Does access to capital affect cost stickiness? Evidence from China^*

* Accepted by Jeong-Bon Kim upon recommendation by Changling Chen

Abstract

We study the effect of limited access to capital on firm cost stickiness, using data from a large sample of Chinese private firms over 1998–2007. Our results show that on average SG&A costs are anti-sticky. For firms in regions with lower levels of financial development, SG&A costs have lower sensitivity to sales increases and exhibit lower stickiness. Overall our findings suggest access to capital as an important determinant of cost stickiness.

Keywords:

• Cost stickiness
• access to capital
• adjustment costs
• financial development

Acknowledgments

We wish to thank Rosemary Fullerton, Vic Anand, Clara Xiaoling Chen, Bart Dierynck, and participants at the 2012 AAA MAS conference and 2012 AAA annual conference for their insightful and constructive comments. Shijun Cheng gratefully acknowledges the support of the Robert H. Smith School of Business, University of Maryland. Wei Jiang is thankful for the support of the National Natural Science Foundation of China, the Humanities and Social Science Foundation of the Ministry of Education of China, the Humanities and Social Science Foundation of the Higher Education Institution of Guangdong Province, China, a major program of the Institute of Enterprise Development at Jinan University of the Key Humanities and Social Science Research Bases of Guangdong Province, and the Fundamental Research Funds for the Central Universities at Jinan University.

Notes

* Accepted by Jeong-Bon Kim upon recommendation by Changling Chen

1. We define adjustment costs as any costs incurred by firms when managers change capacity according to demand shocks, but adjustment costs and SG&A costs are different concepts. Adjustment costs may not necessarily be included in SG&A costs, but adjustment costs affect managerial decisions when managers adjust SG&A costs according to activity changes.

2. Deliberate managerial decisions focus on managers’ behavior such as: their perception of future sales activity (Banker, Byzalovy et al. 2012), incentive for earnings manipulation (Dierynck, Landsman, and Renders 2012; Kama and Weiss 2013), and empire building (Chen, Lu, and Sougiannis 2012).

3. Firms with limited access to capital also receive fewer amounts of loans and have stricter loan conditions such as restricted convents and debt maturity structure. These indirectly increase the firms’ external financing costs as well. For simplicity, we only discuss direct external financing costs in this paper.

4. Our study is similar to Banker, Byzalovy, and Chen (2013) who use the country-level employment protection legislation as an exogenous proxy for downward labor adjustment costs.

5. Dierynck, Landsman, and Renders (2012) document cost stickiness in Belgian private firms, but they eliminate small firms from their sample and do not focus on the effect of access to capital on cost stickiness.

6. Except for the above three aspects, limited access to capital may affect cost stickiness indirectly through the agency problem or managerial forecasts. For example, managers of firms with financial constraints have less free cash flows, and their incentives to retain redundant capacity are weaker, which leads to weaker cost stickiness. Similarly, managers of firms with financial constraints may be more pessimistic, which also leads to weaker cost stickiness. When we control for the agency problem or managerial forecasts in our empirical analysis, our results are robust.

7. Some firms in the database do have annual sales lower than RMB 5 million. Our main results are robust if we exclude these firms from the sample.

8. Firms in Hong Kong, Macau, and Taiwan are not included in our sample.

9. We use the sum of sales costs and administration costs in the database to approximate SG&A costs. For industrial firms in 1998, joint-stock enterprises followed the Accounting System for Joint Stock Limited Enterprises, the others followed the Accounting System for Industrial Firms. Since 2001, all firms followed the Accounting System for Business Enterprises, with the Specific Accounting Standards for Business Enterprises as a supplement. Since 2005, the Accounting System of Small Enterprises has been introduced, and small firms can choose to follow either the Accounting System of Small Enterprises or the Accounting System for Business Enterprises. Although Chinese accounting system experienced several reforms during the period between 1998 and 2007, the definitions and main items of sales costs and administration costs are consistent across our sample period and across private versus publicly traded firms, and are comparable to those of SG&A costs in the US, except minor changes of items of administration costs in 2007. Since 2007, all firms began to follow the newly introduced Accounting Standards for Business Enterprises, which substitutes for the Accounting System for Business Enterprises and the Specific Accounting Standards for Business Enterprises. According to the 2007 Accounting Standards for Business Enterprises, bad debt expense is not included in administration costs in China any more, while it is included in SG&A costs in the US. Our results are robust when we delete the observations in 2007. In addition, we also examine the stickiness of property, plant and equipment (PP&E) costs, which are less subject to accounting disclosure and tax regulations in terms of SG&A change, and find the similar results.

10. The index has been used by a number of prior studies (Wang, Wong, and Xia 2008; Firth et al. 2009; Li, Yue, and Zhao 2009).

11. Two principles are followed when these indices are established. At first, each of these five indices has at least two first level sub-indices. And each first-level sub-index can indicate one fundamental characteristic of marketization within a certain time period. Unreported techniques are applied to eliminate any non-market related noises in these indices. Second, every index is measurable and based on reliable and objective data sources. Survey data is also used if actual data is unavailable.

12. We also find the similar results when we use 1% to deal with the outliers.

13. If we use exchange rate back in period 1999–2007, the firm size in our sample is even smaller because of the even higher exchange rate between the RMB and the US dollar.

14. Although the coefficients on γ₁ is higher than that in Anderson, Banker, and Janakiraman (2003), we argue that it is hard to compare the estimates of the two studies for a few reasons: (1) cross-sample comparisons of coefficients may not be appropriate, especially the within sample variable variances in our sample and Anderson, Banker, and Janakiraman (2003)’s sample are different; (2) the definitions of the SG&A costs are not exactly the same between US and China; (3) we have public companies and SOEs, we re-estimate the coefficients after excluding these firms, and find the coefficient is lower than before; (4) legal and economic institutions are different between China and US.

15. We find the similar results when including the interaction term between RFD and Decrease_Dummy in model (2).

16. Among 1,046,294 firm-year observations in our sample, 322,181 firm-year observations are SOEs. We also separate our sample by SOEs and non-SOEs, and then test the model (2) in these two subsamples. We find that although in both SOEs and non-SOEs, λ₂ is significantly negative, the absolute value of λ₂ is significantly lower in SOEs than that in non-SOEs.

17. In our large panel data, we have firm observations over ten years and across 31 regions. It is natural to assume that error terms (ε_i,t) are correlated within each region and over the time dimension. We use the two-way clustered standard errors by region and by year based on the formula in Cameron, Gelbach, and Miller (2011). In this paper, we assume that firms in regions with different financial development levels would have different ability to get access to capital. Therefore we choose to cluster error terms by region, instead of by firm itself. And our two-way clustering method is robust to an arbitrary pattern of within-region correlations and an arbitrary pattern of serial correlation within each region.

18. As the instruments and target industries involved in the 2005 and 2006 Chinese macroeconomic regulation are different from those in the 2004 macroeconomic regulation, we only use the sample from 2003 and 2004 to examine the effect of the 2004 macroeconomic regulations.

19. We also use the firm-level cost asymmetry measure in Weiss (2010), but do not find significant results. A possible reason is that when we adopt the Weiss (2010) cost asymmetry measure, we lose 97% of our observations, and only relatively large firms remain in the sample. While Weiss (2010) uses quarterly data, we only have yearly observations. To obtain the Weiss (2010) measure using yearly observations for a firm, the firm must (1) have data over at least three and up to five consecutive years, and (2) have sales and costs changing in the same direction. In our sample, only 3% of our firm-year observations, mainly from large firms, satisfy these two conditions simultaneously.

20. We use concentration 50 ratio, which is defined as the market shares of the 50 largest firms over all the firms in the same industry, to proxy for product market competition.