepreneur currently working as paid employee, the opportunity cost of establishing a new firm corresponds to his/her monthly salary or compensation. However, in Vietnam, since it is quite common that entrepreneurs maintain both their salary-paid job and self-employed business, and since the largest fraction of investment capital for their own business comes from savings and salary, compensation growth actually imposes a stimulating effect on new firm formation. The higher the salary increase over time, the sooner the employee saves enough capital for setting up their own business. Besides, the push effect from regional economic environment through public investments on entry in each province is quite strong. Other things held constant, people in a province being endowed with more state-invested capital are more likely to be self-employed than equally able people in another province with less state-invested capital.

The coefficient of the variable measuring the share of urban population in the overall provincial population is positive and statistically significant, therefore supporting the urban incubator theory that identifies in metropolitan areas and urban centres important nurseries of new firms. However, since the beginning of

the process of economic reforms initiated in 1986 with the purpose of creating a "socialist-oriented market economy" (known as doi moi), Vietnam has witnessed a rapid urbanisation and industrialisation erpcess, ultimately leading to the conversion of large rural areas into metropolitan areas and urban centres; these reforms have forced thousands of farmer households to be in the transition from traditional agricultural and rural-based economic activities to urbanized livelihoods (cf. Tran-Nam and Pham 2003). The newly born urban areas have in most cases worked as a "revolving door" to eliminate "entry mistakes" (Santarelli and Vivarelli 2007), rather than as an incubator of new successful firms.

Our findings of a non-statistically significant relationship between unemployment and entry are not at odds with previous empirical studies reporting contradictory evidence with respect to such a relationship (cf. Santarelli, Carree, and Verheul 2009). In this respect, it is worth noting that the period from 2000 to 2008 marks both high economic growth and poor labour market conditions in Vietnam.

On one hand, the country has benefited from the launch of Enterprise Law in 2000, creating a firmer legal basis for the development of the private sector, and from the official approval to hold World Trade Organization membership. In fact, the removal of entry barriers might have inflated the entry numbers in areas where incumbents are enjoying higher growth.

On the other hand, since high economic growth attracts people to migrate to big cities in large number, it is inevitable that while unemployment increases quickly in urban areas, job opportunities are redundant in rural areas. Thus, the insignificant relationship between unemployment rate and net entry may depend upon the interplay of two coexisting forces, i.e., high unemployment rate shows a poor functioning market that hampers start-up efforts and high economic growth stimulates new firm formation to satisfy increasing consumers' demands.

Finally, we find statistically significant spatial autocorrelation among neighbouring regions across all methodological treatments. As expected, neighbouring regions share more common features than those further away.

Conclusions

The main finding in this paper is that in 61 Vietnamese provinces from 2000 to 2008, net entry is associated with the performance of incumbent firms and the overall performance of the economy. Incumbents' growth and GDP growth create changes in the existing production system and stimulate an imitation effect, consistent with the idea that new business founders are "individuals who discover, evaluate, and exploit" (Shane and Venkataraman 2000: 218) opportunities.

There are two main limitations to our study which open up directions for future research. First, we use net entry as a proxy for regional firm formation activities. Since the "revolving door effect" is quite strong in the Vietnamese regions, a significant number of new entries in previous year are likely to cause an almost equivalent number of exits in the subsequent year, in relation to which one might therefore find negative net entry. Second, we perform our analysis at a sectorally aggregated level, without taking into account the effect of industry-specific characteristics on new business formation and net entry. Should new data be released, we will give explicit consideration to the issues of gross entry, taken as the most reliable proxy for entrepreneurship capital (Acs et al. 2009), and of industry-specific determinants of new firm formation (Carree, Santarelli, and Verheul 2008).

1. For example, introducing the iPad Apple demonstrated that there is a market for tablet PCs and induced many imitations. We thank a reviewer for pointing this out.

2. Within a Marshallian partial equilibrium framework, all firms have U-shaped long-run average cost curves with identical values at their minimum points. Positive or negative excess profits cause new firms to enter or existing firms to leave the industry. The expansion or contraction of industry output through changes in the number of firms continues until a long-run equilibrium is established with zero excess profits, i.e., when output prices equal the minimum average cost.

3. Fritsch and Falck (2007) suggest that a high promotion of employment in small firms may also indicate a low minimum efficient scale which can be assumed favourable for firm entry.

4. As most start-ups in Vietnam are very small, the share of micro-sized firms in the total number of enterprises in the region could be a good proxy for its breeding ground for nascent entrepreneur-ship. In 2006, nearly 60 percent of establishments are micro-sized firms (employing less than 10 employees) with approximately 52 percent of employment share (GSO 2006). Here we adopt the World Bank definitions about firm size: micro enterprises, up to 9 employees; small enterprises, up to 49 employees; medium enterprises, up to 299 employees; large enterprises, more than 300 employees.

5. This variable denotes the percentage proportion of the number of employees working in the private sector over the total (private + public) number of people (employed + unemployed) at the working age (over 15 years old) in the province.

6. But it might also reflect the employees' qualification and the regional price level, particularly price differences between rural areas and larger cities.

7. Proxied by total national money to the province.

8. Unobserved factors of new business entry processes in adjacent regions are not independent, but related to this process in a particular region. Since they are reflected in the disturbance terms of the regression equations, we follow Anselin (1988) and Fritsch and Falck (2007) in using the weighted average of the disturbance terms of adjacent regions to account for spatial autocorrelation.

REFERENCES

Acs, Z.J., D.B. Audretsch, P. Braunerhjelm, and B. Carlsson. 2009. The knowledge spillover theory of entrepreneurship. Small Business Economics 32(1): 15-30.

Agarwal, R., R. Echambadi, A. Franco, and M.B. Sarkar. 2004. Knowledge transfer through inheritance: Spinout generation, development and survival. Academy of Management Journal 47(4): 501-522.

Anselin, L. 1988. Spatial Econometrics: Methods and Models. Kluwer, Dordrecht: Springer.

Arellano, M., and S. Bond. 1991. Some tests of specification for panel data: Monte Carlo evidence and an application to employment equations. Review of Economic Studies 58(2): 277297.

Armington, C., and Z.J. Acs. 2002. The determinants of regional variation in new firm formation. Regional Studies 36(1): 33-45.

Audretsch, D.B. 1995. Innovation and industry evolution. Cambridge, MA: MIT Press.

Audretsch, D.B., and M.P. Feldman. 1996. R&D spillovers and the geography of innovation and production. The American Economic Review 86(3): 630-640.

Audretsch, D.B., and M. Fritsch. 1994. The geography of firm births in Germany. Regional Studies 28(4): 359-365.

Audretsch, D.B., M. Keilbach, and E.E. Lehmann. 2005. The knowledge spillover theory of entrepreneurship and technological diffusion. In University entrepreneurship and technology transfer: Process, design, and intellectual property, ed. G.D. Libecap, 69-91. Bingley: Emerald Group Publishing Limited.

Audretsch, D.B., and E.E. Lehmann. 2005. Does the knowledge spillover theory of entrepreneurship hold for regions? Research Policy 34(8): 1191-1202.

Audretsch, D.B., E. Santarelli, and M. Vivarelli. 1999. Start-up size and industrial dynamics: Some evidence from Italian manufacturing. International Journal of Industrial Organization 17(7): 965-983.

Balestra, P. 1995. Fixed effect models and fixed coefficient models. In The econometrics of panel data, ed. L. Matyas, and P. Sevestre, 23-48. Dordrecht: Kluwer Academic Publishers.

Baum, F.C., and M.E. Schaffer. 2003. Instrumental variables and GMM: Estimation and testing. The Stata Journal 3(1): 1-31.

Beesley, M.E., and M.T. Hamilton. 1984. Small firms' seedbed role and the concept of turbulence. The Journal of Industrial Economics 33(2): 217-231.

Brixy, U., and R. Grotz. 2007. Regional patterns and determinants of the survival of new firms in Western Germany. Entrepreneurship and Regional Development 19(4): 293-312.

Carree, M., E. Santarelli, and I. Verheul. 2008. Firm entry and exit in Italian provinces and the relationship with unemployment. International Entrepreneurship and Management Journal 4(2): 171-186.

Caves, R.E. 1998. Industrial organization and new findings on the turnover and mobility of firms. Journal of Economic Literature 36(4): 1947-1982.

Davidsson, P., L. Lindmark, and C. Olofsson. 1994. New firm formation and regional development in Sweden. Regional Studies 28(4): 395-410.

Evans, D.S., and B. Jovanovic. 1989. An estimated model of entrepreneurial choice under liquidity constraints. Journal ofPolitical Economy 97(4): 808-827.

Fotopoulos, G., and N. Spence. 1999. Spatial variation in new manufacturing plant openings: Some empirical evidence from Greece. Regional Studies 33(3): 219-229.

Fritsch, M. 1992. Regional differences in new firm foundation: Evidence from West Germany. Regional Studies 26(3): 233-241.

Fritsch, M., and O. Falck. 2007. New business formation by industry over space and time: A multidimensional analysis. Regional Studies 41(2): 157-172.

Garofoli, G. 1994. New firm formation and regional development: The Italian case. Regional Studies 28(4): 381-393.

Gaygisiz, E., and M.Y. Koksal. 2003. Regional variation in new firm formation in Turkey: Cross-sectional and panel data evidence. ERC Working Papers in Economics, 03/08.

General Statistics Office of Vietnam (GSO). 2006. Statistic yearbook 2005. Hanoi: Statistical Publisher.

Guesnier, B. 1994. Regional variations in new firm formation in France. Regional Studies 28(4): 347-358.

Hart, M., and G. Gudgin. 1994. Spatial variations in new firm formation in the Republic of Ireland, 1980-1990. Regional Studies 28(4): 367-380.

Hausmann, R., and D. Rodrik. 2003. Economic development as self-discovery. Journal of Development Economics 72(2): 603-633.

Helfat, C.E., and M.B. Lieberman. 2002. The birth of capabilities: Market entry and the importance of pre-history. Industrial and Corporate Change 11(4): 725-760.

Holtz-Eakin, D., W. Newey, and H.S. Rosen. 1988. Estimating vector autoregressions with panel data. Econometrica 56(6): 1371-1395.

Johnson, P.S., and D.G. Cathcart. 1979. The founders of new manufacturing firms: A note on the size of their incubator plants. The Journal of Industrial Economics 28(2): 219-224.

Kangasharju, A. 2000. Regional variations in firm formation: Panel and cross-section data evidence from Finland. Papers in Regional Science 79(4): 355-373.

Keeble, D., and S. Walker. 1994. New firms, small firms, and dead firms: Spatial patterns and determinants in the United Kingdom. Regional Studies 28(4): 411-427.

Kihlstrom, R.E., and J.J. Laffont. 1979. A general equilibrium entrepreneurial theory of firm formation based on risk aversion. The Journal ofPolitical Economy 87(4): 719-748.

Kirzner, I. 1997. Entrepreneurial discovery and the competitive market process: An Austrian approach. Journal of Economic Literature 35(1): 60-85.

Klepper, S. 2010. Spinoffs: A review and synthesis. European Management Review 6(3): 159-171.

Klepper, S., and S.D. Sleeper. 2005. Entry by spinoffs. Management Science 51(8): 1291-1306.

Krugman, P. 1991. Geography and trade. Cambridge, MA: MIT Press.

Lévesque, M., and D.A. Shepherd. 2004. Entrepreneurs' choice of entry strategy in emerging and developed markets. Journal of Business Venturing 19(1): 29-54.

Levine, R., and D. Renelt. 1992. A sensitivity analysis of cross-country growth regressions. The American Economic Review 82(4): 942-963.

Parker, S. 2009. Why do small firms produce the entrepreneurs? Journal of Socio-economics 38(3): 484-494.

Penrose, E.T. 1959. The theory of the growth of the firm. New York: John Wiley & Sons.

Piergiovanni, R., and E. Santarelli. 1995. The determinants of firm start-up and entry in Italian producer services. Small Business Economics 7(3): 221-230.

Rand, J., and F. Tarp. 2007. Characteristics of the Vietnamese business environment: Evidence from a SME survey in 2005. Research Report of the Danida Funded Business Sector Program Support.

Reynolds, P. 1994. Autonomous firm dynamics and economic growth in the United States, 1986-1990. Regional Studies 28(4): 429-442.

Reynolds, P., D.J. Storey, and P. Westhead. 1994. Cross-national comparisons of the variation in new firm formation rates. Regional Studies 28(4): 443-456.

Santarelli, E., M. Carree, and I. Verheul. 2009. Unemployment and firm entry and exit: An update on a controversial relationship. Regional Studies 43(8): 1061-1073.

Santarelli, E., and A. Sterlacchini. 1994. New firm formation in Italian industry: 1975-89. Small Business Economics 6(2): 95-106.

Santarelli, E., and M. Vivarelli. 2007. Entrepreneurship and the process of firms entry, survival and growth. Industrial and Corporate Change 16(3): 455-488.

Schumpeter, J.A. 1934. The theory of economic development: An inquiry into profits, capital credit, interest and the business cycle. Cambridge, MA: Harvard University Press.

Shane, S. 2000. Prior knowledge and the discovery of entrepreneurial opportunities. Organization Science 11(4): 448-469.

-. 2003. A general theory of entrepreneurship: The individual-opportunity nexus. Cheltenham:

Edward Elgar Publishing.

Shane, S., and S. Venkataraman. 2000. The promise of entrepreneurship as a field of research. Academy of Management Review 25(1): 217-226.

Shepherd, D.A., and D.R. DeTienne. 2005. Prior knowledge, potential financial reward, and opportunity identification. Entrepreneurship Theory and Practice 29(1): 91-112.

Storey, D.J. 1982. Entrepreneurship and the new firm. London: Croom Helm.

-. 1994. Understanding the small business sector. London: International Thomson Business

Press.

Sutaria, V., and D. Hicks. 2004. New firm formation: Dynamics and determinants. The Annals of Regional Science 38(2): 241-262.

Tran-Nam, B., and C.D. Pham, eds. 2003. The Vietnamese economy: Awakening the dormant dragon. London: RoutledgeCurzon.

Vella, F. 1998. Estimating models with sample selection bias: A survey. Journal of Human Resources 33(1): 127-169.

Verbeek, M., and E. Nijman. 1992. Testing for selectivity bias in panel data models. International Economic Review 33(3): 681-703.

Verheul, I., M. Carree, and E. Santarelli. 2009. Regional opportunities and policy initiatives for new venture creation. International Small Business Journal 27(5): 608-625.

Wagner, J. 2004. Are young and small firms hothouses for nascent entrepreneurs? Applied Economics Quarterly 50(4): 379-391.

Wagner, J., and R. Sternberg. 2004. Start-up activities, individual characteristics, and the regional milieu: Lessons for entrepreneurship support policies from German micro data. The Annals of Regional Science 38(2): 219-240.

Wang, S. 2006. Determinants of new firm formation in Taiwan. Small Business Economics 27(4-5): 313-323.

Wooldridge, J.M. 2002. Econometric analysis of cross-section and panel data. Cambridge, MA: MIT Press.

Appendix A Dependent Variables and Independent Variables

Categories

Observation Mean Standard deviation

Min Max Expected relationship

03 m 7J

Dependent variable

Explanatory variable

Control variables: entrepreneurial opportunities created from regional entrepreneurial indicators

Control variables: entrepreneurial opportunities created from regional structural indicators

New business formation Net entry rates

Growth of incumbent Incumbent growth firms

Entrepreneurial climate Share of micro-sized firms

Labour force in private sector

Entrepreneurial demand Growth of regional GDP per capita

Agglomeration

Market structure

Growth of population

density Urbanisation

Control variables: other individual motivational factors

Control variables: spatial autocorrelation

Education background Innovativeness

Regional economic Public investment

environment Unemployment effect Regional urban unemployment

Income effect

Compensation in private sector

Spillover effects

The ratio of number of new firms per 1,000

persons in labour supply The annual percentage change in revenues of existing incumbent firms (over 3 years old) The percentage share of micro-sized firms in the

total number of enterprises in the region The percentage share of enterprises' employment

in the total regional labour force The annual percentage change in regional gross domestic product per capita at comparative price of 1994 The annual percentage change in regional

population density The percentage share of urban population in the

total regional population The mean number of employees per

establishment The percentage share of technical and R&D personnel in the total regional labour force State-invested capital per person at working age

Annual urban unemployment rate

Log of average compensation per month of

employees working in SMEs Mean of regional start-up rates in the regions

neighbouring each region Mean of residuals of neighbouring regions

0.27 0.39

-0.79 2.73 n.a.

0.33 0.687 -0.18 4.814 Positive

44.39 16.27 6.14 97.05 Positive

1.53 94.73 Positive

-12.3 40.38 Positive

-11 15.49 Positive

5.79 87.5 Positive

14.14 226 Negative

0.69 1.34 0.015 10.24 Positive

1.309 1.350 0.001 11.99 Positive

9.62 12.63

9.61 4.36

1.34 1.343

22.3 15.85

59.65 31.25

5.49 1.13

6.85 0.33

2.28 8.96 Indeterminate

5.93 8.01 Negative

0.267 0.206 -0.16 1.27 Indeterminate

0.004 0.106 -0.37 0.374 Indeterminate

GDP, gross domestic product; MES, mean establishment size; n.a., not applicable; R&D, research and development; S ME, small and medium-sized enterprises.

CD 73 O

Appendix B Correlation Matrix of Dependent Variables and Independent Variables

Net start-up Incumbent growth Micro share Enter labour GDP. capita Population density Urban population MES Technical employment Compensation Unemployment Public investment

Net start-up 1.0000 n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a.

Incumbent growth 0.1231 1.0000 n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a.

Micro share -0.0506 -0.0594 1.0000 n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a.

Enter labour 0.7524* 0.1723* -0.0610 1.0000 n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a.

GDP capita 0.1121 0.1425* 0.0136 0.0640 1.0000 n.a. n.a. n.a. n.a. n.a. n.a. n.a.

Population density 0.2040* 0.0753 -0.0216 0.3498* -0.175* 1.0000 n.a. n.a. n.a. n.a. n.a. n.a.

Urban population 0.6734* 0.0685 0.1775* 0.6491 -0.0111 0.2918* 1.0000 n.a. n.a. n.a. n.a. n.a.

MES 0.1472* 0.1132 -0.455* 0.4966* 0.0822 0.1495* 0.1612* 1.0000 n.a. n.a. n.a. n.a.

Technical employment 0.5666* 0.0837 -0.0227 0.6216* 0.0700 0.2323* 0.5766* 0.2343* 1.0000 n.a. n.a. n.a.

Compensation 0.3889* 0.1742* 0.2115* 0.4501* 0.2140* 0.2358* 0.4791* 0.0269 0.4198* 1.0000 n.a. n.a.

Unemployment 0.2237* -0.0972 0.0095 0.1290* -0.145* -0.0003 0.1582* 0.2770* 0.2715* -0.0784 1.0000 n.a.

Public investment 0.5850* 0.2089* -0.0314 0.5863* 0.1605* 0.2808* 0.4145* 0.1308* 0.5084* 0.5077* -0.2431* 1.0000

o c s 00 m

Tl 73 m z m c

73 </> HZ

* p<.01.

GDP, gross domestic product; MES, mean establishment size; n.a., not applicable.

02 02 Ol

Appendix C Determinants of New Business Formation

Regional net entry rate- —Heteroskedasticity-adjusted static models

5-year-old incumbent firms Existing firms

FE GMM FE GMM

(9) (10) (11) (12) (13) (14) (15) (16)

Intercept -0.216* (0.096) -0.233* (0.099) -0.086* (0.037) -0.149** (0.042) -0.165 (0.097) -0.192 (0.1009) -0.069* (0.027) -0.116** (0.043)

Growth o f firms (t-1) 0.008* (0.004) 0.007* (0.003) 0.008* (0.004) 0.007* (0.004) 0.043 (0.031) 0.0315 (0.0341) 0.038 (0.036) 0.015 (0.039)

Entrepreneurial climate

Share of micro-sized firms (t-1) -0.016 (0.081) -0.016 (0.083) -0.016 (0.03) -0.0135 (0.041) -0.0305 (0.085) -0.0304 (0.088) 0.0102 (0.033) 0.031 (0.039)

Share of enterprises' employment (t-1) n.a. 0.173 (0.537) n.a. 0.179 (0.436) n.a. 0.18 (0.56) n.a. 0.414 (0.434)

Entrepreneurial demand

Growth of GDP per capita (t-1) 0.0703** (0.0039) 0.062** (0.006) 0.07** (0.003) 0.059** (0.006) 0.069** (0.004) 0.059** (0.0064) 0.069** (0.003) 0.061** (0.006)

Agglomeration

Growth of population density (t-1) 0.0494 (0.077) 0.047 (0.083) -0.192 (0.149) -0.167 (0.131) 0.032 (0.081) 0.036 (0.086) -0.189 (0.159) -0.152 (0.14)

Share of urban population (t-1) 0.555 (0.325) 0.557 (0.333) 0.238* (0.119) 0.254* (0.111) 0.478 (0.323) 0.001 (0.004) 0.174 (0.121) 0.168 (0.109)

Market structure

MES (t-1) 0.0007* (0.0004) 0.0008* (0.0004) 0.00004 (0.0003) 0.00001 (0.0003) 0.0008* (0.0004) 0.0008* (0.0004) 0.0001 (0.0002) 0.0004 (0.0003)

Market innovativeness

Share of technical/R&D personnel (t-1) 7.527* (3.057) 7.505* (3.19) 10.318** (1.996) 10.367** (2.334) 8.009** (3.152) 7.96* (3.28) 11.43** (1.93) 12.22** (2.19)

Income effect

Growth of monthly compensation n.a. 0.045* (0.026) n.a. 0.0596* (0.0336) n.a. 0.055* (0.028) n.a. 0.035 (0.032)

per employee (t-1)

Unemployment effect

Unemployment rate (t-1) n.a. -0.0006 (0.004) n.a. 0.009 (0.007) n.a. 0.001 (0.004) n.a. 0.003 (0.006)

Regional economic environment

Public investment per capita (t-1) 0.0207** (0.008) 0.0186* (0.009) 0.021* (0.009) 0.0238* (0.0112) 0.0216* (0.008) 0.019* (0.009) 0.028** (0.009) 0.036** (0.011)

Spatial autocorrelation control

03 03 03

CD 73 O

D m O m

□ö m TD

Spatial spillover effects .F-value

Hansen J statistic Observations

0.467** (0.114) 213.78** na.

0.464** (0.115) 159.79** n.a.

0.279** (0.081) 0.279** (0.081) 0.505** (0.113) 0.506** (0.114) 0.266** (0.077) 0.251** (0.077)

166.47** 0.513 427

127.65** 0.406 427

196.02** n.a.

147.07** na.

175.65** 5.692* 427

130.15** 5.661* 427

* p < .05; ** p < .01 Standard errors are in parentheses.

FE, fixed effects; GDP, gross domestic product; GMM, generalized method of moments; MES, mean establishment size; n.a., not applicable; R&D, research and development.