A Schumpeterian approach to entry barrier and firm profitability: cycle time of technology

ABSTRACT

Entry barrier has long been considered as a major determinant of firm profitability. Although a less competitive market structure has been commonly known as an indicator of an entry barrier, pieces of past empirical evidence are mixed. Moreover, technological factors, such as R&D intensity, have also been considered. However, no satisfactory empirical analysis has been made, mostly due to the lack of a suitable proxy variable that can reflect the technological environment of a sector. This study addresses this problem by trying a new proxy variable, cycle time of technologies (CTT), and shows, using the US firm data, that firms in a sector with a long CTT tend to enjoy higher profitability and values than others. A long CTT of a sector presents a high entry barrier against any entrant because in such sectors, an existing stock of knowledge tends to be important for a longer period of time, making new innovation continuously rely on old knowledge owned by incumbents and protected by patent rights.

KEYWORDS:

• Entry barrier
• profitability
• cycle time of technologies
• Tobin’s Q
• market structure
• CR4

JEL CODES:

• D21
• D43
• L11
• L22
• O32

Acknowledgement

Earlier versions of this paper were presented at several occasions, including the 2018 International Schumpeter Society Conference. The first author acknowledges the funding by the SNU Foundation of Seoul National University. This revision has benefitted from the three rounds of useful comments by two anonymous referees.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

1 Orr (1974b) used the R&D intensity defined at the sector level, whereas Salinger (1984) used the ratio of the R&D expenditure to a stock of property, plant, and equipment.

2 Such heterogeneity was also mentioned in Scherer (1983) or Cohen, Nelson, and Walsh (2000) where, notably, each industry puts differing levels of emphasis on patents.

3 Park and Lee (2006) and Lee (2013: Ch. 3) conducted a sector-level patent analysis and argued that emerging countries tend to file more patents in short CTT sectors and thus specialize in such sectors because these sectors can be their niches, compared with countries with advanced economies.

4 One referee observed that the effect of CTT of a sector can be examined in terms of its effects on sectoral variables, such as market structure or the entry of new firms at a sector level. However, given that there is a stock of the classical and important literature on the effect of market structure/entry barrier on the firms’ profitability, we aimed to position this paper in such literature, comparing CTT with other measures of entry barrier or market structure. Furthermore, the ultimate effect of entry barrier should be judged in terms of its effect on firms, rather than the sector.

5 This study tends to explain the advantage of technological leaders through imitation costs, which according to Mansfield, Schwartz, and Wagner (1981) includes licensing costs, costs of inventing around the existing patents, and so on. In the context of CTT, imitation costs can be taken to include the licensing costs and the investment needed to absorb previous technology.

6 For some new firms with no existing ID, we have assigned a new pdpass. After the merge, firms that were found to have a market share of 1 (100%) were removed. Of course, this action was conducted before the removal of outliers.

7 Negative value for sales makes seeing what market share signifies difficult.

8 An alternative method is first to calculate the average CTT of a technological class (not sectors) and then match each class to each sector. However, this kind of matching is arbitrary given the fundamental difference between the technological classes and industrial sectors. Kortum and Putnam (1997) attempted to link International Patent Classification with North American Industry Classification System. The results are dependent on the specific time period at which the linking took place, which makes it unreliable, especially when different or longer time periods are considered.

9 ROE is more widely used as a standard, whereas the return on assets (ROA) is also used in consideration of variations in firms’ debt-equity ratio, and higher debt ratio means bigger difference in equity vs. asset. ROA can be calculated as the ratio of net income to total assets and was used in some studies focusing on technology, such as Artz et al. (2010). The results using ROA are available upon request.

10 The variable of firm ages in the dataset is not the actual age, but the time period for which the firm was included in the dataset.

11 This criterion is used because negative equity makes ROE extremely difficult, if not impossible, to interpret. For example, when equity is negative, firms with negative net income would actually be considered to be better performing than those with positive net income.

12 Such firms are dropped because some firms have implausibly high ROEs, generally from a suddenly small value of equity due to sudden changes in how assets and liabilities are calculated, which seem to have had little to do with actual performance. For example, Lockheed Martin had a sudden rise in ROE not because it was suddenly more profitable but because the figure for its equity was affected by changes in how pension plans are accounted for.

13 As shown in Table 1, the statistics for the variables are based on the dataset used to run regressions with the ROE as the dependent variable. The exceptions are Tobin’s Q and the current sector CTT, which are based on the dataset used for regressions that have Tobin’s Q as the dependent variable. These exceptions also apply to Tables 2A, 2B, and other tables.

14 This is primarily based on the Hausman test and the Breusch-Pagan test performed for Model G in Table 4A, the full model with ROE as the dependent variable. For the Hausman test, the chi square statistic was 48.37 with degree of freedom 14, suggesting that the fixed effects model is preferred over the random effects model. For the Breusch-Pagan test, the chi square statistic was 694.8 with degree of freedom 1, suggesting that the random effects model is preferred over the pooled effect model. To make comparisons across the different specifications easier, we decided to use the fixed effects model for all other models as well.

15 When we include all sectors with 5 firms, only one sector is added. The result with all explanatory variables remains the same and consistent (shown to the referees and available upon request. Meanwhile, there is a tradeoff between adding more firms (e.g. to firms with at least one patent every 5 or 6 years) and reliability of such constructed measure of CTT (because we include firms with few patents). We have tried to extend the sample to include the firms with at least one patent every 6 years. Sample size is increased to include 420 firms, but 352 firms have no patents in 2013. Then, a median value of the number of patents in such sample becomes zero in 2013, which is a huge problem in regressions. In regressions with this sample, we cannot obtain robust results for ROE, although Tobin’s Q results are fine. However, we extended the sample to include the firms with at least one patent in every 4 years, we are able to obtain the same results as the original one (shown to the referees and available upon request). We tried with the sample for every 5 years, the results are in-between that with every 4 and 6 years. We interpreted that the estimated values of CTT became more unreliable if we include the firms filing less or zero patents. Given this fundamental trade-off, the results with every 3 years seems to be a reasonable choice.