Böck M, Zörner TO (2023)
Publication Language: English
Publication Type: Journal article
Publication year: 2023
URI: https://onlinelibrary.wiley.com/doi/epdf/10.1111/jmcb.13109
DOI: 10.1111/jmcb.13109
This paper investigates the role of credit market sentiment and investor beliefs in credit cycle dynamics and their transmission to businesscycle fluctuations. Using U.S. data from 1968 to 2014, we find that credit market sentiment is indeed able to detect asymmetries in a small-scale macroeconomic model. An unexpected credit market sentiment shock has different impacts in an optimistic and pessimistic credit market environment. While an unexpected movement in the optimistic regime leads to a rather muted impact on output and credit, we find a significant negative impact on these variables in the pessimistic regime. The findings highlight the relevance of expectation formation mechanisms as a source of macroeconomic instability.
The Great Financial Crisis (GFC) of 2008–09 revived interest among economists and policymakers in the role of credit expansion, investor beliefs, and expectations in financial markets as a source of recurrent financial crises. In this paper, we propose an empirical macroeconomic model that links credit market dynamics with shocks to investors sentiment. The general idea is based on a promising strand of literature that finds a strong link between credit market sentiment in explaining different phases of economic activity. For instance, López-Salido, Stein, and Zakrajšek (2017) show that low credit spreads used as a proxy for market sentiment predict both a rise in credit spreads and a decline in economic activity. The importance of investor sentiment raises the question of which measures or events might be influential in providing explanations for macroeconomic instability.
Hence, a growing body of literature, originating from Beaudry and Portier (2004, 2014), argues that sudden changes in expectations due to unexpected news are an independent and significant driver of macroeconomic fluctuations. The main idea of news-driven business cycles is that news about the future total factor productivity (TFP) exerts strong linkages to other macroeconomic fundamentals and is thus important in explaining fluctuations. Recently, Görtz, Tsoukalas, and Zanetti (2022) present an analysis of how such TFP news shocks propagate in an environment with financial frictions. They report strong linkages between innovations due to news shocks as well as shocks explaining movements in risk indicators such as credit spreads. The role of news thus seems to be promising for explaining the origins of macroeconomic instability. However, these approaches focus on news about technological innovations and how these shocks shape agents' expectations concerning the stability properties of an economy. Instead, market sentiment in general tends to react rather quickly to unexpected news which are not necessarily based on fundamentals. As pointed out by Stein (2014), the relevance of risk measures and how agents form their expectations about future risks to monetary policy became apparent a few years after the crisis.
The rather strong assumption of rationality in the formation of expectations is criticized by many scholars originating from Simon (1957). In response, several branches of the literature relax this assumption and address the concept of bounded rationality. One strand of literature incorporates learning into the expectation formation process (Sargent 1993, Evans and Honkapohja 1999), another one deals with rational inattention (Mackowiak and Wiederholt 2009, Sims 2003, Gabaix 2014), and finally the behavioral literature uses simple psychological heuristics (Tversky and Kahneman 1974). Assenza et al. (2011) find several robust heuristics when agents forecast inflation dynamics in a learning-to-forecast experimental study. A more recent contribution by Bordalo, Gennaioli, and Shleifer (2018) operationalizes the representativeness heuristic (Kahneman and Tversky 1972) and presents a new belief formation mechanism—diagnostic expectations—to explain credit cycles.
Prominent models explaining credit cycle instability rely on some form of financial frictions. These models typically endow their agents with rational expectations (RE; Bernanke and Gertler 1989, Kiyotaki and Moore 1997), who cut back on investment and reduce borrowing when their ability to borrow is constrained. Since agents are not fully aware of what their borrowing decisions trigger in the aggregate economy, they face externalities in their choice of leverage. The fragility of the system is then identified through leverage. This has initiated a plethora of empirical studies using balance sheet measures as predictors of recessions (Schularick and Taylor 2012, Jordà, Schularick, and Taylor 2016, Mian, Sufi, and Verner 2017). In summary, according to this strand of the literature, the driving force behind fluctuations can be traced back to some sort of market imperfection.
On the contrary, behavioral theories highlight the importance of overoptimism in the wake of a credit boom, starting with the seminal contributions of Minsky (1977) and Kindleberger (1978). Similar findings emerge from the behavioral finance literature, which emphasizes the time-varying component of credit conditions when the assumption of belief updating via RE is relaxed. The literature introduces behavioral elements into the expectation formation mechanism of agents. For instance, Greenwood and Hanson (2013) show that credit quality of corporate debt issuers deteriorates when the credit market overheats and is thus a better predictor for recessions than rapid aggregate credit growth. Baron and Xiong (2017) highlight the undervaluation of crash risk in the face of credit expansions, while López-Salido, Stein, and Zakrajšek (2017) provide evidence that sentiment on the credit market is a valid predictor of recessions. Using these concepts, Boeck (2023) examines belief formation on financial markets in detail by investigating the responses of belief distortions to a financial risk shock.
While the financial frictions approach mostly assumes exogenous shocks and RE in order to explain fluctuations, other models provide endogenous explanations without relying on the assumption of strict rationality. For example, Matsuyama, Sushko, and Gardini (2016) propose a model of endogenous credit cycles in an overlapping generations setup without leaving the realm of RE. They introduce financial frictions through the limited pledgability of collateral (Tirole 2010). A recent contribution by Kubin et al. (2019) extends this model by endogenizing the pledgability parameter and allowing it to vary over time according to a simple heuristic rule: if the current state of the economy, proxied by net worth, is above or below some threshold, the agents' sentiment switches between an optimistic and a pessimistic credit market regime, reflecting the psychological state of the lenders. This translates into different degrees of what lenders are willing to accept as collateral and thus provides an endogenous behavioral explanation of how the general (or in particular the business) sentiment is anchored in perceptions of agency problems.
Taking this theoretical approach as a starting point, we propose an empirical macroeconomic model that incorporates both an endogenous explanation of credit cycles as well as switching dynamics resulting from periods of optimism and pessimism on the credit market. To this end, we test the conjecture that credit market sentiment can exert disruptive forces on the credit and business cycle. We distinguish between optimistic and pessimistic credit market conditions and expect more severe effects when the financial system is already in distress. Furthermore, we argue that when agents receive bad news about the state of the economy, sentiment toward the credit market deteriorates and credit spreads rise. This is then transmitted to real economy and we expect a decline in credit, investment, and eventually output. To investigate the proposed relationship, we employ a nonlinear structural vector autoregression (SVAR) with monthly data covering the period between January 1968 and December 2014 of the U.S. economy. This setup allows us to examine the impact of a credit market sentiment shock, which is equivalent to an unexpected news shock in the credit market. The identification is based on the different belief formation mechanisms that agents use to forecast risks in financial markets.
Our paper is also related to the literature on structural identification in threshold VARs (TVAR). While the SVAR literature has made great advancements in identifying monetary policy shocks (Gertler and Karadi 2015), the literature looking at financial market feedbacks or modeling financial shocks in an explicit manner is rather scarce. Studies have mostly focused on single-equation models (Krishnamurthy and Muir 2017, López-Salido, Stein, and Zakrajšek 2017) and reduced-form multi-equation models (Gilchrist and Zakrajšek 2012). A recent paper by Caldara and Herbst (2019) adds credit spread variables to a structurally identified multivariate framework but they restrict their analysis to monetary policy shocks. To the best of our knowledge, there are only two other papers that deal with the identification of a credit spread shock. Brunnermeier et al. (2021) use the identification via a heteroskedasticity-based approach to identify a monetary policy shock and two “stress” shocks originating in the financial sector and propagating to the real economy. However, we use a proxy TVAR specification that is closer to Carriero, Galvao, and Marcellino (2018). While they develop a variant of a smooth transition VAR model using a recursive ordering identification strategy, we rely on the forecast error based on the expectation formation mechanism of agents to identify a credit market sentiment shock.
To summarize, we provide several contributions to the literature. First, we use a nonlinear specification to disentangle periods of optimism and pessimism in the credit market. It is a fairly well-established fact in the literature on credit activity that financial markets operate and react strongly in times of distress than in times of tranquility.1 Second, we use a sophisticated shrinkage prior setup that exploits recent developments in the Bayesian literature on VARs (Huber and Feldkircher 2019) for an efficient estimation of our model. Third, we propose a novel identification mechanism inspired by the literature on identification via external instruments (Stock and Watson 2012, Mertens and Ravn 2013, Gertler and Karadi 2015) based on diagnostic expectations as an expectation formation mechanism (Bordalo, Gennaioli, and Shleifer 2018) in a TVAR setting. Fourth, we discuss and compare our results with other expectation formation mechanisms, primarily RE and a set of behaviorally and experimentally confirmed belief formation mechanisms discussed in Anufriev and Hommes (2012). This strategy allows us to identify a credit market sentiment shock, where unexpected news leads to different dynamics in regimes of optimism and pessimism.
Our results show that a credit market sentiment shock has two distinct features. First, there are strong asymmetries between different credit market regimes. When the credit market is calm, an unexpected news shock to the credit market sentiment has short-lived and small to muted effects on the credit and business cycles. On the contrary, in more turbulent times, when pessimistic sentiment is already prevalent in the economy, a shock to credit market sentiment induces severe negative effects on the business and the credit cycle. In addition, it also leads to a drop in prices. The economy starts to recover approximately after 1–2 years. Moreover, a comparison of different expectation formation mechanisms reveals heterogeneous reactions on impact on economic activity, prices, and short-term interest rates.
The remainder of the paper is organized as follows. After introducing our credit market sentiment indicator and the issue of belief formation in macroeconomics in Section 1, which presents the main identifying assumption in the model, we introduce our structural TVAR framework and the technical details of our identification strategy in Section 2. Our main results and the comparison of using different expectation formation mechanisms as well as a sensitivity analysis are discussed in Section 3 while Section 4 concludes. In an online appendix (Boeck and Zörner 2023), we provide additional material on the technical details of the paper and provide additional results.
APA:
Böck, M., & Zörner, T.O. (2023). The Impact of Credit Market Sentiment Shocks. Journal of Money, Credit and Banking. https://doi.org/10.1111/jmcb.13109
MLA:
Böck, Maximilian, and Thomas O. Zörner. "The Impact of Credit Market Sentiment Shocks." Journal of Money, Credit and Banking (2023).
BibTeX: Download