Research

• Abstract: We develop and test a theory of how investor-clientele segmentation shapes financial product design. In specialized ETF markets, institutional investors discipline providers through risk-adjusted performance and fee sensitivity, while retail investors extrapolate from salient past returns. The theory predicts, and the data confirm, that retail-oriented specialized ETFs have higher and more dispersed pre-listing betas, stronger pre-listing alphas, and higher fees than institutional-oriented specialized ETFs. The sharpest evidence supports the model's distinctive prediction: among retail-oriented specialized ETFs, pre-listing alpha and beta are substitutes after positive market returns but complements after negative market returns. Overall, the evidence documents clientele segmentation in financial product markets, linking investor composition to design, pricing, performance, and survival.

Abstract: Forecast error predictability regressions (Coibion and Gorodnichenko (2015)) test whether expectations overreact or underreact to information. We show that expressing level forecasts in growth-rate form, a common practice in the literature, introduces a mechanical bias, leaving current studies highly prone to misidentifying overreaction. This occurs because the revision embeds a lagged realized component positively correlated with subsequent errors. Using SPF and I/B/E/S data, negative coefficients on conventional growth-rate revisions weaken or reverse when using level-form revisions.  A rational noisy-information model replicates this bias. Thus, careful variable construction is essential to correctly identify expectation formation. 

• Abstract: Exchange-traded fund (ETF) providers display the top 10 holdings on a visually distinct, above-the-fold panel of the fund's information page. In specialized ETFs, where these stocks are likely to be read as representative of the fund's theme, we show that this disclosure convention generates a display salience effect: top 10 constituents attract heightened retail attention, experience larger increases in retail ownership, and earn cumulative abnormal returns 1.57% above other constituents around fund inception. The price gain, however, is temporary, with a significant portion of the initial gain reversing over the following months. The effect is concentrated specifically at the rank-10 visibility cutoff and is not driven by firm size or other stock characteristics. The results indicate that standard features of ETF portfolio disclosure can move retail attention and generate temporary deviations in constituent stock prices.

• Abstract: Firms accumulate intangible capital by building it internally or acquiring it externally. We develop a model in which this source choice is shaped by firm states and priced in stock markets. Tangible capital raises the value of source-specific fit and increases the cost of adapting externally developed intangibles to the firm’s operating architecture, while cash relaxes financing constraints on internal development. The model therefore predicts that firms with more tangible capital and more cash rely more on built rather than acquired intangible investment. It also predicts that built investment earns higher expected returns because internal development embeds an option to install: unfinished projects become productive only when continuation is sufficiently valuable. Evidence from U.S. firm data supports both predictions, suggesting that firms choose how to accumulate intangible capital, and markets price the risks embedded in that choice.

• Abstract: This paper examines whether stock-level demand is incorporated into the design of new exchange-traded fund (ETF) products. Using Korean equity ETFs from 2002 to 2024, we show that hot stocks, which exhibit strong prior returns and concentrated retail net buying in the pre-launch window, are more likely to be included in newly launched specialized ETFs. Using an ETF-level measure of hot-stock exposure, we find that ETFs with greater hot-stock exposure i) attract more retail and less institutional net buying after launch, ii) subsequently underperform other specialized ETFs by 0.6% to 1.3% per month, resulting in larger welfare losses for retail investors, and iii) generate higher fee revenue for providers. The evidence shows how stock-level retail demand is repackaged into ETF products in ways that are privately profitable for providers but costly for the retail investors they attract.

• Abstract: This paper revisits the idiosyncratic volatility puzzle (IVOL) by decomposing total idiosyncratic variance into two distinct components: a Beta-related part, driven by deviations of market beta from one, and a True part, reflecting the residual variance orthogonal to beta. Relying on total idiosyncratic variance alone conflates these components and obscures their distinct pricing implications. To address this issue, we implement a conditional double-sorting method that first ranks stocks by their Beta-related variance and then by their True variance. We find that stocks in the lowest quintiles of both components earn large abnormal returns, remain resilient during recessions, and consistently outperform the market with near-one betas. In contrast, stocks in the highest quintiles of both components experience the poorest performance. These findings demonstrate that decomposing idiosyncratic variance provides a sharper lens to distinguish between stocks with truly favorable risk–return profiles and those with truly unfavorable ones.

Park, Hayeon (2024). “Speculation Sentiment in Korea.” , Korean Journal of Financial Studies, 53(1), 49–102.