Is the asymmetric cost behavior affected by competition factors?^*

* Accepted by Hong Hwang upon recommendation by Jeong-Bon Kim

Abstract

Prior studies on the asymmetric cost behavior mainly focus on internal factors. However, managers consider both internal and external factors when making strategic cost decisions. In this paper, we investigate the relation between the asymmetric cost behavior and external competition factors. We find that Selling, General, and Administrative costs are more sticky for firms in different competition environments, proxied by higher product differentiation, higher entry costs, and larger market size. Overall, these findings suggest that the asymmetric cost behavior is affected by external factors as well as internal factors.

Keywords:

• Cost stickiness
• competition factors
• product differentiation
• entry costs
• market size

Notes

* Accepted by Hong Hwang upon recommendation by Jeong-Bon Kim

1. Many subsequent studies after Anderson, Banker, and Janakiraman (2003) use GNP or Gross Domestic Product (GDP) to measure macroeconomic growth (Banker, Byzalov, and Chen 2013).

2. Since 1988, data from most countries are available in Compustat Global (Banker, Byzalov, and Chen 2013). Because some control variables are measured with two lags in the data, the final sample starts from 1990.

3. ENTCOST1 measured by the natural logarithm of the ratio of total assets to sales at the industry level may simply reflect asset rigidity. To obtain a more comprehensive set of measures for entry costs, we use two other proxies, R&D intensity (ENTCOST2) and capital intensity (ENTCOST3) (Cheng 2005).

4. We find similar results when DIFF are measured as the ratio of sales to operating costs.

5. Following Roychowdhury (2006), R&D expenditures are set to zero if they are missing.

6. Endogeneity issues could potentially affect the results of this study. To mitigate this concern, we use a lead-lag model and find similar results.

7. If b₂ is insignificant, then percentage decrease in SG&A costs per 1% decrease in sales is calculated as the sum of b₁ and b₃.

8. For brevity, we report two additional tests. Untabulated results are unchanged (1) three-digit SIC codes (Cheng 2005) or four-digit SIC codes (Karuna 2007) to measure the level of competition within an industry, (2) at least five observations in each year- country-industry group, (3) the measurement of competition factors across countries. Also, Consistent with the results from previous studies, we find significant variations in the results across industries. The effect of competition factors on cost stickiness is mainly concentrated in three large countries (Japan, UK, and US) in our sample, although it still exists in other countries. Finally, when we transform the three factors of product market competition into percentile rank variables to compare the coefficients across different factors, we find that the coefficient on ΔSALE_i,t*DD_i,t*ENTCOST_j,t is the largest, followed by the coefficient on ΔSALE_i,t*DD_i,t*MKSIZE_j,t. The coefficient on ΔSALE_i,t*DD_i,t*DIFF_j,t is the smallest.

9. We use the following measures: (1) country-level legal origin, measured as code-law or common-law country as defined in Table 2 of La Porta et al. (1998); (2) law enforcement, measured as the mean score of three legal variables, the efficiency of the judicial system, an assessment of rule of law, and the corruption index as defined in Table 2 of La Porta et al. (1998), and Panel B of Table 2 in Leuz, Nanda, and Wysocki (2003); (3) shareholder rights, measured as the anti-director rights index (aggregate measure of minority shareholder rights) as described in Table 9 of Djankov et al. (2008); (4) competition environments, measured as a survey index from 1(strongly disagree) to 7 (strongly agree) using survey questions (Competition in the local market is intense and market shares fluctuate constantly) from World Economic Forum (www.weforum.org/gcr) (2015) (Djankov et al. (2002); Dyck and Zingales (2004).

10. We report the mean values of each coefficient and t-statistic from 30 country-level regressions at the bottom of the Table 8. The t-statistics are based on the distribution of the mean values of the coefficients across 30 countries. These are similar to Fama-MacBeth t-statistics.