This paper presents empirical evidence on the existence of structural breaks in the fundamentals process underlying US stock prices. We develop an asset pricing model that represents breaks in the context of a Markov switching process with an expanding set of non-recurring states. Different hypotheses on how investors form expectations about future dividends after a break are proposed and analyzed. A model in which investors do not have full information about the parameters of the dividend process but gradually update their beliefs as new information arrives is shown to induce skewness, kurtosis, volatility clustering and the serial correlation in stock returns after a break.

The focus of my dissertation is the study of stock market predictability. More precisely, I use econometric tools to understand, explain, and predict aggregate and cross-sectional patterns in stock prices. Predictability of aggregate stock market returns and dividend growth is a widely studied topic, of great interest to both academics and practitioners. It is related to theories of market efficiency and information diffusion, both rational and behavioral. It also allows us to determine which types of information generate the movements in stock prices that we observe. Understanding why stock prices move and what factors drive their variation is critical from theoretical and policy-making perspectives. Chapter 1 of my dissertation revisits one of the main findings of the predictability literature, namely that all variation in aggregate stock prices is explained by changes in aggregate risk through discount rates and none by news about firms' expected cash flows. I propose a more comprehensive measure of dividends that includes M&A cash flows and show that dividend growth is predictable and that cash flow news explains around 60% of the observed variation in prices, while the remaining 40% is accounted for by discount rate news. Chapter 2 shows that information about fundamentals of the aggregate economy derived from closely held firms help predict stock returns of public firms. A common feature of most stock market predictors is that they are constructed using financial data of public firms. I construct a new economy-wide dividend-price ratio that takes into account dividends and market capitalization of both listed (public) and non-listed (private) U.S. companies and show that it strongly predicts stock returns both in-sample and out-of-sample. I also find that changes in dividends of private firms lead those of public firms and that the economy-wide dividend-price ratio subsumes the standard dividend-price ratio in predictive regressions. Chapter 3, co-authored with Christopher A. Parsons and Sheridan Titman, explores geographic momentum: a positive lead-lag stock return relation between neighboring firms operating in different sectors. It shows that a portfolio of firms headquartered in the same area, but operating in different sectors, strongly forecasts individual stock returns up to one year ahead. The economic significance of a city-momentum trading strategy is of similar magnitude to that observed with industry momentum. However, while industry momentum is strongest among thinly traded, small firms, and/or those with scant analyst following, geographic momentum is unrelated to these proxies for information processing. We propose an explanation linking this to the structure of the investment analyst business, which is organized by sector, rather than by geographic region.

Economic modeling in a data-rich environment is often challenging. To allow for enough flexibility and to model heterogeneity, models might have parameters with dimensionality growing with (or even much larger than) the sample size of the data. Learning these high-dimensional parameters requires new methodologies and theories. We consider three important high-dimensional models and propose novel methods for estimation and inference. Empirical applications in economics and finance are also studied.

In Chapter 1, we consider high-dimensional panel data models (large cross sections and long time horizons) with interactive fixed effects and allow the covariate/slope coefficients to vary over time without any restrictions. The parameter of interest is the vector that contains all the covariate effects across time. This vector has dimensionality tending to infinity, potentially much faster than the cross-sectional sample size. We develop methods for the estimation and inference of this high-dimensional vector, i.e., the entire trajectory of time variation in covariate effects. We show that both the consistency of our estimator and the asymptotic accuracy of the proposed inference procedure hold uniformly in time. Our methodology can be applied to several important issues in econometrics, such as constructing confidence bands for the entire path of covariate coefficients across time, testing the time-invariance of slope coefficients and estimation and inference of patterns of time variations, including structural breaks and regime switching. An important feature of our method is that it provides inference procedures for the time variation in pre-specified components of slope coefficients while allowing for arbitrary time variation in other components. Computationally, our procedures do not require any numerical optimization and are very simple to implement. Monte Carlo simulations demonstrate favorable properties of our methods in finite samples. We illustrate our methods through empirical applications in finance and economics.

In Chapter 2, we consider large factor models with unobserved factors. We formalize the notion of common factors between different groups of variables and propose to use it as a general approach to study the structure of factors, i.e., which factors drive which variables. The spanning hypothesis, which states that factors driving one group are spanned by those driving another group, can be studied as a special case under our framework. We develop a statistical procedure for testing the number of common factors. Our inference procedure is built upon recent results on high-dimensional bootstrap and is shown to be valid under the asymptotic framework of large $n$ and large $T$. In Monte Carlo simulations, our procedure performs well in finite samples. As an empirical application, we construct confidence sets for the number of common factors between the macroeconomy and the financial markets.

Chapter 3 is joint work with Jelena Bradic. We propose a methodology for testing linear hypothesis in high-dimensional linear models. The proposed test does not impose any restriction on the size of the model, i.e. model sparsity or the loading vector representing the hypothesis. Providing asymptotically valid methods for testing general linear functions of the regression parameters in high-dimensions is extremely challenging -- especially without making restrictive or unverifiable assumptions on the number of non-zero elements. We propose to test the moment conditions related to the newly designed restructured regression, where the inputs are transformed and augmented features. These new features incorporate the structure of the null hypothesis directly. The test statistics are constructed in such a way that lack of sparsity in the original model parameter does not present a problem for the theoretical justification of our procedures. We establish asymptotically exact control on Type I error without imposing any sparsity assumptions on model parameter or the vector representing the linear hypothesis. Our method is also shown to achieve certain optimality in detecting deviations from the null hypothesis. We demonstrate the favorable finite-sample performance of the proposed methods, via a number of numerical and a real data example.

Recent evidence suggests that many economic time series are subject to structural breaks. In the presence of breaks, including historical data prior to the most recent break to estimate a forecasting model will lead to prediction errors that are biased but also may have a smaller variance. This paper examines the trade-off between the bias and variance of forecast errors and proposes a new set of reversed Cusum procedures to determine the window size that minimizes mean squared forecast error. This window size varies over time and depends on the size of the break, the distance to the break and the squared correlation coefficient between predicted and realized values. The forecasting performances of several procedures for determination of window size are compared in a simulation experiment and in a recursive prediction exercise using data on US stock returns. We find evidence that out-of-sample forecasting performance can be improved by explicitly accounting for breaks and adopting the proposed method for optimally determining the window size

Existing results on the properties and performance of forecast combinations have been derived in the context of mean squared error loss. Under this loss function empirical studies have generally found that estimates of optimal forecast combination weights lead to higher losses than equally-weighted combined forecasts which in turn outperform the best individual predictions. We show that this and other results can be overturned when asymmetries are introduced in the loss function and the forecast error distribution is skewed. We characterize the optimal combination weights for the most commonly used alternatives to mean squared error loss and demonstrate how the degree of asymmetry in the loss function and skews in the underlying forecast error distribution can significantly change the optimal combination weights. We also propose estimation methods and investigate their small sample properties in simulations and in an inflation forecasting exercise.

This dissertation is comprised of three free-standing chapters, each focused on topics in the intersection of financial economics and macroeconomics.

Chapter 1 examines the propagation of shocks in economic models with customer-supplier networks. I show that the common assumption of idiosyncratic shocks at the firm or industry levels implies empirically implausible sparsity restrictions on the input-output network structure. Moreover, I provide evidence that substitutability between trade partners is related to technological and product dispersion that is not captured by standard firm and industry definitions, and thus generates non-negligible correlation in shocks. Finally, I show that assets positively exposed to upstream and downstream shocks are useful hedges and earn lower average risk premia than less exposed peers. This is confirmed by statistically significant return spreads and a negative association between correlated shock propagation and aggregate growth.

Chapter 2 studies the role of time-to-build in federal defense when estimating aggregate federal government spending multipliers. We find that the early impact of defense news shocks on GDP is due to a rise in business inventories, as contractors ramp up production for new defense contracts. These contracts do not affect government spending (G) until payment-on-delivery, which occurs 2-3 quarters later. Novel data on defense procurement obligations reveals that contract awards Granger-cause shocks to G identified via Cholesky decomposition, but not defense news shocks. We show that Cholesky shocks to G miss early changes in inventories, and thus result in lower multiplier estimates relative to the narrative method.

Chapter 3 explores the permanent-transitory decomposition of stochastic discount factor (SDF) processes in dynamic asset pricing models, in which the permanent component captures pricing at long payoff horizons. Analytic solutions for the permanent component are limited, and standard numerical methods are not well-suited to solve for them due to the curse of dimensionality and lack of boundary conditions and/or parametric assumptions. We propose a novel algorithm for computing the permanent-transitory decomposition for a general class of asset pricing models without such restrictions. We validate the algorithm's accuracy in several workhorse structural asset-pricing models, and argue that our approach applies to models whose state dynamics follow general and potentially high dimensional Levy processes.

This dissertation consists of three chapters that analyze the economic information contained in option markets. Option markets are forward looking, and thus contain valuable insight into the beliefs of financial market participants. They can be used to study risk premia and to make forecasts. The Chapter 1, Central Bank Tone and Currency Risk Premia, asks how the tone of central bank press conferences impacts risk premia in the currency market. First, I find that option implied risk aversion increases when central banks are hawkish, and decreases when central banks are dovish. Second, I find that hawkish central bank tone predicts higher future variance risk premia, and vice versa. One explanation for this result is that the tone of a press conference indicates to investors the likelihood of central bank intervention, conditional on the state of the economy. Chapter 2, Monetary Stimulus and Perception of Risk, investigates the relationship between monetary stimulus and the perception of risk in financial markets, and how this varies across asset classes. First, I document a positive relationship between monetary stimulus and the perception of risk in equity, commodity, and currency markets. I document a negative relationship between monetary stimulus and the perception of risk in bond markets. Second, I establish a cointegrating relationship between monetary stimulus and implied volatility, indicating a positive long run equilibrium relationship in the levels of monetary stimulus and implied volatility. This relationship is present across asset classes. Third, I document the link between monetary stimulus and expected inflation, a possible mechanism by which monetary stimulus affects the perception of risk across financial markets. Chapter 3, Option Augmented Density Forecasts of Market Return with Monotone Pricing Kernel, considers consider an option augmented density forecast of the market return obtained by transforming a baseline density forecast estimated from past excess returns so as to monotonize its ratio with a risk neutral density estimated from current option prices. We find that monotonizing the pricing kernel leads to a modest improvement in the calibration of density forecasts.

This study examines evidence of structural breaks in models of predictable components in stock returns related to state variables such as the lagged dividend yield, Treasury bill rate, term spread and default premium. We examine a large set of size-and-industry-sorted profolios of US stocks as well as 18 international stock market profolios and find systematic evidence of breaks in the vast majority of porfolios. The breakpoints most frequently identified in the US data are 1966, 1974, 1983, and 1990. The 1966 and 1974 breaks appear to have been driven by the T-bill rate and the default premium coefficients, while the 1983 break reflects changes in the coefficient on the T-bill rate and the term spread and the 1990 break was driven by the dividend yield and default premium coeffciencts. Our evidence also suggests that, while the size of the predictable component in stock returns has come down after the most recent break, many predictors continue to be significant. Although in-sample predictability of returns was lower in the 1990s than in some previous decades, it does not seem to have

My thesis has two themes: The first theme is about studying investors' expectations and the relation to asset prices; while the second theme is about evaluating forecasting performance. Both themes focus on what we can learn from a panel of data. The first chapter of my dissertation studies rational investors' expectation of consumption growth at the presence of structure breaks and asset pricing implications. While the first chapter studies how rational individuals should do, the second and third chapters focus on forecasters' behavior in real world, by developing tools to evaluate forecasters' performance about multiple variables, across many forecasters and at single time periods.

In Chapter 1, we use data on multiple consumption goods to identify infrequent, but persistent breaks to consumption growth dynamics. Over a sixty-year sample, we find four breaks, all of which are associated with major macroeconomic and financial market events such as oil price shocks, the Great Moderation, the end of the tech stock market bubble, and the Covid pandemic. The impact of the breaks on consumption growth is highly uncertain and heterogeneous across consumption goods. We explore the asset pricing implications of our novel empirical evidence in the context of a Lucas tree model in which investors use information on multiple consumption goods to learn about model parameters. We find that break risk in consumption growth, combined with investor learning, helps resolve a number of asset pricing puzzles such as high risk premium and volatility of market returns, as well as cross-sectional anomalies such as momentum.

Chapter 2 is joint work with Allan Timmermann and Yinchu Zhu. Forecasting skills are often identified by comparing predictive accuracy across large numbers of forecasts. This generates a multiple hypothesis testing problem that can trigger many false positives. We develop a new bootstrap test approach for identifying superior predictive accuracy that applies to multi-dimensional panel settings with arbitrarily many forecasts, outcome variables, horizons, and time periods. Our approach controls the family-wise error rate while retaining the ability to identify truly skilled forecasters. An empirical analysis of the IMF's World Economic Outlook forecasts across 185 countries, five variables and several forecast horizons shows how our approach can be used to identify variables and countries for which the IMF's forecasts improve significantly at shorter horizons as well as cases where they fail to improve.

Chapter 3 is also joint work with Allan Timmermann and Yinchu Zhu. We develop new methods for pairwise comparisons of predictive accuracy with cross-sectional data. Using a common factor setup, we establish conditions on cross-sectional dependencies in forecast errors which allow us to test the null of equal predictive accuracy on a single cross-section of forecasts. We consider both unconditional tests of equal predictive accuracy as well as tests that condition on the realization of common factors and show how to decompose forecast errors into exposures to common factors and idiosyncratic components. An empirical application compares the predictive accuracy of financial analysts’ short-term earnings forecasts across six brokerage firms.