International Involvement and Production Efficiency among Startup Firms

Abstract

The economic theory of small firms often requires a reasoning process distinct from those typically used for large multinational enterprises (MNEs), since small firms typically do not possess the resources MNEs commonly employ to outperform similar domestic firms. In the current analysis, we argue that new small firms with international sales can better predict their revenue stream than can comparably aged firms with only domestic sales, and, as a result, international selling firms have higher levels of technical productive efficiency. We employ 7829 firm-year observations from the 2007 to 2011 years of the Kauffman Firm Survey microdata sample in our analysis. A stochastic frontier model empirically supports the result that technical productive efficiency is positively related to the foreign sales ratio. These results hold after controlling for multiple relevant owner and firm characteristics, as well as accounting for potential endogeneity concerns.

Key Words:

• Production efficiency
• stochastic frontier model
• international involvement
• startups
• Kauffman Survey

Jel Classification:

• F23
• F29
• L26

Acknowledgements

The authors acknowledge the support of the Ewing Marion Kauffman Foundation through access to the KFS data. The authors made use of the confidential KFS data, which were securely accessed through the National Opinion Research Center (NORC) Data Enclave at the University of Chicago. All results have been reviewed to make sure that no confidential data have been exposed. Any opinions, findings, and conclusions or recommendations expressed in this material are those of the authors and do not necessarily reflect the views of the Kauffman Foundation. The authors thank Johannes Fernandes-Huessy and Daniel Lee for assistance at the NORC Data Enclave. The authors thank Jon D. Haveman as well as the editor and anonymous reviewers for many helpful comments, criticisms, and suggestions. All mistakes are the authors’ own.

Notes

1. For a review of the literature on the multinationality–performance relationship, see, among others, Yang and Driffield (2012).

2. International diversification as a justification for a firm's presence internationally is not a “consistent” finding. For example, in their sample of firms, Baek and Kwok (2002) show that the international diversification hypothesis works for the foreign but not the US firms.

3. This applies in situations where the foreign exchange rate is stable.

4. Inventory includes materials, works in progress, and final products.

5. Miller and Pras (1980) examined the effects of product diversification, multinational diversification, and exports on the variation in earnings. Using a sample of 246 US corporations in 1961, 1965, and 1968, they found that product and export diversification measures do not significantly relate to return on assets (ROA) variance. Fatemi (1988) examined the financial structures of 136 firms during the 1978–1982 period and found that the foreign sales ratio negatively related to interest-bearing debt, and positively related to both short-term debt and current assets.

6. ROE and ROA are not good measures of performance for younger firms because net income is likely to be variable and because there are likely to be sudden and frequent changes in debt and equity during the early years of a firm's existence. As mentioned earlier, efficiency is also able to capture a measure of shareholder wealth, which is entirely separate from measures of profitability.

7. Production itself was defined as firm revenues plus the change in inventories from the previous year.

8. The Kauffman data include information on six-digit North American Industrial Classification System (NAICS) codes. In order to better conform with the literature, a crosswalk between a set of SIC industry codes and the corresponding NAICS categories was performed for the sake of the analysis. The SIC industry breakdown is similar to that used in Baek (2004).

9. The variables we used to measure international presence, namely: (a) whether there are any international sales, (b) the percentage of international sales, and (c) the location of the primary customer, were largely a function of data availability. Additional follow-up work employing other datasets could use additional measures of international involvement as well as a more direct measure of the foreign sales ratio.

10. All regressions were run in STATA using the “difficult” option to allow for convergence.

11. While the parameters were all estimated jointly, the model can be considered in a two-stage fashion, with the first stage as the production equation (1) and the second stage modeling the error terms u and v. For ease of explanation, the two-stage structure is retained for descriptive purposes, although the simultaneous estimation was the one actually employed. For more on frontier functions, see Kumbhakar and Lovell (2000).

12. A limited number of industry and industry-year results were disclosed due to confidentiality concerns.

13. As an aside, while we take a sample of already privatized firms, notice that it may also be interesting to examine how the fact of privatization itself affects firm performance and efficiency (Arcas & Bachiller, 2008).

14. Race was split into the categories of Asian, Black, White, Hispanic, Native Hawaiian, and Other. Immigrant status employed both whether the owner was foreign born, and whether they were naturalized citizens without additional information with regard to country of origin or immigrant year of entry. Fixed assets were defined as equipment, land or buildings, vehicles, other business property, and other assets. Total assets were defined as fixed assets plus accounts receivables, cash, and inventories. All portions of assets were set to a default value of one because the natural log regression format was being employed. Only the total number of workers was used as a measurement, rather than breaking the analysis into both full-time and part-time workers, although this breakdown was attempted in early parts of the analysis. It was subsequently dropped due to low response levels to the total number of workers stratified by full-time or part-time status.

15. A variable appendix is included, providing more detail on how the variables were constructed.

16. Demographic characteristics were all gathered at the beginning of the sample in year 0 which is why, despite the passage of time, same-industry work experience does not change. It is for this reason that the demographic characteristics should generally remain stable if we do not have selective attrition over time conditional on demographic characteristics in the sample.

17. Results are shown in levels; however, natural logs are used in the later regression structure.

18. This value is very close to Census 2000 measures of the fraction of foreign-born individuals residing in the USA (11%). This number is based on author calculations using Summary File 3 data and is available upon request.

19. Industry 0 is used as the base case and is not displayed.

20. Entertainment, health, education, and law were the relevant categories.

21. Due to the coding structure which we employed, many firms in this portion of the analysis were deleted due to nonresponse or insufficient response regarding revenue or international sales information. Of the remaining firms, 109 had exports in all years of the sample period; over 600 firms had no foreign sales in any year of the sample period; and the remaining firms switched from having international sales to either having no international sales or else not answering the export questions, and vice versa. It was this last category of switching firms which had the largest average SSE.

22. Having an American Indian owner may have an effect, but there were not enough American Indian owners in the sample for us to conclusively determine if this was true. Our results are mostly similar to those from Iqbal et al. (1999) who showed that minority-owned banks – particularly those with African-American owners – are less technically efficient. We find, however, the contrasting result, perhaps because of our particular sample in terms of firm type and size, as well as our sampling time-frame, that Asian-owned firms are not less efficient.

23. Notice that some industry results were suppressed due to confidentiality concerns because the number of observations in the data was less than 30 – which is the cutoff for disclosure requests in the Kauffman data. The industries that were not displayed were #2, 4, 11, 15, and 17.

24. One explanation may include government subsidies and the entrenched methods of production which are employed regardless of the results, or R&D costs.

25. The number of observations in Table 5 decreased due to the decreased number of years of data. Additionally, notice that the deletion of one year of data in a five-year sample may account for a slight change in the size of coefficients.

26. Other work, see in particular Lim and Kim (2011), has examined more generally the question of whether owner nativity affects export levels and how this is mediated by owner education.