Embedded Files
We examine the spillover effects of local initial public offerings (IPOs) on new business formation. An IPO in a local area is associated with a 1% to 4% increase in new business registrations, and this effect is particularly pronounced in counties facing higher economic uncertainty. New business registrations are significantly influenced by the extent of EDGAR downloads related to the IPO firm’s public disclosures and the information in the IPO firm’s S-1 disclosure. The findings highlight the role of IPOs in conveying crucial information through signaling of potential success prospects and additional provision of information through disclosures. A field survey further supports these conclusions.
Journal of Accounting Research (Accepted)
This figure presents the regression coefficient estimates in 7 quarters around the event window, from the regression in equation (1) with controls, time and location fixed effects. Each dot is a point estimate of δ in equation (1) by shifting 𝑃𝑜𝑠𝑡 from Postq−3 to Postq+3. The shaded region represents two-tailed 95% confidence intervals based on standard errors clustered at the county level. Detailed definitions of all variables are provided in Appendix A.
We examine the labor market consequences of the 2020 Regulation S-K requiring human capital disclosure in 10K filings. Using large-sample job-level data and a Generative Large Language Model (GLLM), we observe that public firms subject to the regulation increase their disclosure of diversity, equity, and inclusion (DEI) information in job postings relative to a matched sample of large private firms. The increase in job-posting disclosure is more pronounced among firms facing greater external pressure to increase their workforce diversity. These findings suggest a shift in demand for diverse candidates by public firms following the regulation. Yet, consistent with short-term inelastic labor supply, this demand shift lengthens the recruitment period, with noticeable improvements in workplace gender diversity emerging one year after the regulation, particularly among firms that demonstrate a credible commitment to DEI. Our study documents how securities regulations can impact labor market practices and underscores the challenges involved in improving workforce diversity.
Journal of Accounting Research (Accepted)
The figure reports results from Eq. (1) in event time with 95% confidence interval. The orange dotted vertical line indicates the amendment of Regulation S-K in August 2020, while the red dotted vertical line indicates the deadline when the regulation took effect in November 2020. On the x-axis, we report the months relative to August 2020. On the y-axis, we report IHS(DEI Length).
Early empirical evidence showed a lack of relative performance evaluation (RPE) for executive pay, a surprise given its theoretical appeal. We hypothesize that executive pay transparency can enhance the monitoring of pay practices and increase RPE use. We examine RPE over the two decades centered on the 2006 executive pay disclosure reforms in the United States, which stakeholders—including shareholders, proxy advisors, and compensation consultants—could use to monitor pay plans. Firms that increase disclosures exhibit a significant increase in RPE after the reforms. To understand why, we examine and document that (i) stakeholder attention to pay practices increases after the reform, (ii) stakeholder attention positively relates to increases in RPE, and (iii) say-on-pay voting confirms shareholders’ preference for RPE. Overall, our findings are consistent with executive pay transparency increasing RPE due to enhanced pay monitoring across stakeholders.
Review of Accounting Studies (2025)
This figure shows the point estimates and confidence intervals from pay for systematic firm performance regressions over time. For each calendar year, we regress CEO pay on predicted firm performance using three years of data centered around the calendar year and plot the coefficient and 95% confidence interval on the predicted performance measure. For this graph, the dependent variable is Ln(Total Pay), and the independent variable of interest is predicted Ln(MVE). The green and brown lines represent the average of coefficient estimates in the first and second halves of our sample, respectively. The gray regions are the corresponding 95% confidence intervals obtained through bootstrapping. Specifically, we resample our data, estimate the cross-sectional regressions, and calculate time-series averages 1,000 times. Then we calculate standard errors based on the bootstrapped averages and derive the 95% confidence intervals.
We combine U.S. Census data with SEC enforcement actions to examine employees' outcomes, such as wages and turnover, before, during, and after periods of fraudulent financial reporting. We find that fraud firms’ employees lose about 50% of cumulative annual wages, compared to a matched sample, and the separation rate is much higher after fraud periods. Yet, employment growth at fraud firms is positive during fraud periods; these firms overbuild and hire new, lower-paid employees concurrent with the fraud, unlike firms in distress which tend to contract. When the fraud is revealed, firms shed workers, unwinding this abnormal growth and resulting in most of the negative wage consequences. Wage outcomes are particularly unfavorable in thin labor markets, and lower-wage employees, though unlikely to have perpetrated the fraud, experience more severe wage losses compared to higher-wage employees.
Journal of Accounting and Economics (2024)
This figure shows magnitude estimates from OLS regression analyses estimating equation (1): estimates for earnings effects at fraud firms (y-axis) in the by-event-time years (x-axis). Point estimates are incremental earnings of employees at fraud firms relative to those at matched control firms. We adjust the coefficient estimates from column (3) in Table 3 to percentages. We also show 90% confidence interval estimates as vertical bars through the point estimates. These standard errors are calculated with clustering by pre-fraud employer (i.e., fraud firm or matched control firm).
Motivated by the Financial Accounting Standards Board’s project on the disaggregation of income statement expenses, we study a Korean rule change that allowed firms to withhold a previously mandated disaggregation of cost of sales (CoS). We find that after withholding, firms’ profitability increases by 1.6 percentage points. Our industry-focused results suggest that withholding affects profitability by reducing the transfer of competitive information to peer firms. We then document a range of evidence consistent with the idea that firms withhold disaggregated CoS to protect cost innovations from rivals. First, we construct a novel measure of firms’ cost-innovative potential and show that it predicts withholding and subsequent profitability gains under the voluntary disclosure regime. Second, we document efficiency gains following the withholding of disaggregated CoS. Third, our survey experiment of 1,257 U.S. public firm managers shows that they would reduce investments in process/cost innovations if they were required to disaggregate CoS. Our study highlights to standard setters and academics that CoS disaggregation entails operational consequences for firms.
Management Science (2024)
This figure plots the average Profitability differences between two sets of Korean listed firms that both disclosed the Cost of Sales Schedule (CSS) from 1999 to 2003 when it was mandated: those that also disclosed the CSS from 2004 until 2008, and those that continually withheld the CSS over the same 2004 – 2008 period. The solid line shows average non-disclosers’ Profitability minus that of disclosers, by year. The dotted line shows the average differences from 2000 to 2004 and from 2005 to 2009. Profitability is orthogonalized against the set of control variables described in Section 4.2. Standard errors are estimated for each year and clustered by industry. The Appendix contains variable definitions.
We examine whether financial reporting quality affects worker wages using employer-employee matched data in the United States. We find that low financial reporting quality is associated with a compensating wage differential—that is, a risk premium—using three distinct approaches while controlling for worker characteristics by (1) regressing wages on firm-year–level and firm-level reporting quality, (2) documenting wage changes when workers switch firms, and (3) estimating a structural approach that separates reporting quality from performance-related volatility. We find evidence consistent with two channels: performance pay and turnover risk, where workers bear risks from noise in performance measurement and unemployment, respectively. To mitigate endogeneity concerns, we show that—after the accounting scandals in 2002 and after the announcements of an internal control weakness (ICW)—former Arthur Andersen clients and ICW firms pay wage premiums to employees, with magnitudes between 0.9% and 2.8% of annual wages.
Journal of Accounting Research (2023)
This figure shows wages in event time for workers that switch firms. Firms are classified as high or low financial reporting quality based on Abnormal Accruals. Abnormal Accruals are the absolute value of a residual from a regression of accruals on lagged, current, and future cash flows and sales changes and tangible assets. Workers are then categorized by switches as (i) switching between firms with both high financial reporting quality: “High to High,” (ii) switching from a firm with high financial reporting quality to a firm with low financial reporting quality: “High to Low,” (iii) switching from a firm with low financial reporting quality to a firm with high financial reporting quality: “Low to High,” or (iv) switching between firms with both low financial reporting quality: “Low to Low.” This figure measures wages relative to the pre–switch year. See appendix table A for more detailed variable descriptions. Statistics are rounded to comply with disclosure requirements of the U.S. Census Bureau.
We study the information content of earnings announcements and its relevance for job search using detailed search data from half a million anonymous job seekers. We find evidence consistent with job seekers initiating job-search activity in response to a prospective employer's earnings announcements. Job seekers search more intensely for employers with media coverage and earnings growth, consistent with the attention and information roles of earnings announcements. We find corroborating evidence about the usefulness of earnings announcements' financial information content to job seekers: (1) a survey experiment indicates that job seekers are more willing to apply to firms when provided with evidence of positive performance; (2) job seekers search for financial information during applications and interviews; and (3) financial information is predictive of future job prospects, including job openings and career growth. Overall, our paper suggests that earnings announcements—among other sources—prompt and guide job seekers' search activities.
Journal of Accounting and Economics (2023)
This figure summarizes search activity in response to earnings announcements for all earnings events. This figure plots coefficients from the full specification in Table 3 Column (3). ScaledSearch is a dependent variable. Event Week 0 (EW 0) is the week of the earnings announcement. ScaledSearch is defined as the count of unique job seekers who search for a given firm scaled by the number of TeamBlind users who identify as employees as that firm. ScaledSearch primarily measures search by prospective employees. The sample firms are those matched between the TeamBlind data set and the IBES data set. The sample period is Q3 2018 to Q1 2021. The unit of observation is firm-week-year. Variable descriptions are included in the Appendix. Clustered standard errors by firm are used to report the 95% confidence internal for each coefficient value.
We examine how information about the diversity of a potential employer's workforce affects individuals’ job-seeking behavior. We embed a field experiment in job recommendation emails from a leading career advice agency in the United States. The experimental treatment involves highlighting a diversity metric to jobseekers. Our results indicate that disclosing diversity scores in job postings leads jobseekers to click on firms with higher diversity scores, with such effects varying across jobseeker demographics. A follow-up survey provides evidence on potential explanations for why jobseekers value diversity information. We then examine how jobseekers’ preferences for diversity relate to disclosure choices under the U.S. SEC Human Capital Disclosure requirement. We find that firms in industries characterized by higher jobseeker responsiveness to diversity information tend to voluntarily disclose diversity metrics in their 10-Ks under these new disclosure requirements.
Journal of Accounting Research (2023)
This figure depicts the experimental design of different emails sent to users. The figure describes which information users under different conditions face. For example, under both conditions, jobseekers receive information about employer name, job position, median salary, and location. Only under the Diversity condition do jobseekers receive information about Diversity Scores.
This paper evaluates the role of accrual accounting in improving firms' production decisions and resource allocation across firms. I introduce two imperfect firm-performance measures, cash flows and accounting earnings, into a general equilibrium model with heterogeneous firms under imperfect information. The model demonstrates that improvements in measurement systems lead to more informed decisions on the part of firms and ultimately to allocation of greater resources to high-productivity firms via the product and input markets. Estimated parameter values are consistent with accrual accounting improving managers' information about current productivity by providing a better measure of historical firm performance. Quantitative analysis suggests that introducing accrual-accounting information on top of cash-accounting information leads to a 0.5% increase in aggregate U.S. productivity and a 0.7% increase in aggregate U.S. output via improved resource allocation. The corresponding estimates for China and India, as benchmarks for developing countries, are larger: a 1.5% to 2.3% increase in aggregate productivity and a 2.3% to 3.4% increase in aggregate output. I conclude that accrual accounting plays a significant role in determining aggregate productivity via improved resource allocation.
Journal of Accounting Research (2021)
This figure shows the aggregate productivity and output effect of accrual accounting in U.S., China, and India. These thoughts experiments are based on the quantitative general equilibrium model with the estimated parameters of quality of accrual accounting and cash accounting systems.
In contrast to firm-level relations, researchers have found that aggregate earnings changes and aggregate stock returns are negatively related. In this paper, we construct new measures of aggregate earnings news based on revisions in analyst forecasts. The findings suggest aggregate earnings news is positively related to contemporaneous stock returns. The results also show that aggregate stock returns are positively related to unexpected aggregate forecast errors, and negatively associated with expected aggregate earnings growth. Taken together, these findings suggest the negative relation between aggregate earnings changes and aggregate contemporaneous stock returns results from the expected component of aggregate earnings, rather than aggregate earnings surprises.
Journal of Financial Markets (2016)
This figure plots histograms of regression coefficients from firm-level time-series regressions of stock returns on the earnings news measure. The analyst forecast revision is the change in analyst forecasts for firm i over quarter t, scaled by market capitalization at the end of quarter t−1. The earnings news measure is the residual from the regression. The degrees of freedom in the firm-level time-series regressions are required to be greater than or equal to 20. In Panels A, B, and C the market-level coefficients (labeled “Market”) are calculated using a refined aggregate earnings news measure ηt (similar to Table 8) where the forecast revisions are scaled by lagged market capitalization, and the aggregate quarterly returns. Panel D employs the quarterly cumulative CRSP value-weighted index and the refined aggregate earnings news measure, ηt. The sample consists of S&P 500 non-financial firms, between 1983 and 2010, with analyst earnings forecasts available on IBES. The analysis is based on quarterly data. Panel A reports results for the full sample, while Panel B reports results for the period with increased analyst coverage. Panels C and D report results excluding the recent financial crisis, 2007:Q4–2009:Q1.
This paper develops and estimates a structural model of the labor market for accountants that integrates forward-looking lifetime occupational choices with oligopsonistic employer demand. Using longitudinal resume data covering career transitions of business graduates across six sectors (Big 4, non-Big 4, internal accounting, finance/consulting, technology, and other), we specify a discrete choice dynamic programming model in which workers accumulate sector-specific skills while considering monetary and non-monetary factors, and employers compete for talent under imperfectly elastic labor supply. Estimates indicate that accounting labor markets are characterized by economically significant markdowns (the gap between a worker’s marginal product and their wage) and entry barriers. Counterfactual analyses suggest that reducing employer market power and, to a lesser degree, eliminating entry barriers would increase accountants’ lifetime career values and improve (labor) resource allocation across sectors.
Working Paper (2025)
This figure shows the job category distribution for career positions held by business majors including accounting majors in the United States from 1 to 15 years(s) (inclusive) after graduation using Revelio data. Jobs held for over 15 years are collapsed under the distribution Year 15+. Career positions are separated into one of six categories based on the following criteria. 1) Big4: The position is labeled with an NAICS code beginning with 5412 (Accounting, tax preparation, bookkeeping, and payroll services), and job descriptions include accounting, accountant, audit, or financial reporting at one of the Big 4 accounting firms: Deloitte, PwC, EY, or KPMG. 2) NonBig4: The position is labeled with an NAICS code beginning with 5412, it is not at one of the Big 4 accounting firms, and job descriptions include accounting, accountant, audit, or financial reporting. 3) IntAcct: The position is not captured by categories 1 or 2 and has an aggregated role involving accounting. 4) FinCon: As a finance or consulting job at any firm, the position is labeled with an NAICS code beginning with 52 (Finance and Insurance) or 5416 (Management, Scientific, and Technical Consulting Services). 5) Technology: The position is labeled with an NAICS code beginning with 51 (Information), 5415 (Computer Systems Design and Related Services), or 5417 (Scientific Research and Development Services Techniques). 6) NonAcctJob: All other positions.
This paper provides early evidence on the integration and impact of Generative Artificial Intelligence (GenAI) in accounting at both the accountant and task levels. Using a multi-method approach, we first identify the heterogeneous adoption patterns, perceived benefits, and concerns through survey data from 277 accountants, which we formalize into testable predictions. Leveraging field data from a technology firm providing AI-based accounting software to 79 small- and medium-sized enterprises, we analyze hundreds of thousands of transaction-level records. We document significant productivity gains among adopters, including an 18% increase in weekly client support per standard deviation of AI use, with estimates reaching 59% when comparing the highest to lowest usage levels. We also find a reallocation of approximately 9% of accountant time from routine data entry toward high-value tasks. AI use corresponds to improved financial reporting quality, evidenced by a 12% increase in ledger granularity and a 7.5-day reduction in monthly close time. Experienced accountants selectively intervene when AI confidence scores are low, suggesting complementarity between expertise and AI. A framed field experiment shows that accountants sometimes over-rely on inaccurate AI-suggested classifications, illustrating risks of AI use. Overall, our findings highlight AI’s role in potentially augmenting, rather than replacing, professional accounting expertise.
Working Paper (2025)
This table evaluates how adoption of the Partner Firm’s AI system affects financial reporting quality, measured by the number of unique accounts (# Unique Accounts) and the timeliness of transaction recording (Average Days to Record Transactions). The analysis is conducted at the client-month level using ordinary least squares regressions with client and month fixed effects. The key treatment variable, Post AI in FR, equals 1 starting with the month in which a client first uses the partner firm’s system and remains 1 thereafter. Columns 1 and 2 include this single indicator, whereas columns 3 and 4 replace it with a set of mutually exclusive dummy variables that capture the first through fourth postimplementation months (t+1 to t+3, t+4 and later). All specifications include separate client and month fixed effects; standard errors are clustered by client. Appendix A provides detailed definitions of all variables.
In recent years, U.S. public companies have voluntarily disclosed official workforce diversity data (i.e., EEO-1 reports). To understand the factors leading these corporations to release this information publicly, we conduct a field experiment by reaching out to IR and HR personnel at about 4,000 large U.S. firms that currently do not publicly disclose their EEO-1 reports. We experimentally vary the information content of our requests. We find that companies are more likely to respond when directly considering investors’ rather than employees’ informational demands. On the other hand, we do not find any evidence that companies are more likely to respond when directly considering S&P 100 firms’ disclosure decisions. Both IR and HR departments play key roles in the process. A follow-up analysis identifies deterrents to disclosure, including legal and political considerations. Taken together, our results highlight understudied stakeholders and organizational characteristics that play a role in workforce diversity disclosure.
Working Paper (2025)
This table shows to which extent which companies take investors’ versus employees’ interests into consideration when making public disclosure decisions using 3,879 firms in our experiment. Appendix 1 contains all variable definitions. In this table, we focus on the cross-sectional interaction effects between firm’s Glassdoor DEI score and the investor treatment. We omit the Employees & No Peer treatment (reference group). In all four columns of both panels, we control for the inverse hyperbolic sine of sales and employment, private firm and day-sent (i.e., the day on which the email to the firm was sent) fixed effects as well as the peer treatment arm (and its interaction in Panel B). *** p<0.01, ** p<0.05, * p<0.1.
The under-representation of women within the high-tech sector is well-recognized. Platforms that match buyers and sellers in labor markets have been proposed as a means of increasing women’s participation in high-tech. In this study we conduct a randomized field experiment on a professional job search platform. This allows us to investigate how gender-specific information provided by the platform increases women’s engagement in searching for jobs in the tech industry. The inclusion of gender-specific, disaggregated information versus non-gendered aggregated information diminishes user engagement on the platform for all gender groups. However, further analyses of the experimental data suggest the benefit of gender-specific information for women’s search behavior. To study this underlying mechanism, we conduct a follow-on survey. We find that women exhibit a relatively heightened interest in gender-specific, disaggregated job market information. Overall, our study suggests a nuanced perspective on the informational value of gender-specific data in the job search context.
Working Paper (2025)
This table describes the linear probability model of a user starting a session by clicking on the push notification. The baseline is the control push notification. Note that we combine the two treatments into one indicator variable, since both treatments received the same push notification. The first column restricts the sample to users who are men, the second column restricts the sample to users who are women, and the last two columns use both men and women.
Page updated
Google Sites
Report abuse