ETF Ownership and Corporate Investment

Constantinos Antoniou is Associate Professor of Finance and Behavioural Science at University of Warwick Business School; Avanidhar Subrahmanyam is Distinguished Professor of Finance, Goldyne and Irwin Hearsh Chair in Money and Banking at University of California Los Angeles Anderson School of Management; and Onur Tosun is Assistant Professor of Finance at University of Warwick Business School. This post is based on their recent paper.

Recent work has highlighted that exchange traded funds (ETFs) contribute to a decrease in the pricing efficiency of the underlying securities (Ben-David, Franzoni and Moussawi, 2017). This is because, due to their high liquidity, ETFs attract high-frequency traders. Moreover, since ETFs and the underlying assets are bound by no arbitrage conditions, volatility in ETFs caused by high-frequency trading can propagate to the underlying assets, as arbitrageurs trade to exploit violations of the law of one price. As a result, the stock prices of companies with high ETF ownership are more noisy.

Corporate managers rely on stock prices for information, since stock prices aggregate information about the future prospects of their companies from a large pool of outside investors, and some of this information is not known to management. This reliance on the stock price leads to the well-established positive relationship between stock prices and corporate investment, which indicates that the managers of companies with higher stock prices (which are deemed by the market to have good growth opportunities) invest more heavily.

In this study we hypothesize that, since ETF ownership makes stock prices more volatile and less informative, it adversely affects the ability of managers to learn about the prospects of their firms from stock prices. Therefore, for firms highly owned by ETFs, the positive relationship between investments and stock prices should be weakened.

We test this hypothesis using three different corporate investment measures: capital expenditure, capital expenditure plus R&D, and change in assets. In our models we use Q as a measure of normalized price (calculated as the market value of equity plus the book value of total assets minus book value of equity scaled by book value of total assets), where higher Q values indicate better growth opportunities. The variable of interest for our hypothesis is the interaction between ETF ownership and Q, which we expect to be negative, so that for high ETF owned firms, real investment is less sensitive to stock prices.

The results strongly confirm our hypothesis. Using data on a sample of U.S. firms from 2000 to 2014, and models that include several firm level controls as well as firm and time fixed effects, we find that the coefficient on the interaction between ETF and Q is negative and statistically significant, for all three investment policy measures. The magnitude of the effect is considerable, as the sensitivity of investments to stock price falls by a factor of six, as we move from the 25th percentile of ETF ownership to the 75th percentile.

To allay concerns that ETF ownership is capturing the effect of an omitted variable in our models, we also test our hypothesis using an instrumental variable model, using S&P500 inclusions as the relevant instrument. Since many ETFs track the S&P500 index, the demand for shares by ETFs in newly listed companies increases. Importantly, because inclusion in the index is not based on considerations of future performance, the increase in ETF ownership is plausibly exogenous to firm fundamentals, and therefore unrelated to corporate policies. We find that the interaction between ETF (as predicted by S&P500 additions) and Q continues to be negative and significant for all three of the investment policy measures. This result strengthens the view that the relationships we document are causal.

We also estimate our models in various subsamples. We find that the negative relationship between Q and ETF is more pronounced among larger, older and less volatile firms, with better governance systems, operating in high competition industries, and managed by CEOs with longer tenure. Collectively these results suggest that the adverse effect of ETF ownership on the ability of managers to learn from prices is concentrated in cases where managers are more likely to rely on prices for information.

According to the Q theory of investments the relationship between dividend payments and prices should be negative, as firms with growth opportunities should invest and not pay out money to shareholders. However, our previous findings suggest that ETF ownership may temper this relationship. Specifically, since managers of firms highly owned by ETFs are less able to learn from stock prices about their growth opportunities, they may be more prone to making dividend payments instead. To examine whether this is the case, we estimate the relationship between a firm’s dividend policy and the interaction between ETF and Q. We find that the coefficient on the interaction is positive, consistent with this conjecture.

Our results show that ETF ownership adversely impacts the relationship between prices and corporate policies (investments and dividend payouts). Therefore, ETF ownership may be detrimental to firm performance. For our last test we estimate the relationship between operating performance (sales growth and return on assets) and the interaction between ETF and Q. The results indeed show that the coefficient on the interaction is negative, which suggests that the distortions in the relationship between corporate policies and prices associated with ETF ownership are costly.

It has long been argued that investors would be better off if they pursued passive investment strategies that are cheap to implement. The expansion in the availability of ETFs goes a long way toward achieving this objective. However, recent work, including our own analysis in this paper, suggests that ETFs entail some indirect costs to financial markets, which can affect not just market participants, but also corporate management.

The complete paper is available for download here.

Both comments and trackbacks are currently closed.