Peer-reviewed publications:
“The Polarization of Personal Saving” (with Marina Gindelsky). 2025. Review of Income and Wealth. https://doi.org/10.1111/roiw.70004
Abstract: The Bureau of Economic Analysis and Bureau of Labor Statistics have estimated the first complete distribution of personal saving for the United States by estimating the joint distribution of disposable personal income and personal consumption expenditures. We start with household survey data augmented with data from administrative sources and modify these data such that they aggregate to the national accounts totals for 2004-2022. The augmented household survey data are corrected for suspected underreporting at the top and bottom of the income and expenditure distributions to allocate macro totals to households. While aggregate saving is 3% of personal income in 2022, we find it is negative for the bottom half of the distribution. In fact, expenditures are more than double income for the bottom 10%, but almost six times less than income for the top 1%. Despite a temporary increase in saving during the COVID pandemic, the polarization is large and persistent, and robust to modifying the definitions and sample composition. This paper represents an important step in bridging the gap between micro households and national accounts for saving.
“Democratic Aggregation: Issues and Implications for Consumer Price Indexes.” 2024. Review of Income and Wealth, Vol. 71, No. 1. https://doi.org/10.1111/roiw.12703.
Abstract: This paper constructs and compares consumer price indexes (CPI) using weighting methods that differentially incorporate inflation disparities across households. Plutocratic CPIs, commonly used by statistical agencies, weight households based on their total expenditure, while democratic CPIs equally weight households to better represent average consumer experiences. I estimate democratic versions of the Bureau of Labor Statistics' CPI and Chained CPI (C-CPI) for all urban consumers using the Lowe and Törnqvist formulas, respectively. From December 2002 to June 2021, the democratic CPI-U exceeds its plutocratic counterpart by approximately 0.08 percentage points per year, on average, while the democratic C-CPI-U surpasses the plutocratic by 0.19 percentage points per year. The results indicate a negative correlation between inflation and household expenditure level over the study period. I also find weight frequency to be more important than index formula for explaining why larger differences occur for the C-CPI-U.
“A distributional approach to U.S. personal consumption expenditures: an overview" (with T. I. Garner, B. Matsumoto, and S. Curtin). 2024. Business Economics, Vol. 59, No. 3, pp. 166-173. https://doi.org/10.1057/s11369-024-00358-2.
Winner of the 2024 Abramson Award for Best Paper from the editors of Business Economics
Abstract: We distribute Personal Consumption Expenditures (PCE) across households in the U.S. using microdata from the Consumer Expenditure Surveys (CE) for the period 2017–2021. Since the CE mainly collects data on out-of-pocket spending, we supplement it with imputations based on other survey and administrative data to better match PCE definitions, particularly with respect to health care. Over the study period, out of the total PCE (excluding expenditures by non-profits serving households), the bottom 20% accounted for between 8.4% and 9.5%, while the top 20% accounted for 39.4–41.6%. The 90/10 ratio for equivalized PCE ranged from 3.3 to 3.7, and the Gini coefficient from 0.31 to 0.33.
“Revisiting Taste Change in Cost-of-Living Measurement.” 2022. Journal of Economic and Social Measurement, Vol. 46, No. 2, pp. 109-147. http://dx.doi.org/10.3233/JEM-220485.
Abstract: This paper compares conditional and unconditional cost-of-living indexes (COLI) when tastes change, focusing on the Constant Elasticity of Substitution model. A consumer price index typically targets a conditional COLI, which evaluates price change given set of preferences. An unconditional COLI aims to also capture the welfare effects of changing tastes, but it requires stronger assumptions. Using retail scanner data for food and beverage products, I find COLIs conditioning on current period tastes exceed those conditioning on prior period tastes. Consistent with previous studies, I find an unconditional COLI tends to reflect negative direct contributions from taste change.
“The robustness of conditional logit for binary panel data models with serial correlation” (with D. Kwak and J.M. Wooldridge). 2023. Journal of Econometric Methods, Vol. 12, No. 1, pp. 33-56. https://doi.org/10.1515/jem-2021-0005.
Abstract: We examine the conditional logit estimator for binary panel data models with unobserved heterogeneity. A key assumption used to derive the conditional logit estimator is conditional serial independence (CI), which is problematic when the underlying innovations are serially correlated. A Monte Carlo experiment suggests that the conditional logit estimator is not robust to violation of the CI assumption. We find that higher persistence and smaller time dimension both increase the magnitude of the bias in slope parameter estimates. We also compare conditional logit to unconditional logit, bias corrected unconditional logit, and pooled correlated random effects logit.
“The Geometric Young Formula for Elementary Aggregate Producer Price Indexes” (with A. Sadler, S. Stanley, W. Thompson, and J. Weinhagen). 2022. Journal of Official Statistics, Vol. 38, No. 1., pp. 239-253. http://dx.doi.org/10.2478/JOS-2022-0011.
Abstract: We re-estimate historical U.S. Producer Price Indexes (PPI) using the geometric Young formula at the elementary level. The geometric Young has better axiomatic properties than the modified Laspeyres, and may better approximate a feasible economic target. We find in most cases, indexes that use the geometric Young escalate between 0.1 and 0.3 percentage points less each year than those that use the modified Laspeyres. However, for wholesale and retail trade, as well as some other services, the differences are much larger. As a result, using the geometric Young at the elementary level lowers the U.S. PPI for Final Demand by 0.55 percentage points per year during the study period, a magnitude larger than what has been previously found for the U.S. Consumer Price Index.
“Estimation of average marginal effects in multiplicative unobserved effects panel models” 2017. Economics Letters 160, pp. 16-19. https://doi.org/10.1016/j.econlet.2017.08.020.
Abstract: In multiplicative unobserved effects panel models for nonnegative dependent variables, estimation of average marginal effects would seem problematic with a large cross section and few time periods due to the incidental parameters problem. While fixed effects Poisson consistently estimates the slope parameters of the conditional mean function, marginal effects generally depend on the unobserved heterogeneity. However, I show that a class of fixed effects averages is consistent and asymptotically normal with only the cross section growing. This implies researchers can estimate average treatment effects in levels as opposed to settling for average proportional effects.
Working papers:
“Another Look at the Linear Probability Model and Nonlinear Index Models” (with K. Chen and J.M. Wooldridge). 2024. https://arxiv.org/abs/2308.15338
Abstract: We reassess the use of linear models to approximate response probabilities of binary outcomes, focusing on average partial effects (APE). We confirm that linear projection parameters coincide with APEs in certain scenarios. Through simulations, we identify other cases where OLS does or does not approximate APEs and find that having large fraction of fitted values in [0, 1] is neither necessary nor sufficient. We also show nonlinear least squares estimation of the ramp model is consistent and asymptotically normal and is equivalent to using OLS on an iteratively trimmed sample to reduce bias. Our findings offer practical guidance for empirical research.
“Household Cost Indexes: Prototype Methods and Results” (with J. Klick, W. Johnson, and P. Liegey). 2023. BLS Working Paper 604. https://www.bls.gov/osmr/research-papers/2023/pdf/ec230040.pdf
Abstract: We estimate a family of price indexes known as Household Cost Indexes (HCI) using U.S. data. HCIs aim to measure the average inflation experiences of households as they purchase goods and services for consumption, and similar indexes are produced in the United Kingdom and New Zealand. These differ from the Bureau of Labor Statistics’ headline Consumer Price Index (CPI) products in two main respects. First, the upper-level aggregation of the HCIs weights households equally, unlike most headline CPIs which implicitly give more weight to higher-expenditure households. Second, the HCIs use the payments approach to value owner-occupied housing services explicitly using household outlays. In contrast, the U.S. CPIs use rental equivalence. The HCI for all urban consumers has an average 12-month change of 1.51% over December 2011 to December 2021, compared to 1.86% for the CPI-U. Roughly 95% of the difference is due to the payments approach.
“Distribution of U.S. Personal Consumption Expenditures for 2019: A Prototype Based on Consumer Expenditure Survey Data” (with S. Curtin, T. I. Garner, and B. Matsumoto). 2022. BLS Working Paper 557. https://www.bls.gov/osmr/research-papers/2022/pdf/ec220120.pdf
Abstract: We create a prototype procedure to allocate aggregate U.S. Personal Consumption Expenditures (PCE) across consumer units in the U.S. This allocation is based on data from the Consumer Expenditure Survey (CE). Since the CE collects data on out-of-pocket spending, to match PCE definitions, we assigned administrative and other household survey data for health expenditures to consumer units to reflect the full value of benefits received. Preliminary results for 2019 are that out of the total PCE (excluding expenditures by non-profits serving households), the bottom 10% accounted for 3.7%, while the top 10% accounted for 24.0%. The 90/10 ratio for equivalized PCE is 3.4, and the Gini coefficient is 0.28.
“Changing tastes vs. specification error in cost-of-living measurement.” 2020. BLS Working Paper 531. https://www.bls.gov/osmr/research-papers/2020/pdf/ec200130.pdf.
Abstract: Several recent papers aim to account for changing preferences in cost-of-living indexes (COLI). Workhorse models like Constant Elasticity of Substitution (CES) attribute the errors in demand regressions entirely to preferences, leaving no room for other sources of error. Using a Monte Carlo experiment and retail scanner data, I find evidence that model misspecification can lead to misleading conclusions about the degree of taste change reflected in CES-based price indexes. Nevertheless, under misspecification, a Sato-Vartia index still approximates a conditional COLI that fixes tastes to an intermediate level.
“Exponential panel models with coefficient heterogeneity.” 2018. BLS Working Paper 503. https://www.bls.gov/osmr/research-papers/2018/pdf/ec180030.pdf
Abstract: If heterogeneous slopes are ignored in exponential panel models, fixed effects Poisson may not estimate any quantity of interest. Existing estimation methods often involve treating only a small subset of the slopes as “random effects” and integrating from the likelihood, increasing computational difficulty. I propose a test to detect slope heterogeneity that, unlike the tradi- tional approach, does not amount to testing for information matrix equality. Additionally, I present a correlated random coefficients approach to identification which allows for estimation of the coefficient means and average partial effects. I test these proposed methods using a Monte Carlo experiment and apply them to the patent-R&D relationship for U.S. manufacturing firms.