Debt on the Balance Sheet Emerges as a Powerful Predictor of Stock Returns in India

A firm’s choice of how much debt to carry on its balance sheet may do far more than shape its cost of capital. New research from India suggests that capital structure, the mix of debt, liabilities and equity a company uses to finance itself, behaves like a genuine pricing factor in the stock market, capable of explaining return patterns that the celebrated Fama-French three factor model leaves unexplained. The study, published in Frontiers in Applied Mathematics and Statistics, analyzed 394 non-financial companies listed on India’s National Stock Exchange between January 2010 and December 2024 and found that portfolios built around balance sheet leverage earn persistently different returns from those built around size or valuation alone.

The research, conducted by Jayashree Somu and Nirakar Barik of Vellore Institute of Technology, builds on one of the most influential frameworks in modern finance. In the early 1990s, Eugene Fama and Kenneth French showed that three factors, the excess return of the market, the size of the firm measured by market capitalization, and the book-to-market ratio that separates value stocks from growth stocks, could explain cross-sectional variation in stock returns far better than the older capital asset pricing model. That three factor model became a cornerstone of empirical asset pricing and has since been validated across dozens of developed and emerging markets, from the six major equity markets studied by Capaul and colleagues in the early 1990s to the thirteen developed and emerging markets Fama and French themselves examined in 1998.

India, however, presents a distinctive testing ground. Earlier studies by Connor and Sehgal, and later by Kumar and Sehgal, confirmed that firm size and book-to-market characteristics influence Indian stock returns, with Sehgal and Balakrishnan demonstrating that the Fama-French model outperforms the traditional CAPM in the Indian context. Yet most Indian research has measured leverage through the narrow debt-to-equity ratio, which captures only formal borrowings and ignores operating liabilities such as trade credit and accounts payable. Seminal work by Rajan and Zingales showed that capital structure conclusions are highly sensitive to how leverage is defined, and Nissim and Penman made the theoretical distinction explicit: a firm’s balance sheet contains both financial liabilities and operating liabilities, and the two carry different risk implications and are priced differently by markets. Welch has gone further, arguing that the financial debt-to-asset ratio underlying debt-equity measures is fundamentally flawed because it implicitly treats non-financial liabilities as equity.

To address this measurement gap, the researchers turned to two broader, asset-scaled capital structure ratios: long-term debt to total assets, abbreviated LD/TA, and total liabilities to total assets, abbreviated TL/TA. These measures had previously been employed by Li and colleagues in the United States market, where they revealed differential return patterns across portfolios sorted on the two ratios, and by Duong Le in Australia. The Indian study is, according to its authors, the first in the Indian context to incorporate these asset-scaled capital structure measures into an asset pricing framework both as portfolio sorting variables and as constructed risk factors, distinguishing the pricing of capital structure from the financial leverage documented in earlier Indian work by Maiti and Balakrishnan.

The methodology followed the now-classic Fama-French recipe with careful attention to timing. At the end of June each year, companies were ranked on market capitalization, price-to-book ratio, and the two capital structure ratios, with the accounting measures drawn from financial statements for the fiscal year ending the preceding March. This deliberate lag between fiscal year-end and portfolio formation prevents look-ahead bias, ensuring that the accounting information used to construct portfolios was publicly known before returns were measured. Single-sorted quintile portfolios were built for each characteristic, and double sorting produced 25 size-value, 25 size-LD/TA and 25 size-TL/TA portfolios. From the intersections of two size groups and three subgroups on each characteristic, the researchers constructed mimicking portfolios: the familiar SMB, or small minus big, the value factor LMH, or low minus high price-to-book, and two new capital structure factors, LMXLDTA and LMXTLTA.

The single-sorted results revealed striking patterns. Portfolios of micro-cap stocks earned mean excess returns of 3.3 percent per month against just 0.4 percent for mega-cap portfolios, and value stocks returned 2.0 percent against 1.0 percent for growth stocks, confirming a strong size effect and a value effect in the Indian market. The capital structure results ran in the opposite direction to intuition at first glance: portfolios of stocks with high capital structure outperformed those with low capital structure, with mean returns of roughly 1.5 percent for the highest leverage quintile against 1.2 percent for the lowest. This premium compensates investors for bearing the additional financial risk that heavily indebted firms carry, a mechanism consistent with capital structure theories from Modigliani and Miller onward, including the trade-off theory of Kraus and Litzenberger, the pecking order theory of Myers and Majluf, and the market timing theory of Baker and Wurgler.

The correlation analysis added a crucial nuance. Both capital structure factors were negatively correlated with the value factor, at minus 0.646 for LD/TA and minus 0.671 for TL/TA, suggesting that firms with higher capital structure behave like value firms, which is theoretically expected since financial leverage has been argued to be one of the sources of the value premium. Yet the two capital structure factors were positively correlated with each other at 0.677, well below unity, indicating that each carries distinct information and justifying the inclusion of both in the modified model. The corresponding variance inflation factor of approximately 1.84 sits comfortably below the conventional collinearity threshold of five.

The heart of the study lies in the comparison of three regression models. The traditional Fama-French three factor model, applied to the size-value portfolios, performed reasonably well: 23 of 25 intercepts were statistically indistinguishable from zero, market betas clustered near one, and R-squared values ranged from 0.547 to 0.865. But when portfolios were sorted on size and total liabilities to total assets, the standard model began to falter. Four of 25 intercepts became significantly different from zero, with the unexplained returns concentrated among small firms, indicating that the broad liability-based capital structure measure carries information beyond size and value. Notably, the size-LD/TA sorts were handled adequately by the standard model, showing that long-term debt sorting does not generate return patterns beyond what market beta, size and value already capture.

Enter the Modified Fama-French three factor model, in which the value factor is replaced by the two capital structure factors. The motivation is that leverage has been identified as a potential source of the value premium, meaning the two may capture overlapping dimensions of risk. The results were decisive. For size-value portfolios, the modified model explained the one portfolio the standard model had left unexplained, and average R-squared improved from 75.6 percent to 76.6 percent. For size-LD/TA portfolios, 24 of 25 portfolios were explained, average absolute alpha fell from 0.003 to 0.002, and R-squared rose from 74.3 percent to 76.3 percent. For size-TL/TA portfolios, the improvement was most dramatic: significant alphas dropped from four to one, average absolute alpha declined from 0.004 to 0.003, and average R-squared reached 77.2 percent, the highest among all models tested. The TL/TA factor loadings declined systematically across leverage quintiles within every size row and were statistically significant for a majority of portfolios.

To guard against the possibility that the capital structure factors merely substitute for value rather than adding new information, the researchers also estimated an Extended Fama-French three factor model retaining both the value factor and the capital structure factors. Although the alphas in this extended model appeared indistinguishable from zero, the number of statistically significant intercepts actually increased relative to the modified model across all three portfolio sorts, and average R-squared declined. The Gibbons, Ross and Shanken test, which examines whether all 25 intercepts are jointly zero, corroborated the picture: the modified model produced lower GRS statistics and higher R-squared values than both the traditional and extended models in most combinations. The conclusion is that capital structure is not merely a proxy for value but a conspicuous, independent dimension of risk in the Indian stock market. For investors, the practical message is twofold: small-cap and value-oriented firms offer premiums, and firms with high total liabilities to total assets earn higher returns as compensation for financial risk, though the choice between leverage measures materially affects risk assessment. For fund managers and regulators alike, the findings suggest that risk models relying exclusively on book-to-market may underestimate return variation related to capital structure, and that transparent disclosure of complete liability structures, including operating liabilities, would better protect investors in emerging markets.

Subject of Research: The role of capital structure measures as risk factors in explaining portfolio returns in the Indian stock market

Article Title: Capital structure risk and portfolio returns: evidence from Indian stock market

Article References: Somu, J., & Barik, N. (2026). Capital structure risk and portfolio returns: evidence from Indian stock market. Frontiers in Applied Mathematics and Statistics, 12, Article 1953346. https://doi.org/10.3389/fams.2026.1953346

Image Credits: AI Generated

DOI: 10.3389/fams.2026.1953346

Keywords: capital structure, Fama-French three factor model, Indian stock market, asset pricing, leverage premium, size effect, value effect, portfolio returns, total liabilities to total assets, emerging markets, risk factors, NSE 500