The survival of international new ventures

Abstract

International new ventures (INVs) are a popular mode of entry into foreign markets. INVs, those companies that enter foreign markets at inception, often suffer the two liabilities of newness and foreignness, which may increase the odds of their failure. This paper empirically examines the survival of INVs by comparing them with other sequential modes of international operations (e.g., acquisitions). Data from 275 British firms show that INVs have lower unconditional survival probabilities than other modes of foreign market entry. Our analyses also show that differences in survival probabilities disappear when the firms’ competitive strategies are considered.

Appendix

Once a new firm decides to internationalize its business operations, it must choose between an INV strategy and a sequential entry strategy. The strategy choice is that of a start-up that has already decided to adopt a multinational strategy and is considering whether to implement this immediately or in the future. Thus the decision being examined is a conditional one, which is the standard approach to modal choice in the FDI literature (Czinkota et al., 1996; Devereux and Griffith, 1998). This approach usually assumes that the decision to adopt an INV strategy is analogous to an established firm making a modal strategic choice decision. Firms choose an INV strategy when the expected value of a multinational start-up exceeds the expected value of a sequential approach to multinationality. The variable of interest is the difference between the expected value of launching an INV and the expected value of achieving multinationality sequentially. These two decisions differ in timing, in that the INV is multinational from its inception whereas a firm using FDI sequentially becomes multinational at a later age.

For firm i, the difference between the expected value of creating an INV and adopting a sequential approach to FDI is denoted by INV^* _i . This variable is a function of firm and industry characteristics, which can be collected together in a vector, Z _i . The actual outcome also involves an error term, e _i , attributable in part to unobservable factors (e.g., Casson, 1996; Buckley and Carter, 1998). INV^* _i itself is a latent variable, but the firm's choice of strategy is observable. This generates a binary choice variable, INV _i , (=1 where an INV is launched, and=0 where FDI follows a sequential approach).⁵ This is a standard dichotomous choice model, and for the ith firm it is specified as follows:

Firm performance is determined by firm and industry characteristics, with the binary INV _i variable providing an additive difference. The firm and industry variables that affect performance can be gathered together in a vector X _i , while INV _i functions as a dummy variable. Several of the variables, which affect the strategic choice of INV launch, would also affect performance after the choice is made. Thus many variables will enter both Z _i and X _i . Denoting survival, the measure of performance by SURV _i , the following specification is implied:

As in Eq. (A.1), the error term in Eq. (A.2), u _i , is attributable partly to unobservable factors, some of which are the same as those determining e _i . This means that u _i and e _i are correlated, and that INV _i suffers from problems of endogeneity, because INV _i is not a ‘given’ characteristic but a chosen strategy. Firms select into the two categories in INV _i based on their respective resources and capabilities. These resources and capabilities determine the choice of strategy but also influence the probability of survival (Figure A1). Treating INV _i as a normal exogenous variable leads to selectivity bias (Heckman, 1979), which appears in both the mean and the variance of the estimator of θ in Eq. (A.2). The estimate of θ is biased in the direction of the correlation between the errors u _i and e _i . Also, the estimated standard error of θ is biased downward. Consequently, the probability that it will appear significant is increased.⁶

Figure A1

Simplified choice set for the entrepreneurial firm.

Testing Hypothesis 1 involves estimating Eqs. (A.1) and (A.2). The first step involves estimating Eq. (A.1), and provides a preliminary test of the variables that affect the choice of an INV strategy. Next, direct estimates of Eq. (A.2) are generated: i.e., treating INV _i , the INV dummy, as exogenous. Testing Hypothesis 1 involves comparing these direct estimates with estimates of Eq. (A.2) after accounting for selectivity bias (in effect estimating Eqs. (A.1) and (A.2) simultaneously). Several estimating methodologies for addressing selectivity bias are available, and two separate approaches are used here. Eq. (A.2) is estimated separately for INVs and sequential FDI firms. This is done in a two-step manner, adding the estimated self-selection parameter from the estimation of Eq. (A.1) as a regressor in separate estimations of Eq. (A.2) for INVs and sequential FDI firms. While these estimates are not efficient, they are consistent.⁷ Finally, Hypotheses 2 and 3 are tested using the estimates of Eq. (A.2). Given that the correction of selectivity bias is argued to be the correct specification, the tests based on these estimates are preferred.