The rate of hazard confronting new firms and plants in U.S. manufacturing

Review of Industrial Organization 9: 41-56, 1994 © 1994 Kluwer Academic Publishers. Printed in the Netherlands.

The Rate of Hazard Confronting New Firms and Plants in U.S. Manufacturing

D A V I D B. A U D R E T S C H * CEPR and Wissenschaftszentrum Berlin far Sozialforschung, Reichpietschufer 50, 10785 Berlin, Ger- many

and

T A L A T M A H M O O D * Wissenschaftszentrum Berlin far Sozialforschung, Reichpietschufer 50, 10785 Berlin, Germany

Abstract. Three different factors are hypothesized to shape the hazard function confronting new businesses - the extent of scale economies relative to start-up size, the technological environment, and ownership structure. Using a longitudinal data base tracking the post-entry performance of more than 12,000 U.S. manufacturing establishments, a semi-parametric Cox regression model is used'to estimate the hazard function. The evidence suggests that while the presence of high scale economies, a high- technological environment, and a relatively small initial start-up size tend to elevate the exposure of risk confronting new businesses, these factors apparently exert no influence on the likelihood of survival for new branches and subsidiaries established by existing enterprises.

Key words: Entry, innovation, survival, new firms, hazard, exit.

I. Introduction

Nearly thirty years ago, Mansfield (1962, p.1023) made a plea for a greater emphasis on intra-industry dynamics:

Because there have-been so few econometric studies of the birth, growth, and death of firms, we lack even crude answers to the following basic questions regarding the dynamic processes governing an industry's structure, What are the quantitative effects of various factors on the rates of entry and exit? What have been the effects of successful innovations on a firm's growth rate? What determines the amount of mobility within an industry's size structure?

Within the last several years a wave of empirical studies has emerged attempting to fill the gap of knowledge in industrial organization about the process by which firms, plants, and entire industry" structures evolve over time. For example, following the example of Orr's (1976) analysis of entry in manufacturing industries, Geroski and Schwatbach (1991) recently compiled a collection of systematic studies identifying the determinants of industry entry rates across a broad spectrum of studies. 1

However, the studies identifying the determinants of industry entry rates generally leave an important question unanswered: "What happens to new firms and

42 DAVID B. AUDRETSCH AND TALAT MAHMOOD

plants subsequent to their establishment?" While the small but growing number of studies attempting to identify the extent and determinants of industry exit rates 2 might appear to answer this question at first glance, this is clearly not the case for three related reasons. First, the unit of analysis has been virtually always at the aggregated industry level and not at the unit of the specific establishment or enterprise. Second, most studies have generally focused on the behavior and determinants of rates of exit, which is typically defined as the number (either gross or net) of firms leaving the industry within a specified time period, divided by the number of firms in the industry. 3 Third, establishments of all ages are included in the analysis, with no controlling factors for age. 4

These limitations result in somewhat perplexing conclusions. For example, the consistent finding that industry exit rates tend to be positively related to market growth seemingly implies that the propensity for new firms to fail in high-growth industries is relatively high. However, an alternative interpretation is that either entry in such industries also tends to be relatively high (as has also generally been demonstrated), or else technological or some other type of change is resulting in industrial restructuring, whereby the new firms are more likely to employ different production or management techniques. 5

The purpose of this paper is to overcome these limitations by providing the application of the semi-parametric Cox-Regression technique to U.S. manufacturing firms and plants, compiled from a longitudinal data base consisting of virtually all enterprise and establishment start-ups in 1976 and by tracking the performance of those start-ups over the subsequent decade. This enables us to test the hypotheses that the rate of hazard confronting new establishments is influenced not only by their evolution over time, but also by the extent of scale economies, the technological environment, as well as particular establishment characteristics, such as start-up size and ownership structure.

In the following section of this paper three distinct strands of literature are drawn upon to identify major factors influencing the decision to exit from an industry. In particular, the extent of scale economies relative to start-up size, the technological environment, and ownership status are hypothesized to shape the likelihood of exit. The exact nature of the data base and important measurement considerations are described in the third section. In the fourth section the Cox- Regression technique is introduced and applied to new start-ups in U.S. manufacturing. After the empirical results are presented in the fifth section, a summary and conclusion are provided in the final section.

Not only do we find that the extent of scale economies, the technological environment, and certain establishment-specific characteristics, such as start-up size, influence the risk exposure confronting new businesses, but that the exact nature of these relationships depends upon the ownership status of the establishment as well as the technological environment in which it was founded. In particular, scale economies and a high-technological environment are found to elevate the risk exposure of new-firm startups. These inherent disadvantages confronting

HAZARD CONFRONTING NEW FIRMS AND PLANTS 43

new firms can be offset at least somewhat by a greater start-up size. However, neither the extent of scale economies nor the start-up size of branches and subsidiaries belonging to existing enterprises is found to influence their likelihood of exit.

II. Why Do Businesses Exit?

Three different strands of literature have identified several major influences shaping the decision to exit out of an industry. The first, and most obvious strand of literature (Weiss, 1964; 1976; and 1979; and Caves, et at., 1975) suggests that the probability of a business exiting will tend to increase as the gap between its level of output and the minimum efficient scale (MES) level of output increases. 6 That is, given a long-run average cost function and MES in an industry, the smaller the scale at which an enterprise operates, the greater will be its cost disadvantage. Conversely, for any business of a given size, higher levels of the MES will result in a greater cost disadvantage. As the gap between the size of the business and the MES increases, and the cost disadvantage confronting the business correspondingly increases, the likelihood of exit increases.

Of course, to the extent that prices become elevated above long-run average cost in an industry, the cost disadvantage confronting a sub-optimal scale business is diminished. An important finding of Bradburd and Caves (1982) is that industry growth can contribute to elevated price-cost margins. Thus, the likelihood of exit for any given sub-optimal scale business will tend to diminish in markets where growth is greater.

The second strand of literature points to the role that the technological environment plays in shaping the decision to exit. The technological environment is expected to affect the likelihood of exit in two ways. First, as Mueller and Tilton (1969) and Cohen and Klepper (1990 and 1991) point out, R&D intensive industries increase the amount of financial requirements needed to operate in the industry. Second, as Dosi (1988) and Arrow (1962) argue, an environment characterized by frequent product innovation may also be associated with a greater amount of uncertainty regarding not only the technical nature of the product but also the demand for that product. As the technological uncertainty increases, the likelihood that the business will be able to produce a viable product and ultimately be able to survive tends to decrease.

The third strand of literature argues that ownership status will tend to influence the decision to exit from an industry. Caves and Porter (1976, p. 43) argue that the owners of independent plants have a lower opportunity cost and are therefore willing to accept a lower rate of return than is generally set for establishments belonging to a multi-plant enterprise: "Diversification can remove or combat managerial sources of exit barriers. Multi-industry firms may be controlled by their top managements, but the evaluation and removal of managers of the firm's individual business units is the responsibility of top management. Top management may maintain an effective internal capital market, reviewing divisional perfor-


mance internally and facilitating managerial changes or exit from the business when necessary. The possibility of internal placement of employees displaced from an extinguished business makes it easier for top management to wield the axe. And the greater the breadth and extent of the firm's diversification, the less threat does exit from one industry represent to its continuity, the more dispassionate can be the decision to exit". Their prediction is consistent with the theories by Reyn- olds (1988) and Baden-Fuller (1989) that multi-plant enterprises will be more likely to close subsidiary establishments than are independently owned plants to decide to exit from the industry.

These three factors - the extent of scale economies relative to the start-up size of the business, the technological environment, and ownership structure - are linked to the likelihood of exit using the semi-parametric Cox hazard regression model in the fourth section.

HI. Measurement Issues

The greatest constraint to analyzing the post-entry performance of new businesses has been the lack of longitudinal data sets comprised of individual plants and firms that identify the actual start-up and closure dates. While Dunne, Roberts, and Samuelson (1988 and 1989), Evans (1987a and 1987b), Hall (1987), Phillips and Kirchhoff (1989), and Baldwin and Gorecki (1991) all had access to such a longitudinal data set, none of these studies explicitly estimated the hazard rate model. One reason why the U.S. Bureau of Census data employed by Dunne, Roberts, and Samuelson do not lend themselves to estimation of the hazard model, is that while observations over time are available, they are only identified at five-year intervals.

Thus, we employ a data set which provides bi-annual observations on firms and plants - the U.S. Small Business Administration's Small Business Data Base (SBDB). The data base is derived from the Dun and Bradstreet (DUNS) market identifier file (DMI), which provides a virtual census on about 4.5 million U.S. business establishments every other year between 1976 and 1986.

While the raw Dun and Bradstreet data have been subject to considerable criticism (FitzRoy, 1989; Armington and Odle, 1982), the SBDB data have been adjusted by the U.S. Small Business Administration to clean up the raw data in the original DMI files. Several important studies have compared the SBDB data with analogous measures from the establishment data of the U.S. Census of Manufactures (Boden and Phillips, 1985; Acs and Audretsch, 1990, Chapter Two), and from the establishment and employment records of the Bureau of Labor Statistics (Brown and Phillips, 1989) and have concluded that the SBDB data are generally consistent with these other major data bases providing observations on firms and plants.

The SBDB has already been applied in a number of other studies to address a wide variety of issues. For example, Macdonald (1986) analyzed the impact of

H A Z A R D CONFRONTING NEW FIRMS AND PLANTS

Table I. Life table of establishment founded in 1976

45

Time interval Survival hazard Survival hazard Survival hazard Survival hazard rate rate rate rate

1976-1978 - 0 254 - 0.222 - 0.225 - 0.252 1978-1980 0.775 0.205 0.800 0.234 0.774 0.204 0.776 0.202 1980-1982 0.631 0.354 0.632 0.274 0.631 0.358 0.633 0.358 1982-1984 0.441 0.164 0.479 0.116 0.439 0.166 0.441 0.116 1984-1986 0.374 0.166 0.427 0.167 0.372 0.166 0.373 0.165 1986 0.316 - 0.361 - 0.314 - 0.316 -

No. of obs 12251 590 11661 11154 Sample All Branches All firms Single-

establishments establishment firms

entry and exit, Evans (1987a and 1987b) analyzed the relationship between firm growth and size, Phillips and Kirchhoff (1989) tested the hypothesis that plant growth is dependent upon age, and Acs and Audretsch (1989a, 1989b, and 1990) identified the determinants of market entry and industry turbulence.

The essential building block and unit of observation in the SBDB is the establishment, which is defined as a particular economic entity operating at a specific and single geographic location. While some establishments are legally tied to parent firms through either a branch or subsidiary relationship, other establishments are independent and therefore are, in fact, firms (enterprises) in their own right. Establishments in the manufacturing sector are frequently referred to as plants. The data base links the ownership of all establishments to any parent firms, thereby enabling the performance of establishments which are independent firms to be distinguished from those which are branches and subsidiaries of parent firms. Thus, the data base makes it possible to identify if each record, or establishment, is (1) a single-establishment firm, in which case the establishment is an independent legal entity, as explained above, (2) a branch or subsidiary belonging to a multi- establishment firm, or (3) the headquarters of a multi-establishment firm. Besides a detailed identification of the ownership structure of each establishment, the USELM file of the SBDB links the performance of each establishment within two- year intervals beginning in 1976 and ending (under the current version) in 1986, thereby tracking each establishment over what constitutes a ten-year longitudinal data base.

IV. Regression Method

The techniques of survival analysis (Lawless, 1982) are used to analyse the SBDB data for compilation of the Life Tables (non-parametric procedure) for U.S. establishments which were founded in 1976. Two important functions, shown in Table I, are the survival and hazard functions. The survival function gives for each

46 DAVID 13". AUDRETSCH AND TALAT MAHMOOD

Table II. Life Table according to technological environment

Time interval High-tech a Moderate-tech b Low-tech c Survival hazard rate Survival hazard rate Survival hazard rate

1976-1978 - 0.239 - 0.235 - 0.268 1978-1980 0.786 0.225 0.789 0.196 0.763 0.209 1980-1982 0.627 0.379 0.647 0.315 0.618 0.381 1982-1984 0.427 0.187 0,471 0.139 0.421 0.181 1984-1986 0.354 0.155 0.410 0.165 0.351 0.168 1986 0.303 - 0.347 - 0.296 -

No,of obs. 944 4751 6556

a Includes b Includes

Includes

industries where R&DtSales _->5%. industries where 1% < R&D/Sales < 5%. industries where R&D/Sales < t%.

time period the share of those establishments founded in 1976 which still existed. The hazard function gives for each time point the risk of failure, i.e. the (con-

ditional) probabili ty that an establishment will exit in the next t ime interval, on the condition that this establishment had survived up to the beginning of the time interval.

While the first column in Table I provides the survival and hazard rates for all

establishments founded in 1976, the second column includes only those establishments which were identified as being a branch or subsidiary of an existing multi-

establishment enterprise. The survival and hazard rates for all new firms, but not

branches and subsidiaries are indicated in the third column. Finally, only those

new establishments representing single-establishment firms (excluded are not only new subsidiaries and branches of multi-plant firms, but also establishments which are headquarters of new multi-establishment enterprises) are included in the sam-

ple analyzed in the last column. The extent to which the technological environment , or what Scherer and Ross

(1990) term as the "technological opportunity class", influences the ability of new

establishments to survive is shown in Table II. Those 944 establishments opened in 1976 in industries with an R&D/Sales ratio of at least five percent are included in the "high-technological" opportunity class. By contrast, those 6,556 new estab-

lishments in industries with R&D/Sales ratio less than one percent are classified as being in a "low-technological" opportuni ty class. The remaining 4,751 establishments started in 1976 are considered to be in the "moderate- technological" opportunity class. Table I I shows that the mean survival rates of establishments in the high-and low-technological opportunity classes are considerably less than the mean survival rate for new establishments in the moderate-technological opportuni ty

class. The Cox-Regression model (Cox, 1972; Kiefer, 1988) captures the effects of the

covariates (or explanatory variables) upon death (hazard rates) rather then upon times to death. In addition, it corrects for the problem of censored data.

H A Z A R D CONFRONTING NEW FIRMS AND PLANTS 47

The model is defined in terms of h (t;x), where h is the hazard rate for a business, and x is a vector of covariates. The hazard model is given by,

h(t;x) = ho(t).exp(/3x')

o r

ln[h(t;x)/ho(t)] =/3x'

(1)

(2)

where/3 is a vector of unknown regression coefficients and ho(t) is an unknown non-negative base line hazard rate. The second exponential term incorporates the covariates vector x. Estimates of the regression parameters are obtained as follows: let h < t2 < . . . t~ represent distinct times to death among n observed survival times.

The conditional probability that the ith firm exits at time t /with a covariate vector x~, given that a single exit has occurred at ti, is given as the ratio of the hazards,

exp(flxi')! ~ exp(/3x/) (3) j ~Ri

where ] E Ri corresponds to those establishments which are just at risk prior to time t~. The baseline hazard rate is assumed to be the same for all the observations, and hence it cancels out.

The partial likelihood function derived from Cox (1972 and 1975) is obtained by multiplying these probabilities together for each of the k incidences of exit,

k

PL(/3, X l , . . . , x,, = 1-I (exp(flxi')/~ exp(/3xi') (4) ]~Ri

Maximization of the partial likelihood function yields estimators of/3 with proper- ties similar to those of usual maximum-likelihood estimators, such as asymptotic normality.

The estimated regression coefficient indicates the relationship between the covariate and the hazard function. A positive coefficient increases the value of the hazard function and therefore indicates a negative relationship with survival. A negative coefficient has the reverse interpretation.

In the case of ties among the time until exit, Breslow (1974) proposed to maximize the following likelihood function,

PL(/3, Xl . . . . , x,,) = [exp(/3s~' exp(/3x~')mi] (5) i=1 i i ~ R i

where mi is the number of exits at ti and s~ is the vector sum of covariates of the mi businesses. The SBDB data we introduced in the previous section satisfy these criteria, because they provide enough cases of censored observations and ties, so that the hazard duration function can be estimated with Equation (2).


V. Empirical Results

The Cox-Regression technique described in the previous section is used to test the hypotheses that the exposure to risk of new establishments is influenced by the extent to which scale economies play a role in the relevant industry, the initial start-up size of the establishment, market growth, the technological environment, and the ownership structure of the establishment. Measurement of the minimum scale of output required to exhaust scale economies in an industry has proven to be challenging at best (Scherer, t990; and Caves et al., 1975). Here we adapt the standard Comanor-Wilson (1967) proxy for measuring MES, which is defined as the mean size of the largest plants in each industry accounting for one-half of the industry value-of-shipments, 1977. This measure is derived from the Census of Manufactures of the U.S. Bureau of the Census. While the Comanor-Wilson measure is crude, it has proven in numerous studies at least to reflect the extent to which scale economies play an important role in an industry (Scherer and Ross, 1990). The MES, which is measured in terms of thousands of dollars, is expected to exert a positive influence on the rate of hazard confronting new establishments.

The most reliable and consistent measure of the size of the establishment when it was founded is the number of employees. As explained in the previous section, a larger start-up size is expected to reduce the hazard rate. Market growth is measured as the percentage change in the employment of the four-digit standard industrial classification (SIC) industry within which the establishment operated between 1972 and 1977. This measure is derived from the annual Survey of Manufactures of the U.S. Bureau of the Census. Market growth is expected to increase the growth potential of new establishments, and therefore should decrease the degree of risk confronting them.

To measure the importance of technology in the industry, the 1977 Federal Trade Commission's Line of Business company R&D/Sales ratios are used. An important finding within the last several years is that under certain technological conditions, large firms tend to contribute more of the innovative activity while under other conditions the small firms have the innovative advantage (Acs and Audretsch, 1987 and 1988). In their 1990 study (Chapter Seven), Acs and Aud- retsch found that industries where the small firms have the innovative advantage tend to correspond with what Winter (1984) termed as the "entrepreneurial regime", while industries where the large firms have the innovative advantage more closely resemble Winter's (1984 concept of the "routinized regime". According to Winter (1984), the entrepreneurial regime is particularly conducive to innovative activity by newly established enterprises, while the routinized regime tends to impede innovations from entrants. 7 Under the entrepreneurial regime new enterprises are more likely to enter an industry in the hopes of successfully innovat- ing. However, those new start-ups that are unable to innovate or adapt in some other manner that enhances growth at least to the MES level of output will be


forced to exit from the industry. Thus, under the entrepreneurial regime, or where innovative activity tends to emanate more from the small firms than from the large enterprises, the hazard rate is expected to be greater than under the routinized regime, where the large firms tend to have the innovative advantage.

We employ the same innovation measures as those introduced by Acs and Audretsch (1987 and 1990) to reflect Winter's notion of the technological regime. A complete and detailed description of the Small Business Administration's Inno- vation Data Base can be found in Chapter Two of Acs and Audretsch (1990). The total innovation rate is measured as the total number of innovations in 1982 divided by industry employment. Similarly, dividing the number of innovations contributed by small firms 8 by small-firm employment yields the small-firm innovation rate. The innovation rates are employed rather than the absolute numbers. of innovations in order to standardize the amount of innovative activity for the size of the industry, as well as for the relative economic activity accounted for by large and small enterprises. 9 When the small-firm innovation rate is large relative to the total innovation rate, the industry is better characterized by the entrepreneurial regime. By contrast, the routinized regime is more reflective of the underlying technological conditions when the total innovation rate is high relative to the small-firm innovation rate.

Finally, the ownership structure of the establishment is captured by including a dummy variable taking on the value of one for branches and subsidiaries that belong to a multi-establishment enterprise and zero for establishments which represent independent firms. As suggested in the second section, the hazard rate for newly established firms is expected to be greater than that for new plants opened by existing enterprises.

Using all 7070 manufacturing establishments which were founded in 1976, and for which compatible industry characteristics could be matched, the Cox-Regres- sion model was estimated, and the results are shown in Table III. The positive and statistically significant (at the 95% level of confidence) coefficient of the measure of minimum efficient scale suggests that new establishments face a greater rate of hazard in industries where scale economies play an important role than in markets where the MES is relatively low. lO The start-up size of the establishment is apparently negatively related to the hazard rate. That is, new establishments which are larger experience a lower risk of failure, at least within ten years subsequent to start-up.

The negative coefficient of the industry growth variable confirms the hypothesis that the hazard rate tends to be lower for establishments founded in high-growth industries and greater for those in industries with low or even negative growth. Industry R&D intensity is associated with a higher rate of hazard, as indicated by the positive R&D/Sales coefficient. Similarly, the coefficient of the small-firm innovation rate divided by the total innovation rate is positive, suggesting that under the entrepreneurial regime the hazard rate confronting new establishments

50

Table III. Cox-regression results

DAVID B. AUDRETSCH AND TALAT MAHMOOD

Independent variables (1) (2) (3) (4)

Minimum efficient scale 0.004 0.004 0.005 0.004 (2.02) (2.03) (2.53) (2.13)

Start-up size -0.002 -0.001 -0.002 - (-3.23) (-3.09) (-3.25) -

Growth - 1.393 -1.390 -1.445 -1.404 (-3.15) (-3.14) (-3.27) (-3.17)

R&D/sales 0.020 0.020 -0.002 0.002 (1.57) (1.59) (-0.12) (1.61)

Total innovation rate - - 0.088 - (2.32)

Small-firm innovation rate/ 0.006 0.006 - - Total innovation rate (1.67) (1.74) Branch Dummy 0.086 - 0.083 0.092

(1.1.4) (1 09) (1.22) No. of observations 7070 7070 7070 7070 Chi square 37.1 35.8 37.2 34.6 Log of likelihood -39310 -39310 -39310 -39310

" T-statistics in parenthesis.

tends to be higher than it is under the routinized regime. That is, in industries where small firms tend to be particularly innovative relative to the innovative

activity of the entire industry, the exposure to risk of new establishments is greater. Finally, the coefficient of the branch dummy is positive but the t-statistic cannot

be considered to be significantly different f rom zero. In any case, there is little

evidence supporting the hypothesis that the hazard rate of new branches and

subsidiaries belonging to existing multi-establishment firms is any less than that for new-firm start-ups. Omitting the branch dummy in the second equation of

Table I I I leaves the results for the other variables virtually unaffected. In Equat ion 3 the total innovation rate is included as an additional measure of

technology. While the positive coefficient reinforces the conclusion that the new- establishment hazard rate is greater in high-technology environments, the coefficient of the R&D/Sales variable becomes negative and cannot be considered

statistically significant. This is probably attributable to the high simple correlation between the two measures found by Acs and Audretsch (1990, Chapter Two).

One possible explanation for the insignificance of the branch dummy variable is that the start-up size of branches and subsidiaries tends to be greater than that of new firms. That is, the mean start-up size of the branch and subsidiary establishments was 58.90 employees , .whereas the start-up size of the new enterprises in 1976 was 9.55 employees. However , when the start-up size variable is omit ted f rom the regression in Equat ion 4, the coefficient of the branch dummy variable remains virtually unchanged.

While the results from Table I I I suggest that, after controlling for scale, technological, and start-up size effects, the extent to which new establishments are

HAZARD CONFRONTING NEW FIRMS AND PLANTS

Table IV. Cox-regression results according to ownership and technological environment a

51

All establishments

Independent variables New firms Branches High-tech Moderate-tech Low tech (1) (2) (3) (4) (5)

Minimum efficient scale 0.014 0.003 -0.004 0.009 0.015 (2.15) (0.39) (-1.30) (-3.25) (3.74)

Start-up size -0.003 -0.001 -0.001 -0.001 -0.001 (-3.07) (-0.99) (-0.62) (-2.42) (-2.18)

Growth -1.621 -2.540 -1.538 -0.864 -0.726 (-3.49) (1.28) (-1.32) (-1.30) (-0.94

R&D/sales 0,027 -0.041 0,033 0.001 0,350 (2.08) (-0.77) (1.33) (0.04) (6.87)

Small-firm innovation rate/ 0.005 0.008 0.005 0.019 0.005 Total innovation rate (1.35) (1.24) (0.50) (1,23) (1.14) No. of Observations 6492 322 810 3388 2872 Chi square 40.9 4.5 3.9 23.5 62.7 Log of likelihood -35646 -1132 -3530 -16768 -14487

T-statistics in parenthesis.

exposed to risk does not significantly differ between new branches and subsidiaries of existing enterprises and new-firm start-ups, it is certainly possible that the establishment ownership status effects the relationships multiplicatively or in an interactive manner. Therefore, the first two equations of Table IV contain all new firms (Equation 1) and new branches and subsidiaries of existing enterprises (Equation 2). 11

There are five important points to emphasize in comparing the determinants of the hazard functions between new-firm start-ups and the opening of new plants by existing firms. First, the hazard rate tends to be higher in the presence of scale economies for new-firm start-ups but not for branches and subsidiaries. Second, newly established firms can apparently reduce their hazard rate by increasing their start-up size. The rate of hazard of new establishments belonging to multi-plant firms is not significantly influenced by the initial plant size.

Third, although a high market growth rate reduces the exposure to risk confronting new firms, it has no apparent effect on branches and subsidiaries. Simi- larly, the hazard rate of new firm start-ups is significantly and positively related to R&D intensity, while R&D/Sales apparently exerts no influence on the ability of new plants opened by existing firms to survive. Finally, the impact of the technological regime, or the relative innovative advantage of large and small firms, on the hazard rate is virtually identical, regardless of ownership status.

In general, "the distinct differences between the two hazard rate functions suggest that new-firm start-ups are much more influenced from the external environment, and in particular, from the extent of scale economies and from the technological environment than are new branches and subsidiaries of existing firms. Perhaps the importance of the initial start-up size of new firms is attributable to its role as a


mechanism for offsetting otherwise inherent scale and technology disadvantages. By contrast, the start-up size contributes little towards reducing the exposure to risk confronting new branches and subsidiaries, which apparently are able to exploit the advantages of belonging to an established multi-plant firm.

Because of the important role that technology plays, the Cox-Regression model is also used to estimate the hazard duration rate for new establishments in high- tech, moderate-tech, and low-tech industries 12 There are two important results emerging from Equations 3, 4, and 5 in Table IV. First, while the existence of scale economies tends to raise the hazard rate confronting new establishments in low-and moderate-tech industries, this is clearly not the case in high-tech industries. Second, the start-up size is apparently important in reducing the hazard rate for new establishments in low-and moderate-tech industries, but the exposure to risk faced by new establishments is not influenced by start-up size in high-tech industries. This would suggest that in a high-technological environment, initial size and scale considerations do not seem to play an important role in the ability of the establishment to survive. Rather, innovation is presumably the more important factor in such industries.

VI. Conclusions

One of the striking findings that has emerged in the growing empirical literature of entry in industrial markets is that entry is apparently promoted in industries where small firms tend to have the innovative advantage (Acs and Audretsch, 1989a, 1989b, and 1990). However, the results of this paper strongly suggest that, while entry might be greater in such markets, new establishments are actually exposed to a greater degree of risk. Such industries, which Winter (1984) characterized as the entrepreneurial regime, are particularly turbulent. Not only are entry rates relatively high, but the correspondingly high rate of hazard reflects the high propensity of new establishments to fail within the first few years of their existence. An important characteristic of the entrepreneurial regime, where small new entrants tend to contribute much of the innovative activity, is the ac- companying high rate of firm failure and turbulent industry structure.

The conclusion of Dunne et al. (1988) that the ownership structure of establishments affects their ability to survive is confirmed in this paper. However, the results of the Cox-Regression estimation at the individual establishment level reveal that after controlling for scale, technological, and initial size effects, this ownership structure differential in hazard rates disappears. What does emerge is that the roles that scale economies, technology, and the start-up size play apparently differ depending upon the ownership structure. That is, the hazard rate of new enterprises is greater in the presence of scale economies, a high-technology environment, and low market growth. A large start-up size can at least somewhat offset these inherent disadvantages by reducing the exposure to risk. By contrast, the hazard rate confronting new branches and subsidiaries opened by existing


firms does not appear to be significantly influenced by the existence of scale economies, technology, or the start-up size. That the ten-year survival rate of the branches and subsidiaries established in 1976 is about fifteen percent greater than that for the newly established firms can be attributed to the influence that scale economies, the technological environment and initial start-up size exert on new firms but not on new establishments belonging to multi-plant enterprises.

Finally, the risk of failure confronting all new establishments is conditional upon the technological environment. The scale and start-up size disadvantages confronting new establishments do not play an important rote in high-technological industries, but rather only in low-and moderate-technological markets.

The answer to the question, "What happens to new establishments subsequent to entry?" is "It depends". The results of this paper make it clear that one key element of intra-industry dynamics - the rate of hazard confronting new establishments -is invariably shaped by specific factors external to the establishment, such as the presence of scale economies and the technological environment, as well as factors internal to the establishment, such as the start-up size and the ownership structure.

One important limitation in this study is the restricted sample of plants and firms established in a single year. Future research needs to access a large panel to identify the manner by which factors internal and external to the establishment interact with age to shape the firm and plant trajectories within a dynamic context.

Notes

* We wish to thank Ron Braeutigam, Bruce Kogut, Paul Geroski, Dennis Mueller, and two anonymous referees for their helpful comments, and Jianping Yang for his diligent computational assistance. All errors and omissions remain our responsibility. ** The paper was presented at the 18th Annual Conference of EARIE in September, 1991, Ferrara, Italy. 1 Other aspects of intra-industry dynamics which have been empirically explored in the literature include, the relationship between firm and plant size and their respective growth rates (Hall, 1987) and Evans (1987a and 1987b). Similarly, Phillips and Kirchhoff (1989), Mahmood (1992), Dunne et al. (1989), and Audretsch (1991), Preisend6rfer et al. (1989), and Brtiderl and Schiissler (1990) examine the relationship between establishment age and survival. And the issue of turbulence has been addressed by Beesley and Hamilton (1984), Acs and Audretsch (1990, Chapter Seven)i and Khemani and Shapiro (I988). z See for examples Baldwin and Gorecki (1991), Cable and Schwalbach (1991), Caves and Porter (1976), Dunne, Roberts, and Samuelson (1989), Macdonald (1986), and Marcus (1967). 3 Exit rates are generally used rather than the absolute number of exiting establishments in order to control for industry size, thereby facilitating reliable inter-industry and in certain cases cross-national comparisons. 4 That is, the propensity of all establishments to exit from the industry is being measured, and not just the propensity of new establishments to exit from the industry. In addition, the implicit reference group is all of the existing establishments in the industry and not just the new establishments. 5 For example, Gort and Klepper (1982), Carlsson (1989), Geroski and Pomroy (1990) and Dosi (1988) follow the long tradition in industrial organization dating back to at least Blair (1948) in arguing that the establishment of new firms will result from changes in technology precipitating a shift in the firm- size distribution.


6 For example, Weiss (1976, p. 126) argues that, "In purely competitive long-run equilibrium, no suboptimal capacity should exist at all". 7 The distinction between the entrepreneurial and routinized regimes emanates from the underlying sources of information leading to innovative activity. If information based on non-transferable experience in the market is an important input in generating innovative activity, then incumbent firms will tend to have the innovative advantage over new firms. This is consistent with Winter's (1984) notion of the routinized regime, where the accumulated stock of non-transferable information is the product of experience within the market, which firms outside of the industry, by definition, cannot posses. By contrast, when information external to the industry is a relatively important input in generating innovative activity, newly established firms will tend to have the innovative advantage over incumbent firms (see Arrow, 1962; Gort and Klepper, 1982; Williamson, 1975; and Mueller, 1976). s Small firms are defined as enterprises with fewer than 500 employees. 9 The adoption of the innovation rate and small-firm innovation rate is the same as used by Acs and Audretsch in their 1987 and 1990 studies. 10 It should be emphasized that the MES is measured in terms of thousands of dollars, so that an increase In the MES of one thousand dollars results in an increase in the hazard rate of 0.004. u While none of the coefficients of the variables included in Equation (2) of Table IV can be considered to be statistically significant, we include the hazard function for branches and subsidiaries to distinguish between that for new firms. lz The classification of industries according to the technological environment is identical to that used in Table II. It should be noted that all new establishments (both firms as well as branches and subsidiaries) are included in the sample used to estimate Equations 3-5.

References

Acs, Zoltan J. and David B. Audretsch (1987) 'Innovation, Market Structure and Firm Size', Review of Economics and Statistics 69, 567-575.

Acs, Zoltan J. and David B. Audretsch (1988) 'Innovation in Large and Small Firms: An Empirical Analysis', American Economic Review 78, 678-690.

Acs, Zoltan J. and David B. Audretsch (1989) 'Small-Firm Entry in U.S. Manufacturing', Economica 56,255-265.

Acs, Zoltan J. and David B.Audretsch (1989) 'Birth and Firm Size', Southern Economic Journal 56, 467-475.

Acs, Zoltan J. and David B. Audretsch (1990) Innovation and Small Firms, Cambridge, MA: MIT Press.

Arminton, Catherine and Marjorie Odle (1982) 'Small Business - How Many Jobs?', The Brookings Review 1, 1417.

Arrow, Kenneth J. (1962) 'Economic Welfare and the Allocation of Resources for Innovation', in R. R. Nelson, (ed.), The Rate and Direction o f Inventive Activity, Princeton, NJ: Princeton University Press, pp. 609-626.

Audretsch, David B. (1991) 'New-Firm Survival and the Technological Regime', Review of Economics and Statistics 60, 441-450.

Baden-Fuller, C. W. F. (1989) 'Exit from Declining Industries and the Case of Steel Castings', Economic Journal 99 949-961.

Baldwin, John and Paul Gorecki (1991) 'Entry, Exit and Productivity Growth', in Paul Geroski and Joachim Sehwalbach (eds.), Entry and Market Contestability: An International Comparison, Oxford: Basil Blackwell.

Beesley, M.E. and R.T. Hamilton (1984) 'Small Firms' Seedbed Role and the Concept of Turbulence', Journal of Industrial Economics, 333,217-232.

Blair, John M. (1948) 'Technology and Size', American Economic Review 38, 121-152. Boden, Richard and Bruce D. Phillips (1985) 'Uses and Limitations of USEEM/USELM Data', Office

of Advocacy, U.S. Small Business Administration, Washington, DC. Bradburd, Ralph and Richard E. Caves (1982) 'A Closer Look at the Effect of Market Growth on

Industries' Profits', Review of Economics and Statistics 64, 635-645. Breslow, N. (1974) 'Covariance Analysis of Censored Survival Data, Biometrics, 30, 88-99.


Braderl, Josef and Rudolf Schfissler (1990) 'Organizational Mortality: The Liabilities of Newness and Adolescence', Administrative Science Quarterly 35, 530-547.

Cable, John and Joachim Schwalbach (1991) 'International Comparisons of Entry and Exit', in Paul Geroski and Joachim Schwalbach (eds.). Entry and Market Contestability: An International Compari- son, Oxford: Basil Blackwell.

Carlsson, Bo (1989) 'The Evolution of Manufacturing Technology and its Impact on Industrial Struc- ture: An International Study', Small Business Economics 1, 21-38.

Caves, Richard E., J.Khalilzadeh-Shirazi, and M.E. Porter (1975) 'Scale Economics in Statistical Analyses of Market Power', Review of Economics and Statistics 57, 133-140.

Caves, Richard E. and Michael E. Porter (1976) 'Barriers to Exit', in Robert T. Masson and P. David Qualls (eds.), Essays on Industrial Organization in Honor of Joe S. Bain, Cambridge: Ballinger.

Cohen, Wesley M. and Steven Klepper (1991) 'Firm Size versus Diversity in the Achievement of Technological Advance', in Zoltan J. Acs and David B. Audretsch (eds.), Innovation and Techno- logical Change: An International Comparison, Ann Arbor: University of Michigan Press, pp. 183- 203.

Cohen, Wesley M. and Steven Klepper (1992) 'The Tradeoff between Firm Size and Diversity in the Pursuit of Technological Progress', Small Business Economics 4, 1-14.

Comanor, Williams S. and Thomas A. Willson (1967) 'Advertising, Market Structure, and Performan- ce', Review of Economics and Statistics 49, 423-440.

Cox, David R. (1972) 'Regression Models and Life-Tables (with discussion)', Journal of the Royal Statistical Society 34, 187-220.

Cox, David R. (1975) 'Partial Likelihood', Biometrics 62, 269-275. Dosi, Giovanni (1988) 'Sources, Procedures, and Microeconomic Effects of Innovation', Journal of

Economic Literature 26, 1120-1171. Dunne, Timothy, Mark, J. Roberts, and Larry Samuelson (1988) 'Patterns of Firm Entry and Exit in

U.S. Manufacturing Industries', Rand Journal of Economics 19, 495-515. Dunne, Timothy, Mark, J. Roberts, and Larry Samuelson (1989) 'The Growth and Failure of U.S.

Manufacturing Plants', Quarterly Journal of Economics 104, 671-698. Evans, David S. (1987) 'The Relationship Between Firm Growth, Size, and Age: Estimates for 100

Manufacturing Industries', Journal of Industrial Economics 35, 567-581. Evans, David S. (1987) 'Tests of Alternative Theories of Firm Growth', Journal of Political Economy

95, 654-674. FitzRoy, Felix R. (1989) 'Firm Size, Efficiency and Employment: A Review Article', Small Business

Economics 1, 75-80. Geroski, Paul and R. Pomroy (1990) 'Innovation and Evolution of Market Structure', Journal of

Industrial Economics 38, 299-314. Geroski, Paul and Joachim Sehwalbach (eds.) (1991) Entry and Market Contestability: An International

Comparison, Oxford: Basil Blackwell. Gort, Michel and Steven Klepper (1982) ~Time Paths in the Diffusion of Product Innovations?',

Economic Journal 92, 630-653. Hall, Bronwyn H. (1987) 'The Relationship Between Firm Size and Firm Growth in the U.S. Manufac-

turing Sector', Journal of Industrial Economics 35, 583-605. Jovanovie, Boyan (1982) 'Selection and Evolution of Industry', Econometrica 50, 649-670. Kiefer, Nicholas M. (1988) 'Economic Duration Data and Hazard Functions', Journal of Economic

Literature 26, 646-679. Lawless, Jerald F. (1982) Statistical Models and Methods for Lifetime Data, New York: Wiley. Macdonald, James (1986) 'Entry and Exit on the Competitive Fringe', Southern Economic Journal 52,

640-652. Mahmood, Talat (1992) 'Does the Hazard Rate for New Plants Vary between Low- and High-Tech

Industries?' Small Business Economics 4, 201-210. Mansfield, Edwin (1962) 'Entry, Gibrat's Law, Innovation, and the Growth of Firms', American

Economic Review 52, 1023-1051. Marcus, M. (1967) 'Firms Exit Rates and Their Determinants', Journal of Industrial Economics 10,

10-22. Mueller, Dennis C. and J. Tilton (1969) 'Research and Development Costs as a Barrier to Entry',

Canadian Journal of Economics 56, 570-579.


Mueller, Dennis (1976) 'Information, Mobility, and Profit', Kyklos 29, 419-448. Orr, Dale (1974) 'The Determinants of Entry: A Study of the Canadian Manufacturing Industries',

Review of Economics and Statistics 56, 58-66. Pakes, A. and Ericson, R. (1987) 'Empirical Implications of Alternative Models of Firm Dynamics',

Manuscript, Department of Economics, University of Wisconsin-Madison. Phillips, Bruce D. and Bruce A. Kirchhoff (1989) 'Formation, Growth and Survival: Small Firm

Dynamics in the U.S. Economy', Small Business Economics 1, 65-74. Preisendrrfer,Peter, Rudolf Schiissler and Roll Ziegler (1989) 'Bestandschancen neugegrtindeter Kle-

inbetriebe', Internationales Gewerbearchiv 37, 237-248. Reynolds, Stanley S. (1988) 'Plant Closings and Exit Behaviour in Declining Industries,' Economica

55, 493-503. Scherer, F.M. and David Ross (1990) Industrial Market Structure and Economic Performance, 3rd ed.,

Boston: Houghton Mifflin. Weiss, Leonard W. (1976) 'Optimal Plant Scale and the Suboptimal Capacity', in Robert T. Masson

and P. D. Quails, (eds.), Essays on Industrial Organization in Honor of Joe S. Bain, Cambridge, Massachusetts: Ballinger, pp. 126-134.

Weiss, Leonard W. (1979) 'The Structure-Conduct-Performance Paradigm and Antitrust', University of Pennsylvania Law Review 127, 1104-1140.

Weiss, Leonard W.(1964) 'The Survival Technique and the Extent of Suboptimal Capacity', Journal of Political Economy, 72, 246-261.

Williamson, Oliver E. (1975) Markets and Hierachies, New York, NY: Macmillan Publishing Co. Winter, Sidney G. (1984) 'Schumpeterian Competition in Alternative Technological Regimes', Journal

of Economic Behavior and Organization 5, 287-320.

Documents

The rate of hazard confronting new firms and plants in U.S. manufacturing