Federal Reserve Bank of St. Louis working papers are preliminary materials circulated to stimulate discussion and critical comment.
The geography of wealth: shocks, mobility, and precautionary savings
The spatial distribution of wealth in the United States is very heterogeneous, with important differences within and across US states. We study the distribution of wealth in a country and how it is shaped by the characteristics earnings across regions, and by the frictions individuals face to move and reallocate across space. For this, we develop a tractable model of consumption, savings, and location choice with many regions, incomplete markets, and heterogeneous agents facing persistent and transitory income shocks. Our analysis focuses on the role of income shocks, precautionary savings, mobility, and sorting in shaping the geographic distribution of income and wealth over time. Our theory extends the workhorse macroeconomic model of consumption and savings under uncertainty and risk to an economy with multiple labor markets and costly mobility. Despite the complex spatial and individual heterogeneity, we can characterize the optimal consumption, savings, and mobility decisions of workers in closed form. Mobility frictions increase precautionary savings as workers hedge against sharp fluctuations in consumption generated by their mobility decisions. The spatial distribution of wealth is primarily driven by the interaction between persistent income shocks, saving behavior, and worker sorting across locations. The results highlight the importance of accounting for worker mobility and regional heterogeneity in earnings dynamics when studying the spatial distribution of wealth.
Battles and Diseases in the U.S. Civil War
Wars create short-term fluctuations in mortality. Belligerents might mitigate their own casualties with larger armies that hinder their opponent’s fighting ability. But diseases are frequent in wars and, thus, may reduce the benefits of larger armies. First, I analyze these competing mechanisms in a dynamic model of wartime attrition. Second, I calibrate the model using U.S. Civil War data and find that if the Union had fielded a 50%-larger army in 1861, Union casualties would have been marginally lower. The theory provides the insight for this quantitative result.
Nominal Maturity Mismatch and the Liquidity Cost of Inflation
We document a liquidity channel through which unexpected inflation generates substantial welfare losses. Household balance sheets are nominal maturity mismatched: nominal liabilities have a longer duration than nominal assets. Due to this mismatch, losses from unexpected inflation are concentrated over short time horizons, while gains are spread out over the longer run. This has negative effects on liquidity-constrained households, who cannot easily borrow against their future gains. We quantify the importance of the liquidity channel and show that, for households in the lower half of the wealth distribution, the recent 2021–2022 unexpected inflation shock caused welfare losses valued at 0.5% of lifetime wealth: a monetary loss equal in size to 15% of current- year consumption. More than 75% of that loss is due to the liquidity channel, with the remainder coming from the more commonly studied wealth channel.
College Access and Intergenerational Mobility
This paper studies how college admissions preferences for low income students affect intergenerational earnings mobility. We develop a quantitative model of college choice with quality differentiated colleges. We find that admissions preferences substantially increase low income enrollment in top quality colleges and intergenerational earnings mobility. The associated losses of aggregate earnings are very small.
What the Transcripts Reveal About the FOMC’s Pre-Emptive Easing in July 1995
At their September 2024 meeting, the FOMC began reducing the federal funds target rate and indicating the likelihood of additional reductions by the end of the year and into 2025. The FOMC took this action because of favorable inflation trends and some developing weakness in labor markets. A similar dynamic was at work from 1994 to early 1996. During this period, the FOMC undertook, first, a pre-emptive tightening in policy to combat emerging price pressures and then, second, a pre-emptive easing of monetary policy to counter the expectations of slower real GDP growth or outright recession. One key difference between the two episodes was the marked acceleration in inflation rate in 2021-2022 compared to 1994-95. Nevertheless, the end result of the 1994-96 episode was that the US economy avoided a recession and inflation by the end of 1997 was effectively at a level that is now deemed price stability. The purpose of this article is to outline the key arguments that Chairman Greenspan and the other FOMC participants deployed during the 1995-96 pre-emptive easing episode.
Cross-border Patenting and the Margins of International Trade
This paper investigates the impact of cross-border patenting on the margins of international trade using disaggregated data on international patenting and trade flows. We develop a theoretical framework of trade and firms' patenting decisions that motivates our empirical analysis. The main results reveal that cross-border patenting has a larger effect on the extensive margin of trade compared to the intensive margin. This finding suggests that firms tend to seek patent protection in international markets prior to entering those markets with new products, rather than with their existing products.
The Rapid Adoption of Generative AI
Generative Artificial Intelligence (AI) is a potentially important new technology, but its impact on the economy depends on the speed and intensity of adoption. This paper reports results from the first nationally representative U.S. survey of generative AI adoption at work and at home. In August 2024, 39 percent of the U.S. population age 18-64 used generative AI. More than 24 percent of workers used it at least once in the week prior to being surveyed, and nearly one in nine used it every workday. Historical data on usage and mass-market product launches suggest that U.S. adoption of generative AI has been faster than adoption of the personal computer and the internet. Generative AI is a general purpose technology, in the sense that it is used in a wide range of occupations and job tasks at work and at home.
The Implications of Labor Market Heterogeneity on Unemployment Insurance Design
We digitize state-level and time-varying unemployment insurance (UI) laws on initial eligibility, payment amount, and payment duration and combine them with microdata on labor market outcomes to estimate UI eligibility, take-up, and replacement rates at the individual level. We document how levels of income and wealth affect unemployment risk, eligibility, take-up, and replacement rates both upon job loss and over the course of unemployment spells. We evaluate whether these empirical findings are important for shaping UI policy design using a general equilibrium incomplete markets model combined with a frictional labor market that matches our empirical findings. We show that a nested alternative model that fails to match these findings yields a substantially less generous optimal UI policy compared to the baseline model. Our empirical results are also relevant for researchers estimating the effects of UI policy changes on labor market outcomes.
Mismatch Unemployment During COVID-19 and the Post-Pandemic Labor Shortages
We examine the extent to which mismatch unemployment—employment losses relative to an efficient allocation where the planner can costlessly reallocate unemployed workers across sectors to maximize output—shaped labor market dynamics during the COVID-19 pandemic and the subsequent recovery episode characterized by labor shortages. We find that, for the first time in our sample, mismatch unemployment turned negative at the onset of the pandemic. This result suggests that the efficient allocation of job seekers would involve reallocating workers toward longer-tenure and more-productive jobs, even at the expense of fewer hires. We show that sectoral differences in job separations were the main driver behind this result, while differences in vacancies caused positive mismatch unemployment during the recovery episode. We also establish an empirical link between mismatch unemployment and the surge in the labor cost during the recovery, documenting that sectors with larger mismatch unemployment experienced higher employment cost growth.
Hours Worked and Lifetime Earnings Inequality
We document large differences in lifetime hours of work using data from the NLSY79 and argue that these differences are an important source of inequality in lifetime earnings. To establish this we develop and calibrate a rich heterogeneous agent model of labor supply and human capital accumulation that allows for heterogeneity in preferences for work, initial human capital and learning ability, as well as idiosyncratic shocks to human capital throughout the life-cycle. Our calibrated model implies that almost 20 percent of the variance in lifetime earnings is accounted for by differences in lifetime hours of work, with 90 percent of this effect due to heterogeneity in preferences. Higher lifetime hours contribute to lifetime earnings via two channels: a direct channel (more hours spent in production at given productivity) and a human capital channel (more hours spent investing in human capital, which increases future productivity). Between a third and a half of the effect of lifetime hours on lifetime earnings is due to the human capital channel. Our model implies that policies that limit long hours have important effects on both the mean and variance of lifetime earnings.
Measuring Trends in Work From Home: Evidence from Six U.S. Datasets
This paper documents the prevalence of work from home (WFH) in six U.S. data sets. These surveys measure WFH using different questions, reference periods, samples, and survey collection methods. Once we construct samples and WFH measures that are comparable across surveys, all surveys broadly agree about the trajectory of aggregate WFH since the Covid-19 outbreak. The surveys agree that pre-pandemic differences in WFH rates by sex, education, and state of residence expanded following the Covid-19 outbreak. The surveys also show similar post-pandemic trends in WFH by firm size and industry. Finally, we highlight that an important source of quantitative differences in WFH across surveys is WFH by self-employed workers; by contrast, surveys closely agree on rates of WFH among employees.
Economic Surveillance using Corporate Text
This article applies simple methods from computational linguistics to analyze unstructured corporate texts for economic surveillance. We apply text-as-data approaches to earnings conference call transcripts, patent texts, and job postings to uncover unique insights into how markets and firms respond to economic shocks, such as a nuclear disaster or a geopolitical event—insights that often elude traditional data sources. This method enhances our ability to extract actionable intelligence from textual data, thereby aiding policy-making and strategic corporate decisions. By integrating computational linguistics into the analysis of economic shocks, our study opens new possibilities for real-time economic surveillance and offers a more nuanced understanding of firm-level reactions in volatile economic environments.
A High-Frequency Measure of Income Inequality
To identify shocks in VARs using short-run sign or exclusion restrictions, the highest-frequency data possible is usually preferred. For income inequality, there is tension between high frequency and high quality. Annual datasets that survey large numbers of people provide high-quality estimates of income. Higher frequency surveys generally provide a sparser sampling of individual income. Some previous studies have used the the higher frequency data, presumably to match the frequency necessary to identify the shock. Using data obtained from the higher frequency, lower respondant surveys might result in misleading conclusions. We combine the two surveys and construct a set of quarterly-frequency income quantiles that are scaled to the annual data but fluctuate according to the high-frequency survey. We show that using these data yields very different economic conclusions than simply using the raw high-frequency income series. In particular, we show in two simple applications to technology shocks and house price shocks that one obtains different conclusions about the permanence and/or the direction of the responses of income inequality.
The Impact of the 2017 Tax Cuts and Jobs Act on U.S. Multinationals’ Intangible Assets
This paper investigates the impact of the 2017 Tax Cuts and Jobs Act (TCJA) on U.S. multinationals’ intangibles. We develop a theoretical model that incorporates key provisions of the TCJA—the Global Intangible Low-Taxed Income (GILTI) and the Foreign-Derived Intangible Income (FDII)—and derive testable implications for changes in licensing and patent transfer patterns. Using data on international royalty flows and patent assignments, we test the model’s predictions. Our findings suggest that the TCJA may have impacted profit shifting strategies through intangibles, aligning with our model’s predictions.
Scalable vs. Productive Technologies
Do larger firms have more productive technologies or are their technologies more scalable, or both? We use administrative data on Canadian and US firms to estimate flexible nonparametric production functions. Our estimation results in a joint distribution of output elasticities of capital, labor, and intermediate inputs---therefore, returns to scale (RTS)---along with total factor productivity (TFP). We find significant heterogeneity in both RTS and TFP across firms. Larger firms operate technologies with higher RTS, both across and within industries. Higher RTS for large firms are entirely driven by higher intermediate input elasticities. Descriptively, these align with higher intermediate input revenue shares. We then incorporate RTS heterogeneity into an otherwise standard incomplete markets model with endogenous entrepreneurship that matches the observed heterogeneity in TFP and RTS. In this model, we find that the efficiency losses of financial frictions are more than twice as large relative to the conventional calibration that loads all heterogeneity on TFP and imposes a common RTS parameter.
Trade and Inequality in an Overlapping Generations Model with Capital Accumulation
We study the lifecycle aspect of within-country inequality that stems from capital and labor services supplied by individuals. Our environment is a combination of a multicountry trade model and an overlapping generations model with production and capital accumulation. Trade liberalization increases the measured total factor productivity in each country, which increases the marginal product of capital and incentivizes capital accumulation. Higher capital stock and higher measured productivity raise the marginal product of labor and, hence, wages. Inequality, measured by the ratio of old agents' income to young agents' income, evolves over time due to capital accumulation during the transition from autarky to an open-economy world. Immediately after liberalization, inequality increases. Over time, capital accumulates at a diminishing rate and inequality declines.
Dissecting the Great Retirement Boom
Between 2020 and 2023, the fraction of retirees in the working-age population in the U.S. increased above its pre-pandemic trend. Several explanations have been proposed to rationalize this gap, such as the rise in net worth due to higher asset returns, the labor market's deterioration due to higher unemployment risk, the expansion of fiscal support programs, and increased mortality risk. We quantitatively study the interaction of these factors and decompose their relative contribution to the recent rise in retirements using an incomplete markets, overlapping generations model with a frictional labor market. We find that all of these channels contribute to excess retirements, with labor market conditions being a more important driver in 2020-2021 and fiscal programs playing a larger role in 2022-2023. We show that our model's predictions on aggregate labor market moments and cross-sectional moments on retirement patterns across the wealth distribution are in line with the data.
An Empirical Analysis of the Cost of Borrowing
We examine borrowing costs for firms using a security-level database with bank loans and corporate bonds issued by U.S. companies. We find significant within-firm dispersion in borrowing rates, even after controlling for security and firm observable characteristics. Obtaining a bank loan is 132 basis points cheaper than issuing a bond, after accounting for observable factors. Changes in borrowing costs have persistent negative impacts on firm-level outcomes, such as investment and borrowing, and these effects vary across sectors. These findings contribute to our understanding of borrowing costs and their implications for corporate policies and performance.
Sovereign Debt Restructuring and Credit Recovery
This paper focuses on the significant growth of domestic credit once the debt is restructured and shows that is not correlated with the size of the haircut. Second, it performs an event study around Ecuador’s sovereign default and restructuring of 2008-2009 to study changes in domestic bank lending behavior. After external debt restructuring, private lending increased the most for banks highly exposed to public debt. Finally, it provides a simple model were uncertainty about the return on government external debt during default has spillover effects on the domestic economy by creating dispersion in beliefs across domestic banks, which leads to a misallocation of credit. External debt restructuring eliminates domestic belief heterogeneity by making the return on bonds observable to everyone. This simple framing is not only consistent with the substantial growth in domestic credit upon debt restructuring but also with its independence from the haircut size observed in the data.
This document contains a technical description of the dynamic stochastic general equilibrium (DSGE) model developed and maintained by the Research Division of the St. Louis Fed as one of its tools for forecasting and policy analysis. The St. Louis Fed model departs from an otherwise standard medium-scale New Keynesian DSGE model along two main dimensions: first, it allows for household heterogeneity, in the form of workers and capitalists, who have different marginal propensities to consume (MPC). Second, it explicitly models a fiscal sector endowed with multiple spending and revenue instruments, such as social transfers and distortionary income taxes. Both of these features make the model well-suited for the analysis of fiscal policy counterfactuals, and monetary-fiscal interactions. We describe how the model is estimated using historical data for the US economy and how the COVID-19 pandemic is accounted for. Some examples of model output are presented and discussed.
Correct order of authors: Hubmer, Halvorsen, Salgado, Ozkan. We use 1993--2015 Norwegian administrative panel data on wealth and income to study lifecycle wealth dynamics. By employing a novel budget constraint approach, we show that at age 50 the excess wealth of the top 0.1%, relative to mid-wealth households, is accounted for by higher saving rates (38%), inheritances (34%), returns (23%), and labor income (5%). One-fourth of the wealthiest---the "New Money"---start with negative wealth but experience rapid wealth growth early in life. Relative to the "Old Money," the New Money are characterized by even higher saving rates, returns, and labor income. We use these dynamic facts to test six commonly used models of wealth inequality. Although these models can generate the high concentration of wealth seen in the cross-section, they tend to put too much weight on (accidental) bequests and fail to capture the contribution of the New Money. A model with heterogeneous returns that decrease in wealth, and non-homothetic preferences is consistent with the new facts on the dynamics of wealth accumulation.
Work from Home and Interstate Migration
Interstate migration by working-age adults in the US declined substantially during the Great Recession and remained subdued through 2019. We document that interstate migration rose sharply following the 2020 Covid-19 outbreak, nearly recovering to pre-Great recession levels, and provide evidence that this reversal was primarily driven by the rise in work from home (WFH). Before the pandemic, interstate migration by WFH workers was consistently 50% higher than for commuters. Since the Covid-19 outbreak, this migration gap persisted while the WFH share tripled. Using quasi-panel data and plausibly exogenous changes in employer WFH policies, we address concerns about omitted variables or reverse causality and conclude that access to WFH induces greater interstate migration. An aggregate accounting exercise suggests that over half of the rise in interstate migration since 2019 can be accounted for by the rise in the WFH share. Moreover, both actual WFH and pre-pandemic WFH potential, based on occupation shares, can account for a sizable share of cross-state variation in migration.
A Theoretical Treatment of Foreign Fighters and Terrorism
The paper offers a game-theoretical model that includes three participants – the terrorist organization, its foreign fighters, and the adversarial host government. In stage 1, the terrorist group induces foreign fighters to emigrate through wage incentives, while the host government deters these fighters through proactive border security. Foreign fighters decide whether to emigrate from their source country (extensive margin) in stage 2, after which these fighters determine their level of attacks (intensive margin) in stage 3. Comparative statics to the Nash equilibrium are tied to changes in the employment or opportunity cost in the source country, as well as to changes in radicalization. Our basic model provides a theoretical foundation to recent empirical results. An extension involves a four-stage game with the host government assuming a leadership role prior to the terrorist group choosing its wage incentive.
After 40 Years, How Representative Are Labor Market Outcomes in the NLSY79?
In 1979, the National Longitudinal Study of Youth 1979 (NLSY79) began following a group of US residents born between 1957 and 1964. It has continued to re-interview these same individuals for more than four decades. Despite this long sampling period, attrition remains modest. This paper shows that after 40 years of data collection, the remaining NLYS79 sample continues to be broadly representative of their national cohorts with regard to key labor market outcomes. For NLSY79 age cohorts, life-cycle profiles of employment, hours worked, and earnings are comparable to those in the Current Population Survey. Moreover, average lifetime earnings over the age range 25 to 55 closely align with the same measure in Social Security Administration data. Our results suggest that the NLSY79 can continue to provide useful data for economists and other social scientists studying life-cycle and lifetime labor market outcomes, including earnings inequality.
CORRECT ORDER OF AUTHORS: Aakash Kalyani, Nicholas Bloom, Marcela Carvalho, Tarek Hassan, Josh Lerner, and Ahmed Tahoun. We identify phrases associated with novel technologies using textual analysis of patents, job postings, and earnings calls, enabling us to identify four stylized facts on the diffusion of jobs relating to new technologies. First, the development of new technologies is geographically highly concentrated, more so even than overall patenting: 56% of the economically most impactful technologies come from just two U.S. locations, Silicon Valley and the Northeast Corridor. Second, as the technologies mature and the number of related jobs grows, hiring spreads geographically. But this process is very slow, taking around 50 years to disperse fully. Third, while initial hiring in new technologies is highly skill biased, over time the mean skill level in new positions declines, drawing in an increasing number of lower-skilled workers. Finally, the geographic spread of hiring is slowest for higher-skilled positions, with the locations where new technologies were pioneered remaining the focus for the technology’s high-skill jobs for decades.
CORRECT ORDER OF AUTHORS: Aakash Kalyani, Nicholas Bloom, Marcela Carvalho, Tarek Hassan, Josh Lerner, and Ahmed Tahoun. We identify phrases associated with novel technologies using textual analysis of patents, job postings, and earnings calls, enabling us to identify four stylized facts on the diffusion of jobs relating to new technologies. First, the development of new technologies is geographically highly concentrated, more so even than overall patenting: 56% of the economically most impactful technologies come from just two U.S. locations, Silicon Valley and the Northeast Corridor. Second, as the technologies mature and the number of related jobs grows, hiring spreads geographically. But this process is very slow, taking around 50 years to disperse fully. Third, while initial hiring in new technologies is highly skill biased, over time the mean skill level in new positions declines, drawing in an increasing number of lower-skilled workers. Finally, the geographic spread of hiring is slowest for higher-skilled positions, with the locations where new technologies were pioneered remaining the focus for the technology’s high-skill jobs for decades.
CORRECT ORDER OF AUTHORS: Aakash Kalyani, Nicholas Bloom, Marcela Carvalho, Tarek Hassan, Josh Lerner, and Ahmed Tahoun. We identify phrases associated with novel technologies using textual analysis of patents, job postings, and earnings calls, enabling us to identify four stylized facts on the diffusion of jobs relating to new technologies. First, the development of new technologies is geographically highly concentrated, more so even than overall patenting: 56% of the economically most impactful technologies come from just two U.S. locations, Silicon Valley and the Northeast Corridor. Second, as the technologies mature and the number of related jobs grows, hiring spreads geographically. But this process is very slow, taking around 50 years to disperse fully. Third, while initial hiring in new technologies is highly skill biased, over time the mean skill level in new positions declines, drawing in an increasing number of lower-skilled workers. Finally, the geographic spread of hiring is slowest for higher-skilled positions, with the locations where new technologies were pioneered remaining the focus for the technology’s high-skill jobs for decades.
The Creativity Decline: Evidence from US Patents
Economists have long struggled to understand why aggregate productivity growth has dropped in recent decades while the number of new patents filed has steadily increased. I offer an explanation for this puzzling divergence: the creativity embodied in US patents has dropped dramatically over time. To separate creative from derivative patents, I develop a novel, text-based measure of patent creativity: the share of technical terminology that did not appear in previous patents. I show that only creative and not derivative patents are associated with significant improvements in firm level productivity. Using the measure, I show that inventors on average file creative patents upon entry, and file derivative patents with more experience. I embed this life-cycle of creativity in a growth model with endogenous creation and imitation of technologies. In this model, falling population growth explains 27% of the observed decline in patent creativity, 30% of the slowdown in productivity growth, and 64% of the increase in patenting.
Consumption Dynamics and Welfare Under Non-Gaussian Earnings Risk
CORRECT ORDER OF AUTHORS: Fatih Guvenen, Serdar Ozkan, and Rocio Madera. The order of coauthors has been assigned randomly using AEA’s Author Randomization Tool. Recent empirical studies document that the distribution of earnings changes displays substantial deviations from lognormality: in particular, earnings changes are negatively skewed with extremely high kurtosis (long and thick tails), and these non-Gaussian features vary substantially both over the life cycle and with the earnings level of individuals. Furthermore, earnings changes display nonlinear (asymmetric) mean reversion. In this paper, we embed a very rich “benchmark earnings process” that captures these non-Gaussian and nonlinear features into a lifecycle consumption-saving model and study its implications for consumption dynamics, consumption insurance, and welfare. We show four main results. First, the benchmark process essentially matches the empirical lifetime earnings inequality—a first-order proxy for consumption inequality—whereas the canonical Gaussian (persistent-plus-transitory) process understates it by a factor of five to ten. Second, the welfare cost of idiosyncratic risk implied by the benchmark process is between two-to-four times higher than the canonical Gaussian one. Third, the standard method in the literature for measuring the pass-through of income shocks to consumption—can significantly overstate the degree of consumption smoothing possible under non-Gaussian shocks. Fourth, the marginal propensity to consume out of transitory income (e.g., from a stimulus check) is higher under non-Gaussian earnings risk.
Sluggish news reactions: A combinatorial approach for synchronizing stock jumps
Stock prices often react sluggishly to news, producing gradual jumps and jump delays. Econometricians typically treat these sluggish reactions as microstructure effects and settle for a coarse sampling grid to guard against them. Synchronizing mistimed stock returns on a fine sampling grid allows us to better approximate the true common jumps in related stock prices.