Growth of new start-up firms: evidence from the Japanese manufacturing industry

Abstract

This paper investigates firm growth as the post-entry performance of new start-up firms. Using data on Japanese manufacturing firms founded during 1992–1996, the determinants of growth among new start-up firms are identified. The effect not only of firm-specific characteristics, but also of entrepreneur-specific, industry-specific and local characteristics on firm growth are examined. It is found that younger and small-sized firms are more likely to grow among the start-ups. It is also found that entrepreneur's age and educational background affect the growth of start-ups. It is not shown, however, that industry specialization induces the growth of start-ups.

Acknowledgments

This research was done in part while I was at the Department of Economics, University of California, Berkeley. I would like to thank Bronwyn H. Hall and Chris Shannon for advising me and for hosting me as a visiting scholar. I am also grateful to Nobuyuki Harada, Hideaki Miyajima, Hiroyuki Odagiri, Shujiro Urata and Takehiko Yasuda for valuable suggestions. This research was supported by a grant from the Japan Society for the Promotion of Science. Needless to say, any remaining errors are my own.

Notes

¹ For example, Khemani and Shapiro (1986) identified the determinants of new firm entry in Canada. For a survey of empirical studies on entry in the industry-level estimation, see Siegfried and Evans (1994).

² In order to confirm the presence of Gibrat's law, which hypothesizes that growth is independent of size, a number of empirical studies have examined the relationship between the growth and size of firms or plants. While some studies (for example, Simon and Bonini, 1958) supported this law, subsequent studies (for example, Evans, 1987a, b; Hall, 1987) have shown the negative relationship between growth and size. For a review of these studies, see Sutton (1997).

³ Heshmati (2001) examined the effect of human capital and firm-specific characteristics on the growth of micro and small firms, although its sample was not restricted to start-ups either.

⁴ It is noted that the samples of these studies were not restricted to start-ups.

⁵ For more details of the sample selection model, see, for example, Amemiya (1985) and Greene (1999).

⁶ Data on profits was not used because there were many missing data in the sample.

⁷ The TSR Data Bank Service provides the exit date only if the firm exits due to bankruptcy.

⁸ The age of firms in the sample is at most 8 years when we obtained our sample.

⁹ The regulation of minimum paid-in capital was added to the Japanese commercial law on 1 April 1991. At least 10 million yen is required as paid-in capital when a joint-stock company is founded after 1 April 1991. Therefore, it is often said that a number of joint-stock companies had been founded before 31 March 1991. It is noted that the foundation period in our sample corresponds to after April 1991.

¹⁰ The exits include not only bankrupts, but also exits with solvency. Except for the firms regarded as survival or exit, 73 firms were extinct to be merged. However, since the merged firms do not actually exit and it is difficult to identify newly established firms, the firms were excluded in our sample.

¹¹ While the Census of Manufactures is published every year, the Establishment and Enterprise Census of Japan is usually published every 5 years.

¹² Baumol (1959) proposed a theoretical model that maximizes gross sales subject to earning adequate profits.

¹³ From the data source, one can obtain data on sales (thousands of yen) in 1997, 1998 and 1999 or in 1998, 1999 and 2000 based on the accounting period of firms. Apparently, all the firms do not have the same accounting month and, hence, one controls for the difference of the accounting period between firms. If the accounting month for a firm is the Mth month, then, for example, S ₁₉₉₇ for the firm is given by:

where and are sales in 1987 and 1998 based on its accounting period, respectively. If data for both of the successive accounting periods are not available, one obtains a proxy with data for either of the periods. In addition, some firms have changed the accounting month during the observation period. If the accounting month is changed, then the number of months until the new accounting month is substituted for the value of 12 in the above equation.

¹⁴ The values of S ₁₉₉₉ for some firms are not so much different from those of S ₁₉₉₈, partly because, as already explained, these values are recalculated by sales data based on the accounting periods. Therefore, the variable for firm growth is measured by sales growth for 2 years.

¹⁵ A time lag is not taken and, hence, it is here assumed that the growth of start-ups does not affect industry growth, since its effect is comparatively small.

¹⁶ See http\\www.meti.go.jp\english\information\data\cPlan010525e.html.

¹⁷ When the independent variables in columns (i) and (ii) of Table 6 regress on the inverse Mills ratio, the values of are 0.924 and 0.952, respectively. Thus, collinearity problems may occur in the models.

¹⁸ Entrepreneurs’ educational background was also measured by a dummy variable for post-graduate school, but its coefficient was not significant. The ratio of entrepreneurs finally enrolled in post-graduate school, however, is merely about 0.5% in the sample.

¹⁹ Bates (1990) argued that the level of owner education is a major determinant of the loan amounts that commercial banks extend to small business formations.

²⁰ When using firm's age or its logarithm, we found the negative relationship as well.

²¹ Porter (1990) emphasized the presence of local competition in order to foster the pursuit and rapid adoption of innovation, and Feldman and Audretsch (1999) found that local competition drives innovation. When including LCOMP in the model of firm growth, however, we did not obtain a significant result.

²² In addition, the dummy variables for major metropolitan areas subsequent to Tokyo, such as Osaka and Yokohama, were included in the model of firm growth, but their coefficients were not significant.