Accessibility menu
Marketing and Finance Seminars
2022-2023
2022
Hidden in Plain Sight: Consumer Responses to Pseudo-Secrets in Marketing
The present research introduces and conceptualizes the paradoxical phenomenon of “pseudo-secrets” in marketing and examines its appeal and impact on real consumer behavior in the marketplace. Restaurants ranging from gourmet Michelin-starred to mainstream fast-food chains offer secret menu items, and hidden stores and “speakeasy” bars feature camouflaged entrances and secret passcodes. Paradoxically, many of these hidden places and products are famous for being a secret. We argue that pseudo-secrets often hold important symbolic value: they make consumers feel socially central. Accordingly, we demonstrate that pseudo-secrets increase word-of-mouth about the brand, and this effect is mediated by consumers’ feelings of social centrality – the subjective experience of feeling connected and focal in one’s network, and attenuated when the symbolic value of the secret is low. We further demonstrate that pseudo-secrets can even create WOM about unexciting items and experiences, reaching the level of WOM found for their iconic and desirable counterparts. Our multi-method approach, combining field experiments, company proprietary data, and lab studies, further demonstrates how marketers can effectively apply these insights and design pseudo-secrets in various product categories and consumption contexts.
How Do Brand Networks Break In Face Of A Crisis?
Brand communities have an unparalleled power to integrate customer value with brand growth. Customers rely on brand communities tointeract with each other, to connect with the brands they love, to solve problems, and to personalize their consumption experiences. However, customers also resort to these communities to coordinate a negative collective crisis response. An uncontrolled reaction of online brand communities to brand crises can deteriorate brands' value and market performance, and push loyal and engaged consumers away from the brand social network. In this project, we assess the effect of brand crises on customer participation in online brand communities, and on ease and speed of information spread in the brand networks. We use data from 300 online brand communities, and exploit the quasi-experimental exposure of community members to over 7000 brand crisis episodes reported by news providers between 2010 and 2019. In a series of difference-in-difference analyses, we find that brand crises (i) increase the weekly participation of consumers in brand communities, (ii) affect the patterns of information-sharing in the brand networks, and (iii) have a differential impact on different consumer types. The "brand loyal" consumers effectively disengage from their brand communities following the crisis event -- therefore, the average boost in brand-related activity is attributable to "brand strangers", people who only activate after a crisis. However, we show that the decrease in engagement is mitigated among the active consumers who had proportionally more experience, loyalty, or status within the brand community. Accordingly, we suggest that brand crises are a serious threat to the integrity of online brand communities, but that consumer loyalty and commitment has the potential to preserve the functioning of brand spaces online in case of serious reputation threats. The insights from this paper support businesses and organizations managing online communities in situations of external stress and unexpected reputational threats.
How Watching Live Streams Creates Connection and Enhances Enjoyment
Peer-to-peer live streaming is a rapidly growing phenomenon: in 2020, users spent over 27 billion hours viewing live streams online, nearly doubling the hours spent in 2019. We examine the viewing experiences of over 2,700 consumers in both naturalistic and carefully controlled environments and find consistent positive effects of viewing online live streams (versus identical pre-recorded videos) on feelings of social connection. We find evidence that liveness itself enhances feelings of connection to the broadcaster, and that this benefit of viewing live streams is driven by an elevated sense of presence, or "being there," in events that are viewed in real-time. We also find that increased salience of other viewers enhances feelings of connection to others who are watching the same events at the same time. The social connection afforded by viewing live streams enhances consumers' enjoyment and increases their propensity to continue watching similar content. In a world where people increasingly turn to technology to satisfy their social needs, live streams present a novel opportunity for consumers to feel connected and for marketers, platform developers, and media personalities to enhance the experiences of their viewers.
Reviews have a powerful influence on product success and, accordingly, have received ample research attention. However, so far, interest in reviews has effectively stopped at checkout. This research shows that exposure to reviews—and, particularly, negative reviews—matters beyond the point of purchase, and can shape the consumption experience itself.
First, we demonstrate, for the first time, that pre-consumption exposure to reviews leads consumers to experience the reviewed products differently from naïve consumers (who experience products without reading reviews). Reviews affect the consumption experience, and not merely reports of the experience, and do so regardless of product quality and whether consumers choose the product or have it chosen for them. Second, we show that the effect is asymmetric, such that negative reviews produce more negative consumption experiences than naïve consumption, whereas positive reviews do not significantly affect experience. Third, we identify negativity bias as a mechanism for this asymmetry: We show that pre-consumption exposure to negative reviews (vs. positive reviews or naïve consumption) elicits a stronger focus on negative attributes of the product experience and, consequently, make it worse. Fourth, we show that negative consumer information has a more pronounced (negative) impact on other consumers' consumption experience compared to negative information generated by a marketer.
Overall, this research advances the review literature by identifying and characterizing (potentially detrimental) effects of reviews after the point of purchase, something they were neither aimed nor expected to affect. Our work generates novel insights that are immediately applicable to the practice of marketing.
2020-2021
The Temporal Slippery Slope: Decline in Sequential Ratings within Batches of Online Reviews
We show that reviewers (on platforms like IMDB and Goodreads) tend to provide ratings in "batches," or several ratings for different items in a short time period. Our analysis of more than 16M ratings and more than 1M reviewers shows that the ubiquity of ratings given in "batches" ranges between 12.41% to 92.88% for various platforms. We further demonstrate that ratings within a batch have unique and consistent distribution. Findings from two large-scale databases—incorporating data from more than 100M ratings and 500K reviewers in a time resolution of seconds—reveal that ratings (by the same reviewer) within these "batches" drop as a function of time. In four experimental studies we replicate the decrease in sequential ratings and shed light on the role of doubt in explaining this drop. We show that reviewers have stronger doubt about later ratings. Such doubt is reflected not only via the decrease in rating score over time, but also via the length of time that passes between sequential ratings. The drop is attenuated when the reviewer has low (compared to high) levels of doubt about later ratings. Our research contributes to the literature on sequential decisions, the batch (or “burst”) phenomenon, and doubt in evaluation and online reviews. We also offer useful practical implications for platforms seeking to create a more reliable ranking scale.
Good Engagement/Bad Engagement – The Positive and Negative Effects of Online Engagement on Consumption
In recent years, billions of dollars are spent, by both online and offline retailers, on website design aimed at increasing consumers’ online engagement. Yet, it is not clear that increasing consumers’ online engagement is the optimal strategy to increase sales. In this talk I will present three studies that examine the potential negative and positive effects of digital engagement. In the first paper, we study the relationship between online engagement and offline sales, utilizing a quasi-experimental setting whereby a leading premium automobile brand launched a new interactive website gradually across markets, allowing for a treatment-control comparison. The paper finds surprising evidence suggesting that increased online engagement reduces offline car sales. This negative effect is due to substitution between online and offline engagement, as the high-engagement website decreased users’ tendency to submit online requests that lead to personal contact with a car dealer. The second study focuses on video as an engagement feature. Using a field experiment of 53 different apps, we find that for 85% of the apps, video inclusion has either a negative or no significant effect on downloads. Moreover, the results show that viewing is negatively associated with installs, while viewing to completion is positively associated with installs, and that video has negative effects on other engagement activities. We further examine the effect of using video in an online lab experiment and find heterogenous effects by user interest and gender. In the third study, two lab experiments were conducted to isolate and measure the effects of three engagement features (gallery scroll, read more description, and read reviews). Results suggest that app description has a negative effect on downloads for users who selected an app category which differed from their favorite one, while no causal effect was identified for gallery scroll and reviews.
How Work Experiences Shape Entrepreneurial Career Ambitions
Research suggests that employees in startup firms are more likely to become founders of entrepreneurial ventures than employees in large corporations. However, a particular mechanism has not yet been identified. In the current research, we are drawing on role categorization and self-efficacy theories to explain this phenomenon. Specifically, we propose that employees who work in startup firms are more likely to categorize the roles they enact and the skills they acquire as pertaining to entrepreneurship. As a result, they become more confident in their abilities to succeed as entrepreneurs and more likely to consider an entrepreneurial career path. To examine our prediction, we adopt an experimental approach and randomly assigning participants to complete identical work categorized either as “startup” related or “large corporation” related. In two experiments, we found that subjects who performed work for a “startup” were more likely to indicate intentions to start a business, and also to take more entrepreneurship classes, than participants who performed the exact same tasks for a “large corporation”. In addition, entrepreneurial self-efficacy mediated the relationship between work categorization and entrepreneurial intentions. In my presentation I will further discuss on how these findings could help generate a set of insights for entrepreneurship programs and intrapreneurship.
How We Make Sense of People, Brands, and Other Agents
Humans navigate a world of agents: people first, but also animals, societal groups, and corporations, as well as more and more AI. This complex landscape of autonomous entities yields to a simple but powerful human judgment of others’ intent for good or ill, plus their ability to enact it. Agents mapped in a warmth x competence space explain phenomena from stereotyping under increasing diversity to economic decision-making under uncertainty. Evidence from surveys and experiments over time and place—including some adversarial collaborations—suggest apparently universal dimensions that allow humans to respond to a rapidly changing world.
Combating Fake News: A Consumer Psychology Perspective
An increasing proliferation of misinformation and “fake news” has been widely reported and documented. Reality is now under attack from advertising-optimized information architectures mediating our contemporary reality. Disinformation campaigns proliferate on online forums and social media. My research program aims to answer questions regarding why we believe and share fake news and how to prevent or correct inaccurate beliefs. In a paper titled “Perceived Social Presence Reduces Fact-Checking” (Proceedings of the National Academy of Science 2017), we find that consumers are less likely to fact check ambiguous news headlines when they feel they are in the presence of others compared to when they are alone. We find that this reluctance to fact check is caused by reduced vigilance in group settings such as social media. A second paper examines how to involve consumers in fact-checking news articles. We propose a method to leverage the input of general consumers (crowdsourcing), algorithms (supervised learning), and experts (third-party fact-checkers) to rate the veracity of scientific articles on news media. We propose and test the use of similarity judgments to facilitate unbiased consumer responses. In a third research project, I address the issue of fake news from a different perspective and examine sharers of fake news. Who are they, and what motivates them to share fake news? We contrast fake news sharers, fact-check sharers, sharers of news articles from general media outlets and a random sample of social media users across five dimensions-demographics, political ideology, social media usage, emotions and personality. We access these characteristics by collecting their personal information as posted on Twitter as well as the content of their tweets. Fake news sharers differ from the other groups on multiple characteristics, but they also show similarities to fact check sharers on their emotional profile. Our findings can help social platforms to screen, prioritize and scrutinize messages posted by potential fake news sharers before false messages are widely disseminated. A fourth project in this research stream titled “Social Marginalization Motivates Indiscriminate Sharing of COVID-19 News on Social Media” (Journal of the Association of Consumer Research 2020), finds that people who feel socially marginalized are more likely to share COVID-19 news indiscriminately. They are likely to share news that is factually untrue and true, as well as news that seems surprising and unsurprising. This effect is driven by a general motivation to seek meaning. Helping people obtain a temporary sense of meaning by endowing them with a feeling of power can reduce indiscriminate news sharing. Taken together, this research program aims to guide policy discussions on how to combat the spread of fake news.
When The Data Are Out: Measuring Behavioral Changes Following a Data Breach
As the quantity and value of data increase, so do the severity of data breaches and customer privacy invasions. While firms typically publicize their post-breach protective actions, little is known about the social, behavioral, and economic aftereffects of major breaches. Specifically, do individual customers alter their interactions with the firm, or do they continue with “business as usual”? We address this general issue via data stemming from a matchmaking website, one for those seeking an extramarital affair, that was breached. The data include de-identified profiles of paying male users from the United States, and their activities on the website since joining, and up to 3 weeks after, the disclosure of the data breach. A challenge in making causal inference(s) in the setting of a massive and highly publicized data breach is that all users were informed of the breach at the same time. In such cases of “information shock”, there is no obvious control group. To resolve this problem, we propose Temporal Causal Inference: for each group of users who joined in a specific time period, we create an appropriate control group from all users who had joined prior to it. This procedure helps control for, among other elements, potential trends in both individual and temporal site usage that broadly fall under the rubric of “normal” usage trajectories. Following construction of suitable control groups, we apply and extend several causal inference approaches. In particular, we adapt Athey, Tibshirani and Wager’s (2019) Causal Forests (among other forest-based methods) into Temporal Causal Forests, to better align ‘temporal’ inference settings. The combination of Temporal Causal Inference and Temporal Causal Forests methods allows us to extract insights regarding the homogenous (average) treatment effect, along with nontrivial heterogeneity in responses to the data breach. Our analyses reveal that there is a decrease in the probability of being active in searching or messaging on the website, and a notable increase in the probability of deleting photos, ostensibly to avoid personal identification. We investigate several potential sources of heterogeneity in response to the breach announcement, and conclude with a discussion of both managerial consequences and policy considerations.
2019-2020
Ineffective Altruism - Marketing seminar
Despite well-meaning intentions, people rarely allocate their charitable donations in the most cost-effective way possible. Whereas most early research on ineffective altruism focused on donation decisions in the absence of effectiveness metrics, the growth of the effective altruism movement has led to much improved information and transparency intended to help people make more effective donation decisions. These changes to the decision making environment yield new research questions about how people decide how much and to whom they should donate. In this talk, I plan to cover two papers: In the first paper, we demonstrate that a common form of effectiveness information provision—the cost per unit (e.g., cost of a malaria net, cost to save of life) has a perverse effect on how much people give. Specifically, people give less when the cost is cheaper. This result arises because people want their donation to have a tangible impact, and when the cost of such an impact is lower, people can achieve it with a smaller donation. In the second paper, we demonstrate that when choosing between charities to support, few choose the charity with the highest return, even when it is obvious which option is “best”. The key reason why is people construe of this decision as a matter of subjective preferences, and thus are willing to do less good overall if they can support the cause they prefer. I will end the talk by discussing how effectiveness information provision is not a cure-all for improving donation decisions, but can be improved upon with better framing.
A Potato Salad with a Lemon Twist: Using a Supply-Side Shock to Study the Impact of Opportunistic Behavior on Crowdfunding Platforms Marketing seminar
Crowdfunding platforms are peer-to-peer two-sided markets that enable amateur entrepreneurs to raise money online for their ventures. However, in allowing practically anyone to enter, such platforms enable opportunistic suppliers to flood the market with offerings, many of which are of low quality. This situation creates choice overload for potential backers and may thus influence their investment decisions. To empirically study the implications of this phenomenon for crowdfunding performance, we use a quasi-natural experiment in the form of an exogenous media shock that occurred on Kickstarter.com. The shock was followed by a sharp increase in the number of campaigns, particularly low-quality ones, offered on the supply side of the market; no such increase was observed on the demand side of the market. These unique conditions enable us to estimate how crowdfunding platforms are affected by the presence of an atypically large number of low-quality campaigns, while controlling for fluctuations in demand. We use two identification strategies, which enable us to control for changes in quality, to show that an increase in low-quality supply significantly decreases the performance of the average crowdfunding campaign, manifested in a lower likelihood of success (reaching funding goals) and less money raised per campaign. We also offer a new measure to estimate campaign quality and study the moderating role of campaign quality in the observed effects. We find that high-quality campaigns are less affected than low-quality campaigns by the influx of low-quality offerings. In the talk, I would also discuss theoretical implications as well as managerial implications for entrepreneurs and platform designers. This talk is based on an MISQ forthcoming paper, co-authored with Hilah Geva and Gal Oestreicher-Singer.
Algorithmic Bias? Marketing seminar
We explore data from a field test of how an algorithm delivered ads promoting job opportunities in the science, technology, engineering and math fields. This ad was explicitly intended to be gender neutral in its delivery. Empirically, however, fewer women saw the ad than men. This happened because younger women are a prized demographic and are more expensive to show ads to. An algorithm that simply optimizes costeffectiveness in ad delivery will deliver ads that were intended to be gender neutral in an apparently discriminatory way, because of crowding out. We show that this empirical regularity extends to other major digital platforms.
2018-2019
2017-2018
National Culture and the Value Implications of Corporate Social Responsibility
We examine why corporate social responsibility (CSR) practices vary across countries and firms, and the value implications of such variation. Using Thomson Reuters’ ASSET4 database on the CSR practices of 4,279 firms from 49 countries over 2003–2015 and employing a hierarchical linear model, we find that the national cultural dimension of individualism is positively associated with firm-level CSR practices. We further find that income inequality at the country level and board diversity and corporate transparency at the firm level link individualism to CSR practices. Moreover, both between and within countries, we find a positive association between firm-level CSR practices and firm value, with two firm-level channels—cost of equity and bankruptcy probability—linking CSR practices to firm value. Finally, we find that the positive association between firm-level CSR practices and firm value is stronger in more individualistic countries.
Impulsive Consumption and Financial Wellbeing: Evidence from an Increase in the Availability of Alcohol
Increased availability of alcohol might harm individuals if they have time-inconsistent preferences and consume more than planned before. We study this idea by examining the credit behavior of low-income households around the expansion of the opening hours of retail liquor stores during a nationwide experiment in Sweden. Consistent with store closures serve as commitment devices, expanded operating hours led to higher alcohol consumption and greater consumer credit demand, default, and negative consequences in the labor market. Our calculation shows that the effects of alcohol consumption on indebtedness could amount to 3.2 times the expenditure on alcohol.
Financing Durable Assets
This paper studies how the durability of assets affects financing. We show that more durable assets require larger down payments making them harder to finance, because durability affects the price of assets and hence the overall financing need more than their collateral value. Durability affects technology adoption, the choice between new and used capital, and the rent versus buy decision. Constrained firms invest in less durable, otherwise dominated assets and buy used assets. More durable assets are more likely to be rented. Economies with weak legal enforcement invest more in less durable assets and are net importers of used assets.
Money Management in Equilibrium
The money management firm is an ideal place to study both asset pricing and corporate finance questions. There are few corporations that allow the level of transparency that a mutual fund firm offers – these firms offer a window into human decision making that is hard to match elsewhere. Consequently, the area offers a unique and rich opportunity for new and innovative research. Historically, researchers have ignored these opportunities because the literature on money management has inconsistently applied the rational expectations equilibrium concept. When applied consistently, the rational expectations equilibrium approximates the observed equilibrium in the money management space at least as well as it does in the stock market. Just as the application of the rational expectations equilibrium transformed the stock pricing literature, the consistent application of this paradigm to the money management literature has the potential to transform the money management literature.
Transparency and Talent Allocation in Money Management
We construct and analyze the equilibrium of a model of delegated portfolio management in which money managers signal their investment skills via fund transparency. To lower the costs of transparency, high-skill managers rely on their performance to separate from low-skill managers over time. In contrast, medium-skill managers rely on transparency to separate, especially when it is difficult for investors to tell them apart through performance alone. Low-skill managers mimic high-skill managers in opaque funds, hoping to replicate their performance and compensation. The model yields several novel empirical predictions that contrast transparent and opaque funds.
Why Do Boards Exist? Governance Design in the Absence of Corporate Law
We study how owners trade off the costs and benefits of establishing a board in a historical setting, where boards are optional and their role can be identified from cross-sectional differences in authority allocation across the general meeting, the board, and management. We find that boards arise when numerous small shareholders own a sizeable fraction of equity, thereby aggravating collective action problems. Boards monitor but also mediate among large and small shareholders as voting caps limit blockholders' influence and help to ensure that small owners' interests are represented on the board. In some _rms, boards arise mainly to advise.
Complementarity between Audited Financial Reporting and Voluntary Disclosure: The Case of Former Andersen Clients
Analytical research suggests mandatory periodic reporting disciplines disclosure and encourages timely voluntary disclosure. We examine this hypothesis using the shock to financial reporting quality experienced by former Arthur Andersen clients after they were forced to switch auditors. Consistent with the confirmatory role of mandatory reporting, we find that former Andersen clients increase disclosure following the switch. They increase forecasting frequency and enhance forecasting precision and specificity. They also show less return concentration around earnings announcements in bad-news quarters, consistent with timelier release of bad news (Roychowdhury and Sletten 2012). We supplement our main findings with a battery of tests to rule out the role of alternative shocks in our results. Our findings demonstrate complementarity between financial reporting quality and voluntary disclosures.
2022-2023
We study dealers’ liquidity provision in the currency market. We show that at times when dealers’ intermediation capacity is constrained their cost of liquidity provision increases disproportionately relative to dealer-provided volume. As a result, the elasticity of dealers’ liquidity provision weakens by at least 80% relative to periods when they are unconstrained. We identify constrained periods based on leverage ratios, Value-at-Risk measures, credit default spreads, and debt funding costs.
We interpret our novel empirical findings within a parsimonious model that sheds light on the key mechanisms of how liquidity provision by dealers tends to weaken when intermediary constraints are tightening.DYNAMIC DISCLOSURE GAMES
We study strategic disclosure timing by correlated firms in the presence of risk-averse investors. Risk premia rise before disclosures, drop when disclosures occur, and then begin rising again. Disclosures are always good news, but disclosures that are only moderately good news induce clustering of disclosures by other positively correlated firms, because a disclosure by any firm reduces the values to others of keeping their disclosure options alive. We present empirical evidence that firms
strategically time disclosures as predicted by the model.The Financial Transmission of a Climate Shock: El Nino and US Banks
This paper studies how a climate shock is transmitted through the financial system. Our empirical strategy combines data on climate and banking with El Nino, a natural experiment producing quasi-random variation in US climate. El Nino generates heterogeneous changes in lending across counties, which aggregate at the bank level. We quantify the effects of El Nino on the demand and supply of credit and implement a LASSO analysis to identify the characteristics of banks resilient to this shock. Our findings show that supply factors induce the lending reduction and banks with lower physical capital are more resilient to El Nino.
Fee the People: Retail Investor Behavior and Trading Commission Fees
We show retail investors are highly responsive to changes in trading commission fees. Using a triple-difference research design around the removal of fees for retail investors on the international retail broker platform, eToro, we show investors responded by trading approximately 30% more frequently, in smaller order sizes, and increasing portfolio turnover. Removing fees also spurred retail investors to reallocate their portfolios and diversify. Retail investors’ gross return performance did not significantly change around the fee removal despite trading more often, but retail investors earned significantly higher returns on a net basis after accounting for fees incurred in the pre-period. Finally, using demographic information, we show removing fees disproportionately affected inexperienced investors with lower deposit amounts and lesser technological sophistication both by expanding the extensive margin of investors and changing trading activity for the intensive margin of investors. Together, our results suggest commission fees play an influential role as a speed bump for retail investor participation, trading activity, and diversification.
Measuring Time-Varying Disaster Risk: An Empirical Analysis of Dark Matter in Asset Prices
To confront the challenge that disaster risk is “dark matter” in finance, we construct an objective measure of disaster risk, which is able to predict half of GDP crashes in a sample of 20 advanced economies between 1870 and 2021. Despite this significant predictability, we find no supportive, and often contradictory, evidence of higher predicted disaster risk being associated with a higher equity premium, volatility, or dividend/price ratio of the equity market index; higher corporate bond spreads, or higher term spreads. Our results suggest that the subjective disaster risk mirrored by asset prices lags objective disaster risk by two years.
2022
Common Fund Flows: Flow Hedging and Factor Pricing
Active equity funds care about fund size, swayed by fund flows responding to primitive economic fluctuations. Funds hedge against common-flow shocks by tilting their portfolios toward low-flow-beta stocks, while retail investors tilt theirs toward the opposite. In equilibrium, common-flow shocks earn a risk premium, leading to a multi-factor asset-pricing model like the ICAPM, even with myopic agents using naive asset-pricing models. Empirically, fund flows obey a strong factor structure with the common component earning a risk premium, and funds actively hedge against fund flow shocks —more aggressively so when flow-hedging motives rocket following natural disasters and unexpected trade-war announcements.
Asset Pricing with Panel Trees under Global Split Criteria*
We introduce a class of interpretable tree-based models (P-Trees) for analyzing (unbalanced) panel data, with iterative and global (instead of recursive and local) split criteria. We apply Ptree to split the cross section of asset returns under no arbitrage, generating a stochastic discount factor model and effective test portfolios for asset pricing. P-Trees capture nonlinear feature interactions, accommodate time-series splits, and allows interactions between macroeconomic states and asset characteristics. In an empirical study of U.S. equities, data-driven cutpoints in P-Trees reveal that long-run reversal, volume volatility, and industry-adjusted market equity interact to drive cross-sectional return variations, and that inflation constitutes the most critical regime-switching. P-Trees consistently outperform known observable and latent factor models for pricing individual asset and portfolio returns, while delivering profitable and transparent trading strategies utilizing characteristic interactions. Notably, factor portfolios from P-Trees generate a monthly risk-adjusted alpha of 2.13% and an annualized Sharpe ratio of 1.71. The methodology is broadly applicable for building trees with multi-period leaves and economic restrictions as split criteria to guard against overfitting and improve model performance.
Key Words: CART, Cross-Sectional Returns, Interpretable AI, Latent Factor, Machine Learning.
We present a novel finding that high macroeconomic uncertainty is associated with greater accumulation of physical capital, despite a contemporaneous
reduction in investment. To reconcile this evidence, we show that high un-
certainty predicts a persistent decrease in the utilization and depreciation of
existing capital, which dominates the investment slowdown. We construct and
estimate a general-equilibrium model to explain our novel findings alongside the
existing evidence on the relationship between uncertainty, economic growth,
and asset prices. In the model, precautionary saving is achieved by lowering utilization, instead of increasing investment. Lower utilization persistently
decreases depreciation, conserving capital for the future, and simultaneously
discourages new investment. This channel amplifies stock price exposure to
uncertainty risks, especially for rms with more flexible utilization, which we
confirm in the data. We further show the importance of our mechanism to generate a negative impact of uncertainty shocks in an extended New-Keynesian
framework
Governments around the world have gone on a massive fiscal expansion in response to the GFC and Covid crises, increasing government debt to levels not seen in 75 years. How will this debt be repaid? What role do conventional and unconventional monetary policy play? We investigate debt sustainability in a New Keynesian model with an intermediary sector, realistic fiscal and monetary policy, endogenous convenience yields, and substantial risk premia. During a large economic crisis, increased government spending and lower tax revenue lead to a large rise in government debt and raise the risk of future tax increases. Quantitative easing (QE) contributes to lowering the debt/GDP ratio and reducing the risk of future tax increases. QE is state- and duration-dependent: while a temporary QE policy deployed in a crisis stimulates aggregate demand, permanent QE crowds out investment and lowers long-run output.
We study the performance of collateralized loan obligations (CLOs) to understand the market imperfections giving rise to these vehicles and their corresponding economic costs. CLO equity tranches earn positive abnormal returns from the risk-adjusted price differential between leveraged loans and CLO debt tranches. Debt tranches offer higher returns than similarly rated corporate bonds, making them attractive to banks and insurers that face risk-based capital requirements. Temporal variation in equity performance highlights the resilience of CLOs to market volatility due to their closed-end structure, long-term funding, and embedded options to reinvest principal proceeds.
Deviations from a policy rule underpin empirical identification of monetary policy shocks. We cast light on how deviations arise by analyzing internal policy deliberations of the Federal Open Market Committee (FOMC). We show that policymakers’ beliefs about higher-order moments of economic distributions— specifically perceptions of uncertainty and skewness—significantly impact policy stance beyond economic forecasts typically used in rule estimates. To capture those otherwise unobservable decision-making features, we construct text-based proxies for policymakers’ uncertainty, sentiment, and policy stance from the FOMC meeting transcripts over the 1987–2015 period. Heightened uncertainty generally amplifies the policymakers’ response to the macroeconomy. However, while an increased uncertainty about the real economy drives an easier stance, inflation uncertainty leads to more hawkishness. We show that policymakers’ inflation uncertainty is associated with their skewed beliefs about rising inflation, which do not materialize in our sample. The results depart from the certainty equivalence arising in classic monetary models and contrast with the frequently-referenced conservatism in policymaking under uncertainty. Instead, the evidence suggests that policymakers act aggressively to avoid low-probability costly outcomes which are endogenous to their policy actions.
We study how firm characteristics are correlated with stock price levels by measuring the longterm discount rates (defined as the internal rate of return) of anomaly portfolios over a long horizon. Utilizing a simple novel non-parametric estimation methodology, which proxies ex-ante equity payout expectations with ex-post realizations, we reveal that the patterns of long-term discount rates are out-of-line with the average short-term holding period returns for multiple prominent anomalies. The set of stylized facts uncovered correspondingly shed new light on the mechanisms underlying various asset-pricing anomalies. Moreover, they indicate that long-term discount rates better characterize firms’ equity financing cost than short-term expected returns; with a representative example, we demonstrate how structural models that posit a tight connection between the two could imply counterfactual patterns in price levels.
2020-2021
Pricing Without Mispricing
We investigate whether various asset pricing models could hold in an efficient market. Assuming decade-old information should be priced correctly, we test whether a model assigns zero alpha to investment strategies that use only such information. The CAPM passes this test, but prominent multifactor models do not. Multifactor betas may help capture expected returns on mispriced stocks, but persistence in those betas distorts the stocks’ implied expected returns after prices correct. Such effects are strongest in large-cap stocks, whose multifactor betas are the most persistent. Hence, prominent multifactor models distort expected returns, absent mispricing, for the largest, most liquid stocks.
Do Common Factors Really Explain the Cross- Section of Stock Returns?
The empirical ability of stock characteristics to predict excess returns challenges the notion of a trade-off between systematic risk and expected return. We measure individual stocks’ exposures to all latent common factors using a novel high-dimensional method. These latent factors appear to earn negligible risk premia despite explaining essentially all of the common time-series variation in stock returns. We use machine learning methods to construct out- of-sample forecasts of stock returns based on a wide range of characteristics. A zero-cost beta-neutral portfolio that exploits this predictability but hedges all undiversifiable risk delivers a Sharpe ratio above one with no correlation with any systematic factor, thus rejecting the key prediction of the arbitrage pricing theory.
Private renegotiations and government interventions in debt chains
We propose a model of strategic debt renegotiation in which businesses are sequentially interconnected through their liabilities. This financing structure, which we refer to as a debt chain, gives rise to externalities, as a lender’s willingness to provide concessions to its privately-informed borrower depends on how the lender’s own liabilities are expected to be renegotiated. We highlight how government interventions that aim to prevent default waves should account for these private renegotiation incentives and their interlinkages. In particular, we contrast the consequences of targeted subsidies vs. debt reduction programs following economic shocks such as a pandemic or financial crisis.
Structural Deep Learning in Conditional Asset Pricing
We develop new nonparametric methodology for estimating conditional as- set pricing models using deep neural networks, by employing time-varying conditional information on alphas and betas carried by firm-specific characteristics. The method first applies cross-sectional deep learning, period-by-period, to estimate spontaneous conditional expected returns, defined as the conditional expectation of asset returns given characteristics and factor realizations. We additionally estimate the long-term expected return as the predicted mispricing component and the product of the estimated risk exposures times the price of risk, where local kernel smoothing is applied to capture the return dynamics that arise from time-varying alphas and betas. Contrary to many applications of neural networks in economics, we can open the “black box” and provide an economic interpretation of the successful predictions obtained from neural networks, by decomposing the neural predictors a risk based and mispricing component. We formally establish the asymptotic theory of the deep-learning estimators, which apply to both in-sample fit and out-of-sample predictions. Empirically, we find a large, time varying mispricing component, and that the mispricing component is slowly decaying over time, but not monotonically. Mis- pricing tends to be high during times of high market volatility which is linked to periods of economic turmoil. Finally, we also illustrate the “double-descent- risk” phenomena associated with over- parametrized predictions, which justifies the use of over-fitting machine learning methods.
The Changing Economics of Knowledge Production
Big data technologies change the way in which data and human labor combine to create knowledge. Is this a modest technological advance or a data revolution? Using hiring and wage data from the investment management sector, we estimate firms' data stocks and the shape of their knowledge production functions. Knowing how much production functions have changed informs us about the likely long-run changes in output, in factor shares, and in the distribution of income, due to the new, big data technologies. Using data from the investment management industry, our results suggest that the labor share of income in knowledge work may fall from 29% to 21%. The change associated with big data technologies is two-thirds of the magnitude of the change brought on by the industrial revolution.
MEASURING THE WELFARE EFFECTS OF ADVERSE SELECTION IN CONSUMER CREDIT MARKETS
Adverse selection is known in theory to lead to inefficiently low credit provision, yet empirical estimates of the resulting welfare losses are scarce. This paper leverages a randomized experiment conducted by a large fintech lender to estimate welfare losses arising from selection in the market for online consumer credit. Building on methods from the insurance literature, we show how exogenous variation in interest rates can be used to estimate borrower demand and lender cost curves and recover implied welfare losses. While adverse selection leads to large equilibrium price distortions, we find only small overall welfare losses, particularly for high-credit-score borrowers.
Interbank Credit Exposures and Financial Stability
This paper investigates how interbank credit exposures affect financial stability. Policy makers often see such exposures as undermining stability by exacerbating cascading losses through the financial system. I develop a model that features a trade-off between cascading losses and risk-sharing. In contrast to previous studies I find that reducing interbank connectivity may destabilize the financial system via the bank-run channel. This is because it decreases the risk-sharing benefits of interbank connectivity. A bank-run model features two islands that are connected via a long term debt claim. Varying the size of this claim (interbank connectivity), I study how the decision to `run on the bank' is affected. I run a simulation of the model, calibrated to the U.S. banking system between 1997-2007. I find that large bankruptcy costs are required to trump the risk-sharing benefits of interbank credit exposures
Predicting Returns with Text Data
We introduce a new text-mining methodology that extracts information from news articles to predict asset returns. Unlike more common sentiment scores used for stock return prediction (e.g., those sold by commercial vendors or built with dictionary-based methods), our supervised learning framework constructs a score that is specifically adapted to the problem of return prediction. Our method proceeds in three steps: 1) isolating a list of terms via predictive screening, 2) assigning prediction weights to these words via topic modeling, and 3) aggregating terms into an article level predictive score via penalized likelihood. We derive theoretical guarantees on the accuracy of estimates from our model with minimal assumptions. In our empirical analysis, we study one of the most actively monitored streams of news articles in the financial system - the Dow Jones Newswires - and show that our supervised text model excels at extracting return-predictive signals in this context. Information in newswires is assimilated into prices with an inefficient delay that is broadly consistent with limits-to-arbitrage (i.e., more severe for smaller and more volatile firms) yet can be exploited in a real-time trading strategy with reasonable turnover and net of transaction costs.
Choosing Investment Managers
We study how plan sponsors choose investment management firms from their opportunity set when delegating $1.6 trillion in assets between 2002 and 2017. Two factors play an influential role in choice: pre-hiring returns, and pre-existing personal connections between personnel at the plan (or consultant advising the plan), and the investment management firm. Post-hiring returns for chosen firms are significantly lower than those for unchosen firms. The post-hiring returns of firms with relationships are, at best, indistinguishable from those without relationships, and often significantly worse. While relationships are conducive to asset gathering by investment managers, they do not appear to generate commensurate benefits for plan sponsors via higher gross returns or lower fees.
Do Venture Capitalists Stifle Competition?
We find that common ownership leads VCs to stifle competition among startups, but only in limited circumstances. Our evidence is from pharmaceutical startups, where common ownership is widespread. We examine how a startup responds after seeing a competitor make progress on a related drug project. If the two startups share a common VC, the lagging startup is less likely to advance its own project and obtain VC funding, which reduces competition between the startups. These anticompetitive effects, however, are limited to concentrated product markets, technologically similar projects, early-stage projects, and VCs with larger equity stakes and less-diversified portfolios.
Cancer and Mortality: The Home Equity Channel
We show that incompleteness in health insurance has a substantial effect on mortality among cancer patients, and we identify the channels through which this effect occurs. As theory predicts, high-wealth patients draw on that wealth (using credit markets) to fund out-of-pocket costs, while low-wealth patients rely on the implicit insurance provided by bankruptcy and other legal institutions to bear out-of-pocket costs. However, wealth matters for mortality: After instrumenting home equity wealth using house price variation, we find that high-wealth patients are substantially more likely to perform treatment, which prolongs survival. Thus, the implicit insurance provided by bankruptcy and other laws is insufficient to induce life-saving health choices. Finally, and surprisingly, we find that modest-wealth households, whose assets would be completely protected in bankruptcy, nonetheless exhaust their wealth in order to cover out-of-pocket costs. This finding suggests important limits to the role of bankruptcy as a form of implicit insurance.
Environmental Externalities of Hedge Fund Activism
We show that inWe study the effect of hedge fund activism on corporate environmental behaviors. Using plant-chemical level data from the EPA, we find that activism campaigns are associated with a 17 percent drop in emissions for chemicals at plants of targeted firms. Campaigns are associated with changes across a wide range of chemicals, including those emitted into the air, water, and ground and those that are harmful to humans. Evidence suggests this change in environmental behavior stems from a drop in production rather than an increase in abatement activities. The net effect on environmental efficiency is positive, with emissions falling by 8 percent per unit of output. Overall, our findings highlight the idea that the benefits of activism are not necessarily confided to shareholders, but may also extend to other stakeholders (e.g., the local community) affected by firms' emissions.
Time-Series Effcient Factors
"Factors in prominent asset pricing models are positively autocorrelated. We derive a transformation that turns an autocorrelated factor to a \time-series efficient"" factor. Time-series efficient factors earn significantly higher Sharpe ratios than the original factors and contain all the information found in the original factors. Momentum strategies profit from the same predictable variation in factor premiums as time-series efficient factors. An asset pricing model with time-series efficient factors, such as an efficient Fama-French five-factor model, therefore prices momentum. "
2019-2020
Awareness of credit information sharing: Evidence from field experiments
We examine the effect of lender credit information sharing on first-time borrowers’ loan take-up and default decisions using a pair of natural field experiments. Upon receiving credit warnings after taking out a loan, borrowers’ default likelihood decreases by 6%, suggesting an improvement in repayment effort. Upon receiving the same information after loan approval but before take-up, borrowers are 3% more likely to take out the loan, suggesting that credit reporting allows them to establish a credit history. Default likelihood is comparable between the two experiments, implying that credit reporting has little effect on borrowers’ adverse selection.
A sample of some current research with a focus on a method for blending theory and data in developing a model
The talk will briefly highlight ongoing research focused on assessing aggregate disadoption, the impact of category layouts, and fund allocation by individual investors. It will then focus on a method which combines theory (intuition) and data (the empirical relations among variables) to "automatically" build a model linking the variables.
How Market Power affects the Price-Inventory Relationship: Evidence from Car Dealerships
This paper investigates the effect of market power on dynamic pricing in the presence of inventories. Specifically, we analyze how automotive dealerships adjust prices to inventory levels under varying degrees of market power. First, we show that inventory fluctuations create scarcity rents for cars that are in short supply. Our empirical results show that a dealership moving from a situation of inventory shortage to a median inventory level lowers transaction prices by about 0.6% ceteris paribus, corresponding to 37% of dealers' average per vehicle profit margin or $165 on the average car. Then, we show that dealer's ability to adjust prices in response to inventory depends on their market power, i.e., the quantity of substitute inventory in their selling area. Specifically, we show that the slope of the price-inventory relationship (higher inventory lowers prices) is significantly steeper when dealers find themselves in a situation of high rather than low market power.
How does media coverage affect corporate disclosures? Finance seminar
We analyze the effect of media coverage on corporate voluntary disclosure. Specifically, we examine a model where a journalist, who covers a firm, reports noisy information which may be disclosed voluntarily also by the firm's manager. We show that, in equilibrium, media coverage crowds out the manager's voluntary disclosure, the manager delegates to the journalist some information provision, and investors are more confident with the journalist's reports about low values. Next, we analyze a setting of a sequential provision of information, where we examine the manager's response to news reports. We show that the manager may withhold information that is better than was reported by the journalist but respond to the news report with a disclosure of information that is less favorable.
Equilibrium Counterfactuals - Finance seminar
We incorporate structural modellers into the economy they model. Using the traditional moment-matching method, they ignore policy feedback and estimate parameters using a structural model that treats policy changes as zero probability (or exogenous) "counterfactuals." Estimation bias occurs since the economy.s actual agents, in contrast to model agents, understand policy changes are positive probability endogenous events guided by the modellers. We characterize equilibrium bias. Depending on technologies, downward, upward, or sign bias occurs. Potential bias magnitudes are illustrated by calibrating the Leland (1994) model to the Tax Cuts and Jobs Act of 2017. Regarding parameter identi.cation, we show the traditional structural identifying assumption, constant moment partial derivative sign, is incorrect for economies with endogenous policy optimization: The correct identifying assumption is constant moment total derivative sign accounting for estimation-policy feedback. Under this assumption, model agent expectations can be updated iteratively until the modellers' policy advice We analyze the effect of media coverage on corporate voluntary disclosure. Specifically, we examine a model where a journalist, who covers a firm, reports noisy information which may be disclosed voluntarily also by the firm's manager. We show that, in equilibrium, media coverage crowds out the manager's voluntary disclosure, the manager delegates to the journalist some information provision, and investors are more confident with the journalist's reports about low values. Next, we analyze a setting of a sequential provision of information, where we examine the manager's response to news reports. We show that the manager may withhold information that is better than was reported by the journalist but respond to the news report with a disclosure of information that is less favorable.
2018-2019
2017-2018
Maturity Driven Mispricing of Options
Options on US equities typically expire on the third Friday of each month, implying that either four or five weeks elapse between two consecutive expiration dates. We find that options that are held from one expiration date to the next achieve significantly lower returns when there are four weeks between expiration dates. The average difference in returns ranges from 12 basis points per week for delta-hedged put portfolios to 89 basis points for straddles (6.5% and 58.5% annualized, respectively). We find consistent results from an alternative price-based measure of mispricing. We perform multiple tests to examine the risk of our option portfolios and do not find any underlying risk patterns that can potentially explain our results. We therefore argue that the mispricing we identify is due to investor inattention to the exact expiration date, and provide further supporting evidence based on earnings announcements and price patterns closer to maturity. Our results survive a series of robustness tests and are unlikely to be driven by transaction costs. Overall, our evidence points to a potential strong behavioral bias among option traders.
Text Selection
Text data is inherently high-dimensional, which makes machine learning regularization techniques natural tools for its analysis. Text is often selected by journalists, speechwriters, and others who cater to an audience with limited attention. We develop an economically-motivated high dimensional selection model that can improve machine learning from text in particular and from sparse counts data more generally. Our highly scalable approach to modeling coverage selection is especially useful in cases where the cover/no-cover choice is separate or more interesting than the coverage quantity choice. We apply this framework to option-implied volatility (VIX) prediction using newspaper coverage, and find that it substantially improves out-of-sample fit relative to alternative state-of-the-art approaches. This advantage increases with the sparsity of the text.
Valuation Uncertainty and Short-Sales Constraints: Evidence from the IPO Aftermarket
We use the IPO setting to provide evidence that accounting measures of valuation uncertainty combine with short-sales constraints to generate significant equity market mispricing. The IPOs that we predict to be most susceptible to overpricing in the immediate aftermarket have first-day returns of +47% and lockup expiration returns of −9%. Our detailed analysis of securities lending market data confirms that these IPOs experience severe short-sales constraints that peak around the lockup expiration. Our paper both explains the anomalous pricing of IPOs and highlights the importance of valuation uncertainty and short-sales constraints in explaining equity mispricing.
Picking Friends Before Picking (Proxy) Fights: How Does Mutual Fund Voting Shape Proxy Contests
This paper studies mutual fund voting in proxy contests using a comprehensive sample of voting records over the period 2008 – 2015, taking into account selective targeting by activists. We find that firm, fund, and event characteristics generate substantial heterogeneity among investors in their support for the dissident, including their reliance on proxy advisors. Notably, active funds are significantly more pro-dissident than passive funds, and we uncover evidence consistent with a large unobserved fund "inherent stance" that cannot be explained by observable fund or event characteristics. In particular, we document a positive correlation between the propensity for targeting by activists and pro-activist voting by mutual funds, both based on the observables and unobservables. This finding suggests that a relatively pro-activist shareholder base is a key factor driving activists' selection of targets.
Glued to the TV: Distracted Investors and Stock Market Liquidity
We study the causal effect of trading on stock market liquidity. We exploit episodes of sensational news (exogenous to the market) that distract retail investors. On “distraction days” we find that trading activity, liquidity, and volatility all decline among stocks owned predominantly by retail investors. These findings, complemented by additional tests, establish that retail investors contribute to liquidity by serving both as noise traders and as liquidity providers. They also identify adverse selection as an important driver of illiquidity, thereby countervailing recent work that assigns a leading role to inventory risk or questions the usefulness of adverse selection measures.
Disclosure Harmonization and Corporate Acquisitions
This paper examines the effect of disclosure harmonization on the market for corporate control. For identification, we exploit a major regulatory change; the implementation of the Transparency Directive of 2004 (TPD), a legislation aimed at harmonizing information requirements for security issuers across the European Union (EU). We find that the TPD is followed by a substantial decrease in the number of control acquisitions. This pattern is concentrated in countries with higher pre-levels of takeover activity and lower pre-levels of acquisition costs. Consistent with the decrease of takeover activity being related to an increase in acquisition costs, we find that, under the TPD, targets’ takeover premiums increase and acquirers’ stock returns around the announcement decrease. Finally, additional tests suggest that these patterns are likely driven by mandatory disclosure of shareholders’ holdings (rather than by provisions related to issuers’ financial reporting). Overall, our evidence suggests that the disclosure harmonization introduced by the TDP decreased differences in takeover activity across EU countries. However, rather than stimulating less active takeover markets, the directive could have slowed down more dynamic markets; an effect likely driven by the higher acquisition costs associated with tighter ownership disclosure rules.