The lead article in Volume 23, Issue 2 of the Review of Finance is Labor and Capital Dynamics under Financing Frictions by Ryan Michaels, Beau Page, and Toni Whited. This paper uses a new quarterly panel dataset to examine how financing frictions affect firms’ investment and hiring decisions. In particular, it shows that higher leverage reduces wages, even though it does not reduce employment. To rationalize the findings, the authors build a model where higher leverage increases the probability of default, reducing the surplus to be shared with workers and thus wages.
There is significant existing research on how financing frictions affect investment. But how do they affect employment? Prior research on this question typically uses datasets that are both (i) narrow, in terms of number of firms covered and (ii) low-frequency, i.e. annual. The authors’ primary empirical contribution is a new dataset that is (ii) broad and (ii) high-frequency, i.e. quarterly. The latter is particularly important as wages and employment change more often than annually.
The authors start with Compustat quarterly data on investment. However, since Compustat only has annual data on employment and almost no data on wages, they combine it with the Bureau of Labor Statistics’ Longitudinal Database of Establishments to obtain data on firms’ total wage bill and employment. The authors successfully replicate well-known findings in the literature, such as the positive correlation between wages and sales. They have two main new results:
are negatively correlated with leverage, both along the cross section between
firms and along the time series for a given firm.
- This result points to the importance of financing frictions, since in a Modigliani-Miller world, leverage would be irrelevant.
- Providing further evidence of the importance of financing frictions, the correlation between wages and leverage is stronger for firms without bond ratings, i.e. that are more financially constrained
- Employment has little correlation with leverage. At a quarterly frequency, the correlation is insignificant; at an annual frequency, the correlation is only weakly significant.
In addition to the new data, the paper also contributes new theory. Prior research on the link between financing and employment was predominantly empirical and focused on estimating elasticities rather than studying the underlying drivers of these elasticities. The authors build a model to rationalize their empirical findings. New theory is indeed required here, because in prior models, firms insure workers against unemployment risk. Higher leverage increases unemployment risk and so would require firms to pay higher wages as insurance – in contrast to the findings in the data, which indicate that higher leverage is associated with lower wages.
In the model, the firm chooses capital and labor (both of which have adjustment costs) as well as financing. It has three financing sources: (i) risky debt, (ii) equity, which is subject to underwriting costs that cause it to be the least-preferred financing source, and (ii) liquid assets. The model’s main novelty is that wages not set exogenously, but determined endogenously as part of a bargaining process. The bargaining process in turn is affected by the firm’s leverage, yielding a novel channel through which financing frictions affect wages, but not employment – exactly as found in the data.
The intuition is as follows. When the firm is more levered, default risk is higher, which reduces the surplus that the firm generates by hiring workers. With less surplus to bargain over, wages fall. This explains the negative relationship between leverage and wages, and also incentivizes firms to lever up to keep the wage bills in check. However, employment does not change, due to labor adjustment costs.
All three components of the model – financing frictions, investment with costly capital adjustment, and labor demand with endogenous wage bargaining – are critical to this result. Prior models with labor demand assume that workers are hired on the spot market, and so wages are not affected by financing frictions. Theories with financing frictions and labor demand exclude capital, which is a crucial factor to include as it can act as collateral.
Putting it All Together
The authors take their model to the data. The model generates a -0.17 semi-elasticity of wages with respect to leverage, very close to the -0.14 observed empirically. One of the key contributions of the model is that it allows the authors to investigate various counterfactuals. They study what happens if a random selection of firms are shocked with increased borrowing costs of the same magnitude as in Chodorow-Reich (2014). They find large effects: a 50 basis point increase in external funding costs leads to a 5% drop in employment demand. This is a striking result since there is no significant correlation between leverage and employment in the cross-section (finding (2) above). It also highlights that government policies to improve access to finance may have important real effects on employment.
The model thus provides a useful starting point for understanding the negative relationship between wages and leverage. Moreover, it offers a potential explanation for a topical puzzle – why wage growth in the US is so weak despite employment being strong. Other explanations, such as declining competition, have been posed, but the authors suggest that financial frictions may also be an important cause. The authors acknowledge that their model has greater power to explain the negative correlation between wages and leverage over the time series than along the cross section. The forces that the model identifies as being relevant over the time series – bargaining power, collateral, and uncertainty – may also be important along the cross section. In future research, it would be fruitful to investigate how much.