Research

2026

1. R&R

   Aggregate Fluctuations from Firm Comovement

   Nahyeon Bak and Daisoon Kim

   R&R, Journal of Monetary Economics, 2026

   Abs PDF

   This paper shows that correlated idiosyncratic fluctuations within industries are an important source of macroeconomic volatility. Standard methods, such as cross-sectional demeaning, obscure this channel by mechanically driving average pairwise correlations toward zero asymptotically. We develop a non-parametric bounds approach that quantifies the contribution of such clustered comovement directly from firm-level data. Applied to U.S. firms, we find that within-industry comovement explains 10–15% of GDP volatility in normal times, 15–30% in downturns, and up to 45% during the Great Recession, helping account for the Great Moderation. These findings highlight the importance of cross-firm interdependencies in business cycle fluctuations.

2. working paper

   Brand Experience and Package Size in Dynamic Consumer Choice

   Nahyeon Bak, Daisoon Kim, and Chanheung Cho

   2026

   Abs PDF

   Consumers choose which brand and how much to buy. Using cough-drop market data, where infrequent purchases allow clean observation of consumption spells, we document that purchase size increases with within-spell brand experience and decreases following brand switching. We explain these patterns with a model of brand and quantity choice under learning about quality. As experience accumulates, declining uncertainty and selection into preferred brands increase purchase size, while brand switching renews uncertainty and induces smaller trial purchases. These patterns are consistent with learning rather than switching costs alone. Purchase quantities provide a novel behavior margin for understanding consumer behavior.

3. work in progress

   Contamination, Bias, and Correction in Text-Based Measurement

   Nahyeon Bak, Daisoon Kim, and Hanbat Jeong

   2026

   Abs

   Natural language processing, machine learning, and artificial intelligence are increasingly used to construct proxies for latent variables from unstructured data in empirical research (e.g., Economic Policy Uncertainty). We formalize a text-generation environment in which semantic-field activation and lexical co-selection cause word choices to co-vary with economic fundamentals, generating state-dependent measurement error in text-based proxies. Thus, using these variables as regressors generates both attenuation bias and contamination-induced omitted-variable bias. Exploiting residual nonlinear dependence, we propose two complementary correction methods: orthogonalized instrumental variables and copula-based control-function estimators that recover latent-variable effects using only the observed proxy and fundamentals. Finally, we provide diagnostic tools and practical guidance for empirical applications using text-derived variables. More broadly, our framework highlights that inference based on machine-generated variables depends not only on the extraction algorithm quality but also on the statistical properties of the underlying data-generation process.

2024

1. publication

   Home Market Effects and Increasing Returns with Non-constant Marginal Costs

   Nahyeon Bak, Daisoon Kim, and Mishita Mehra

   Journal of Economic Geography, 2024

   Abs DOI

   We reexamine the role of increasing returns in production, central to trade and economic geography theories, focusing on the home market effect. We extend the conventional multi-industry new trade model to introduce (1) nonconstant marginal costs and (2) nonhomothetic production in factors. If factors that are more (less) intensively used in fixed costs than variable costs also have higher relative prices in large countries compared to small countries, then large countries exhibit larger (smaller) firm sizes and specialize in industries with decreasing (increasing) marginal costs. Notably, different levels of fixed costs have a limited impact on these patterns.

2. working paper

   Industry Characteristics and the Home Market Effect: Evidence from Manufacturing Trade

   Nahyeon Bak, Daisoon Kim, and Mishita Mehra

   2024

   Abs PDF

   This paper examines how supply-side industry characteristics—particularly returns to scale from non-constant marginal costs—shape home market effects. Using data on narrowly defined industries, we find that industries with stronger returns to scale are more concentrated in larger countries only when non-production activities rely more on non-tradable inputs (labor) than production activities. In such cases, higher labor costs in larger countries deter firm entry, resulting in larger average firm sizes that better exploit scale economies. Consistent with prior research, we also find that industries with higher markups and trade costs tend to concentrate in larger countries.

2019

1. working paper

   Structural Estimation of a Dynamic Brand Choice Model

   Nahyeon Bak

   2019

   Abs

   This paper tests competing theories of consumer behavior using a structural demand model that incorporates variety-seeking. Existing studies struggle to jointly explain two key features of purchase behavior: persistent brand choice and frequent brand switching among variety-seeking consumers. To address this, I develop a dynamic panel demand model with discrete choices that allows for heterogeneity in switching behavior, capturing both persistent and frequent switchers within a unified framework. The estimates yield two main findings. First, consumers learn rapidly about product attributes, challenging standard slow-learning models such as Bayesian updating. Second, state dependence in brand choice declines with the time elapsed since the last purchase, indicating that habit persistence is short-lived. These results are robust to alternative specifications with and without explicit controls for variety-seeking behavior.