How does leverage affect the value relevance? Evidence from Turkey

ABSTRACT

High-levered firms have serious concerns related to avoiding covenant violations and meeting the needs of their creditors. Accounting information of those firms should be less value relevant for market participants. Based on a sample of Turkish listed firms over 2009–2018, we analyse whether the value relevance of accounting information is significantly lower for high-levered firms. For this purpose, we group observations with no net debt and divide the rest into quintiles based on leverage levels. We conclude that the value relevance of both earnings and book value of equity is lower for the high-levered quintile than the rest. Moreover, the value relevance of earnings is moderated more than the value relevance of book value of equity for the high-levered quintile. Last, book value of equity is more dominant in the valuation of the high-levered quintile than the valuation of the rest.

KEYWORDS:

• Value relevance
• leverage
• capital structure
• accounting information
• Turkey

JEL CLASSIFICATION:

• G12
• G14
• M41

Acknowledgments

We are very grateful for precious comments and feedbacks of Flora Muiño, Beatriz García Osma, two unanimous referees, Ali Coşkun, Fatih Kiraz and Efe Can Gürcan.

Disclosure statement

No potential conflict of interest was reported by the author.

Notes

1. https://www.ifrs.org/use-around-the-world/use-of-ifrs-standards-by-jurisdiction/#analysis (Accession Date: 6 April 2020)

2. In Turkey, IFRS has been the mandatory financial reporting regime for listed firms as of 2005.

3. Their earnings figure is Earnings Before Interest and Taxes.

4. It is statistically slightly different version of the method of performing separate analyses for different sub-samples.

5. Debt-covenant-related concerns eventually include default which is extremely costly.

6. DeFond and Jiambalvo (1994) report that violation firms experienced management changes and/or going concern-related problems in the year of violation. These two reasons significantly affect outcomes reported for abnormal total accruals in that year. After removing the impact of these two reasons on regression outcomes, total abnormal accruals are not reported as statistically significant while abnormal working capitals remain significantly positive in the year of violation.

7. These outcomes belong to financial health measured by senior debt ratings.

8. Since all dependent and independent variables are deflated by lagged MVE, the market value data cover one more year than the financial statement data.

9. We decide to use total debt instead of total liabilities because most liability items are not interest-bearing and are not expected to result in financial distress by their natures.

10. Dividing the sample into quintiles based on certain levels is a prevalent approach in the literature. For instance, Aleksanyan (2009) and Aleksanyan and Karim (2013) divide their sample into size-based quintiles while Ciftci and Darrough (2015) divide their sample into quintiles based on R&D intensity levels and Venter et al. (2014) divide their sample into quintiles based on a special type of earnings. Indeed, Ciftci and Darrough (2015) divide their sample into five plus one, one includes firms with no R&D intensity. We prefer following this five plus one approach when we divide our sample into quintiles.

11. In this setting, we first plot the distribution by putting MVE on the y-axis and BVE on the x-axis. We realise that the last quintile is distinctively located below the others. Then, we replace the x-axis by earnings and realise that the last quintile is distinctively located below the others except for the zeroth and third quintiles which are located below the last quintile especially for high-level of loss observations. Those distribution graphs are available from the authors upon request.

12. As this discussion is far beyond the scope of our study, we refer the reader to Gujarati (1970a, 1970b).

13. Onali et al. (2017) and Ertuğrul and Demir (2018) provide concrete evidence for using the fixed-effects method.

14. Our unreported statistics indicate that almost 30% of the whole sample records losses.

15. Those outcomes are available from the authors upon request.

16. Since our period of analysis begins as of 2009, this deflation requires total asset figures belonging to 2008. We lose observations majorly belonging to 2009 as we do not have the required data of 2008.

17. In that research setting, although the mean VIF figure of each regression is not problematic, the maximum VIF figures are reported as very close to the critical value of 10 in certain regressions. Hence, we drop quintile dummies in those regressions and re-perform analyses. Our outcomes remain unchanged.