Using a large sample of forecasts of 22 firm-level variables, we document three facts: (i) the relationship between forecast revisions and future forecast errors is strongly non-linear, (ii) the distributions of the underlying processes have fat tails, and (iii) extreme realizations tend to mean-revert. Next, we formally show that these three facts are consistent with a model in which the underlying process is non-Gaussian, but forecasters fail to recognize this. Finally, we provide additional evidence in support of our theory by showing that it can explain evidence from an online forecasting experiment, that the non-linear relationship between errors and revisions is also present in macroeconomic forecasts, and that it provides an explanation for the presence of non-linearity in the momentum of stock returns.

Using existing and newly designed household surveys, we show that households close to durable purchases acquire more information about interest rates and have more accurate, less dispersed, and less uncertain interest rate expectations. We use this evidence of selective inattention to calibrate an incomplete markets model with durable consumption and endogenous information acquisition about interest rates. Relative to exogenous inattention, selective inattention shifts the composition of aggregate consumption responses to interest rate cuts toward durables, accelerates the impact of larger rate cuts, and generates dampened responses to changes in rate volatility that are closer to empirical evidence.

We develop and implement a novel method to study personal financial advice from Large Language Models (LLMs). Studying this advice is challenging because it depends on the model used (i.e., supply), the questions individuals ask (i.e., demand), and their evolving circumstances. We address these challenges by surveying a representative sample of adults and asking them to write prompts seeking spending and investing advice from an LLM. We then simulate the lifetime paths that result from following this advice under realistic asset and labor market conditions. Applying our method to GPT-5.2 and Gemini 3.0 Flash, we document three facts about AI-generated financial advice. First, following LLM advice would move most survey respondents closer to the prescriptions of life cycle theory relative to their current behavior, including broader participation in diversified equity funds, equity shares that decline with age, and sizeable saving buffers. Second, replacing individual-written prompts with academic prompts moves LLM advice even closer to life cycle theory, with better consumption smoothing and less reliance on simple heuristics. Third, LLM advice varies systematically with individual characteristics, such as gender and financial literacy. These differences accumulate over the life cycle into wealth differences at retirement of 4-5% between groups and reflect both demand (i.e., systematic variation in the prompts written by different individuals) and supply (i.e., differences in advice for a given prompt). These facts highlight the potential of generative AI to improve financial decision-making, but suggest that its impact is likely heterogeneous across households and depends on how the technology is used.

Traditional dynamic programming requires a mathematical model of the transition function for the state vector. Leveraging reinforcement learning techniques, we develop a framework to solve dynamic optimization problems that does not require modeling the data-generating process (DGP) of exogenous states. Instead, the method samples realizations of these states directly from the data, allowing the modeler to be "agnostic" about the DGP. We apply our method to a canonical life cycle consumption-saving problem, solving the model without specifying the DGP for income. Using income data from the CPS, we find that the welfare loss from using a standard parametric income process relative to placing no restrictions on the DGP is small. We conclude by verifying that our method achieves a global optimum when given a known DGP and discussing directions for future work.

We study the role of risk preferences and frictions in portfolio choice using variation in 401(k) default options. Patterns of active choice in response to different default funds imply that, absent participation frictions, 94% of investors prefer holding stocks, with an equity share of retirement wealth declining with age---patterns markedly different from observed allocations. We use this quasi-experiment to estimate a life-cycle model and find a relative risk aversion of 2.5, EIS of 0.25, and $160 portfolio adjustment cost. The results suggest that low levels of stock market participation in retirement accounts are due to participation frictions rather than nonstandard preferences such as loss aversion.

We document the growth of retail options trading and provide evidence that retail investors are drawn to options by anticipated spikes in volatility. Retail investors purchase options in a concentrated fashion before earnings announcements, particularly those with greater expected abnormal volatility. Comparing across asset markets, we also find retail investors disproportionately trade options over stocks as anticipated announcement volatility increases. In doing so, retail investors display a trio of wealth-depleting behaviors: they overpay for options relative to realized volatility, incur enormous bid-ask spreads, and do not close their positions until weeks after announcements. These translate to retail losses of 5-to-9% on average, and 10-to-14% for high expected volatility announcements.

Student loans with income-contingent repayment insure borrowers against income risk but can reduce their incentives to earn more. Using a change in Australia's income-contingent repayment schedule, I show that borrowers reduce their labor supply to lower their repayments. These responses are larger among borrowers with more hourly flexibility, a lower probability of repayment, and tighter liquidity constraints. I use these responses to estimate a dynamic model of labor supply with frictions that generate imperfect adjustment. My estimates imply that the labor supply responses to income-contingent repayment limit the optimal amount of insurance in government-provided student loans. However, these responses are too small to justify fixed repayment contracts: restructuring existing student loans from fixed repayment to a constrained-optimal income-contingent loan---while keeping the tax and transfer system unchanged---increases borrower welfare by the equivalent of a 0.8% increase in lifetime consumption at no additional fiscal cost.

Analyst forecasts outperform econometric forecasts in the short run but underperform in the long run. We decompose these differences in forecasting accuracy into analysts’ information advantage, forecast bias, and forecast noise. We find that noise and bias strongly increase with forecast horizon, while analysts’ information advantage decays rapidly. A noise increase with horizon generates a mechanical reversal in the sign of the error-revision (Coibion--Gorodnichenko) regression coefficient at longer horizons, independently of over/underreaction. A parsimonious model with bounded rationality and a noisy cognitive default matches the term structures of noise and bias jointly.

Over the past couple of decades, the number of volatility indices has increased rapidly. Although the dynamics of realized volatility spillover have been studied extensively, very few studies exist that examine the spillover between these implied volatility indices. By using DAG-based structural vector autoregression, this paper provides evidence that implied volatility spillover differs from realized volatility spillover. This paper finds that Asia, more specifically Hong Kong, plays a central role in implied volatility spillover during and after the 2008 financial crisis.

Intuitively, one would expect that internet search volume would contain valuable information about investor sentiment for a company. With the development of new data sources, such as Google Trends, this relationship can be more easily and objectively examined. This paper seeks to examine the relationship between a company’s stock price volatility and its Google search volume. A small cross-section of twenty companies is considered, and the goal of this paper is to demonstrate the power of Google Trends data in hope of initiating further research. Using a conventional GARCH framework for financial market volatility, an economically and statistically significant contemporaneous relationship between Google search volume and equity volatility is found.