The dynamic interplay of capability strengths and weaknesses: investigating, страница 12

For hypotheses related to changes in strength and weakness sets, the variables of interest are munificence and prior performance. Environmental munificence refers to the availability of resources in the environment (Castrogiovanni, 1991). Following Sutcliffe’s (1994) approach, industry sales were used to create a time trend regression analysis. Munificence is the unstandardized regression coefficient in the regression model divided by the mean of industry sales. The greater this value, the higher the industry’s munificence score. Importantly, similar to our performance measure, the data used for these regressions were not obtained through the Sesame surveys, but were independently collected by the Banque de France as part of their population-level data collection effort. Thus, these figures are not influenced by sample size or by any specific firms in the sample, but reflect the entire population of firms in the focal industry. Lastly, prior performance is proxied by the

location. Specifically, we used the sample mean per industry to establish the spline. For this empirically established spline, we developed two different approaches for the final variables. For one, we simply summed the strengths and weaknesses; for the other, we followed the Deephouse approach. All three of these alternative approaches produced substantially similar results (strength and weakness sets directly and interactively affect performance). We thank a thoughtful reviewer for this suggestion.

firm’s return on assets for the year of the first survey. Again, these data were gathered independently, thus, common method variance is not a concern.

Control variables

Several control variables were included in both analyses. First, because the initial data were collected over three years, we control for the year the questionnaire was administered. Two dummy variables control for the year effect: Year 1 and Year 2. Second, we controlled for the size of the firm, because larger firms are likely to have larger revenues, costs and, perhaps, more capabilities, etc. Size is the log of the number of employees in the firm. Third, we controlled for the potential effect of membership in a trade group, because prior research suggests that group membership can affect a firm’s competitive position based on access to group resources and capabilities beyond those held by the firm (Hoskisson et al., 2004). Group is a dummy variable, with a 1 indicating membership in a business group. Fourth, to control for potential differences between the firms that are publicly traded and private firms, we included a dummy variable for this dichotomy. Public is coded with a 1 indicating public ownership. Fifth, we controlled for the age of the firm by including the logarithm of the number of years since founding. Sixth, we controlled for the firm’s level of product diversification based on a Herfindahl measure of revenues across a potential of ten separate product categories. All of these control variables were included in both sets of analyses.

Additionally, for the tests of the performance hypotheses, we included both prior performance and munificence, as discussed earlier, as controls. When testing the hypotheses related to changes in strength and weakness sets, we added three more controls. These included the number of years between surveys. This variable is a count of the years between the initial and the second survey. Also, we controlled for the initial level of strengths or weaknesses. These initial-level variables capture the dynamics of rivalry in each market.

Lastly, in both sets of analyses, we controlled for any other unobserved industry-level effects with industry dummies entered at the two-digit level. Industry clustered robust standard errors were applied to our regression modeling to provide conservative tests of our hypotheses.

RESULTS