Research
Working papers
Abstract: This paper investigates the real impacts of green investing driven by both financial value and ethical values. Specifically, it examines how firms interact with investors who have a “warm glow” effect in the presence of environmental regulatory risks. Green investing encourages firms to reduce social costs, even in scenarios where sustainable investors lack bargaining power and profit-driven capital is perfectly elastic in supply. From a values perspective, the adoption of clean technology is incentivized by having access to financing at lower rates, while from a value perspective, the size of the division employing clean technology expands due to its lower return variance. Sustainable investing has the potential to enhance social welfare by reducing social costs. However, it also contributes to structural inequality by shifting financial gains from sustainable investors to firms. To mitigate such inequality, policymakers could consider implementing tax transfers to benefit sustainable investors. An alternative to achieving impacting investing without incurring structural inequality is lobbying. Green investors can be better off in a lobbying game by obtaining a “warm glow” effect without sacrificing financial returns. Furthermore, high competition for limited green capital might drive up costs, thereby constraining the impacts of green investing.
2. "Forecasting the Forecasts of Others on Social Networks" with Fei Tan (JMP - Economics)
Abstract: While real-world networks are often strongly connected, significant divergence in opinions remains prevalent. This paper seeks to explain this “failure of consensus” puzzle by introducing a dynamic game model where agents observe their neighbors’ opinions via social networks but are subject to either intentional or unintentional communication noise. The information individuals receive from others often contains noise that individuals need to filter so that agents have to “forecast the forecasts of others”. This process involves higher-order beliefs, making the equilibrium challenging to characterize. We show how such equilibrium behaviors can be characterized given any network structure and continuous action space. The results suggest that divergence in opinion might exist, even when agents are rational and the network is strongly connected as long as some agents share noisy information. Our results also indicate that agents with high-precision information are motivated to share their information truthfully when they exhibit strategic complementary, but not when they exhibit strategic substitution.
3. "Climate Preparedness and Cross Section" with Siddhartha Chib and Kuntara Pukthuanthong
Abstract: In this paper, we build a theoretical two-period investment-based asset pricing model to examine the link between firm-level expected return and its preparation for climate risk. Climate risk is modeled by a climate change-induced event (CCIE) that can occur in the second period. The damage this causes is inversely dependent on the spending in the first period on climate preparedness. The firm is endowed with an initial capital stock, profitability, and readiness. We show that the expected return to equity in equilibrium depends positively on climate change preparedness, firm profitability, size, and negatively on investment. An empirical five-factor-based counterpart of the theoretical model shows promise in the pricing of stocks and ETFs.
4. "Policy Rule Regressions with Survey Data" with Fei Tan and Zheliang Zhu
Abstract: Endogeneity bias is ubiquitous and can be sizable in the ordinary least squares (OLS) estimation of policy rule coefficients. This paper introduces a simple procedure that leverages survey data to rectify the bias under flexible information assumptions. We decompose policy rule regressors (e.g., inflation and output gap) into their forecasts made before policy decisions and the associated forecast errors. The forecasts are readily available in survey data and, by construction, orthogonal to the forecast errors and policy shocks under complete information. We further orthogonalize the forecast errors to remove the bias in the presence of information rigidities. Using Monte Carlo simulations, we showcase the efficacy of this procedure in a prototypical new Keynesian model under distinct information settings. As an empirical application, we employ Bayesian methods to compare the performance of the standard OLS approach using real-time data and our approach based on survey data in estimating monetary and fiscal policy rules. Marginal likelihood estimates reveal that our approach consistently outperforms across nearly all model specifications considered.
Abstract: ESG investing is associated with various risks, such as physical risks and regulatory transition risks. Is the ESG-related factor a risk factor and helps with cross-section pricing? If it is indeed a risk factor, which additional risk factors should be incorporated alongside ESG to potentially influence pricing in the cross-section? I employ the Bayesian model scan method, which evaluates various risk factor combinations. I consider 10 prominent factors in the literature, plus one ESG factor constructed by the 3 by 2 method. Each factor is either a risk factor or a non-risk factor, which generates 2047 model combinations. Among those combinations, The Bayesian model scan result suggests a 4-factor model (BS4), including Mkt, EG, ESG, and PEAD, which appears in the top 3 models. The proposed BS4 model shows promise in pricing risk factors of other asset pricing models, as well as a broad range of return anomalies and stocks.
6. "Estimating the Signaling Channel of QEs" with with Joe Kachovec
Abstract: This paper proposes measuring the signalling effects of open-ended quantitative easing programs by using an augmented Taylor rule for a central bank’s interest rate policy. Monetary policy regimes are estimated using a Bayesian VAR subject to change-points to conduct a monetary rule “regime change” statistical test. The period of near zero policy rates (ZLB) and quantitative easing (QE) is found to be a monetary regime change point. The Taylor coefficients are estimated via OLS and time varying VAR. During well-defined QE-programs the Taylor coefficient on the size of the Federal Reserve’s balance sheet is statistically insignificant. During open-ended QE-programs the magnitude of this coefficient is negative and statistically different from zero.
Abstract: U.S. stocks with high ESG scores outperformed the ones with low ESG scores in recent years. However, theoretical models value ESG characteristics usually suggest that “green” assets have low expected returns. To fill this gap, I model ESG investing in that the representative agent has heterogeneous habit persistence to asset holdings in a two-tree Lucas model. More specifically, the habit-forming agent dislikes variations in habit-adjusted “green” holdings more than in “brown” holdings. In equilibrium, “green” trees have a high expected return because the agent is more risk-averse to “green” holding in bad times, when the “green” holding is low relative to its history than in good times, which is compensated by high expected returns.
Work in progress
1. "Tale of Two Tightenings: Sustainable Finance Regulations in the Business Cycle"
Abstract: Climate change caused by carbon emissions has drawn remarkable attention in recent years. The carbon tax has been proven to effectively reduce carbon emissions but involves high political costs. This paper instead introduces two alternative sustainable finance regulations to reduce carbon emissions, which are implemented by financial sectors independent of the government, The first regulation is tightening borrowing constraints in “brown” firms. The second one is increasing the “green” capital requirements in the bank sector. I quantify to impacts of the two regulations and compare them with the carbon tax by merging a two-sector DSGE framework with an environmental model.
Link to slides Preliminary draft upon request
2. "Non-revisiting regime shifts in Taylor principles"
Abstract: This paper extends the Taylor principle to a framework where the reaction coefficients in the monetary policy rule follow a regime-switching process that does not revisit previous states. Specifically, monetary policy operates in distinct regimes that never recur, with the economy either remaining in its current state or transitioning to a subsequent one. I derive a long-run Taylor principle that ensures unique equilibria within a standard model by applying backward induction. Furthermore, I develop algorithms to quantitatively estimate a general New Keynesian model incorporating non-revisiting regime-switching in monetary policy.
Link to slides Preliminary draft upon request