Search

Why are women underrepresented in politics? This paper documents gender differences in the career paths of novice politicians by studying the persistence of candidates after they win or lose elections. I track the political trajectories of over 11,000 candidates in local California elections and use a regression discontinuity approach. Losing an election causes 50 percent more attrition among female than male candidates: an electoral loss causes men to be 16 percentage points less likely to run again within the next four years, whereas the drop for women is 25 percentage points. Yet the gender gap in persistence depends on the setting: I find no evidence of a gap among candidates for high female representation offices or among more experienced candidates. These results are inconsistent with behavioral explanations of women’s differential attrition. Instead, the results suggest that in low information environments, voters may penalize novice female politicians, which deters women from running again. I discuss the implications of the results for the gender gap in officeholding.

This paper estimates a panel model with endogenously time-varying parameters for COVID-19 cases and deaths in U.S. states. The functional form for infections incorporates important features of epidemiological models but is flexibly parameterized to capture different trajectories of the pandemic. Daily deaths are modeled as a spike-and-slab regression on lagged cases. The paper's Bayesian estimation reveals that social distancing and testing have significant effects on the parameters. For example, a 10 percentage point increase in the positive test rate is associated with a 2 percentage point increase in the death rate among reported cases. The model forecasts perform well, even relative to models from epidemiology and statistics.

We apply advances in analysis of mixed frequency and sparse data methods to estimate ``unsmoothed'' private equity (PE) Net Asset Values (NAVs) at the weekly frequency for individual funds. Using simulations and a large sample of buyout and venture funds, we show that our method yields superior estimates of NAVs relative to a simple approach based on comparable public asset and as-reported NAVs. The market beta of an average buyout [venture] fund, around 1.05 [1.20], is notably lower than previous studies suggest. The overall risk of a median buyout [venture] fund is 34% [40%] per year if measured in terms of the standard deviation of total returns, an increase of 10 [13] percentage points relative to the series based on as-reported NAVs. However, we find significant variation in systematic and idiosyncratic risk within and across vintages, and the risk-return profile based on the samples from the 1990s is not representative of currently operating funds. Our method easily accommodates additional data, such as individual holdings details, and related M&A activity. As an example application, we show how the cross-section of appraisal biases and nowcast errors evolved around the 2008 Financial crisis.

We study the results of a massive nationwide correspondence experiment sending more than 83,000 fictitious applications with randomized characteristics to geographically dispersed jobs posted by 108 of the largest U.S. employers. Distinctively Black names reduce the probability of employer contact by 2.1 percentage points relative to distinctively white names. The magnitude of this racial gap in contact rates differs substantially across firms, exhibiting a between-company standard deviation of 1.9 percentage points. Despite an insignificant average gap in contact rates between male and female applicants, we find a between-company standard deviation in gender contact gaps of 2.7 percentage points, revealing that some firms favor male applicants and others favor women. Company-specific racial contact gaps are temporally and spatially persistent, and negatively correlated with firm profitability, federal contractor status, and a measure of recruiting centralization. Discrimination exhibits little geographical dispersion, but two-digit industry explains roughly half of the cross-firm variation in both racial and gender contact gaps. Contact gaps are highly concentrated in particular companies, with firms in the top quintile of racial discrimination responsible for nearly half of lost contacts to Black applicants in the experiment. Controlling false discovery rates to the 5% level, 23 companies are found to discriminate against Black applicants. Our findings establish that discrimination against distinctively Black names is concentrated among a select set of large employers, many of which can be identified with high confidence using large-scale inference methods.

Based on the MOD09GA/MYD09GA 500-m surface reflectance, a new MODIS snow-cover-extent (SCE) product over China has been produced by the Northwest Institute of Eco-Environment and Resources (NIEER), Chinese Academy of Sciences. The NIEER MODIS SCE product contains two preliminary clear-sky SCE datasets — Terra-MODIS and Aqua-MODIS SCE datasets, and a final daily cloud-gap-filled (CGF) SCE dataset. The formers are generated mainly through optimizing snow-cover discriminating rules over different land-cover types, and the latter is produced after a series of gap-filling processes such as aggregating the two preliminary datasets, reducing cloud gaps with adjacent information in space and time, and eliminating all gaps with auxiliary data. Validation against 362 China Meteorological Administration (CMA) stations shows during snow seasons the overall accuracies (OA) of the three datasets are all larger than 93 %, the omission errors (OE) are all constrained within 9 %, and the commission errors (CE) are all constrained within 10 %. Biases ranging from the lowest 0.98 to the medium 1.02, to the largest 1.03 demonstrate on a whole the SCEs given by the new product are neither overestimated nor underestimated significantly. Based on the same ground reference data, we found the new product’s accuracies are clearly higher than those of standard MODIS snow products, especially for Aqua-MODIS and CGF SCE. For examples, compared with the CE of 23.78 % that the standard MYD10A1 product shows, the CE of the new Aqua-MODIS SCE dataset is 6.78 %; the OA of the new CGF SCE dataset is up to 93.15 %, versus 89.54 % of the standard MOD10A1F product and 84.36 % of the standard MYD10A1F product. Besides, as expected snow discrimination in forest areas is also improved significantly. An isolated validation at four forest CMA stations demonstrates the OA has increased by 3–10 percentage points, the OE has dropped by 1–8 percentage points, and the CE has dropped by 4–21 percentage points. Therefore, our product has virtually provided more reliable snow knowledge over China, and thereby can better serve for hydrological, climatic, environmental, and other related studies there.

The age distribution influences capital flows through aggregate saving and labor supply. To quantify this, I build a dynamic model featuring overlapping generations and international trade among 28 countries since 1970. The equilibrium is replicated by a model with a representative household in each country that experiences an endogenous discount factor, which summarizes the co-evolution of demographics and relevant prices, affording computation of the exact transition. On average, a one-year increase a countrys mean age boosts its current account by 0.4 percent of GDP. Bilateral trade frictions dictate the cross-country response of capital flows to changes in a countrys demographics.

Suppose that technology reduces price-adjustment costs (e.g., the costs of printing and changing price tags), and as a result prices at grocery stores change more dynamically. Will this change mean less stability or more stability for grocery supply chains? In other words, will more dynamic pricing downstream mean more last-minute purchases, more overtime work, and more empty space in trucks and warehouses? Or will it mean more regular and more standardized orders, smoother schedules, and less waste? To answer this question, we fit a pricing and ordering model to data from a large Chinese supermarket chain (daily prices, sales, inventories, and shipments from products from seven categories at 78 stores for 3.5 years) and then simulate a counterfactual grocery chain in which the estimated price-adjustment costs are set to zero. We find that the removal of price-adjustment costs stabilizes the supply chain, reducing both its shipment volatility, its sales volatility, and its bullwhip (the difference between the shipment and sales volatility).

We evaluate electronic monitoring as an alternative to prison for nonviolent offenses. Leveraging plausibly exogenous variation in sentencing outcomes generated by quasi-random assignment of judges, we find electronic monitoring reduces reoffending at both extensive and intensive margins. Compared with prison, electronic monitoring is estimated to reduce the probability of reoffending by 22 percentage points five years after sentencing and by 11 percentage points ten years after sentencing, with the cumulative number of offenses reduced by 40% ten years after sentencing. These results demonstrate that electronic monitoring has sustained crime-reducing effects.

In this paper we study the impact of firm relocations on commuting distance and the probability of married couples and cohabiting couples with children separating. We use Swedish register data for the period 2010–16 and select employees of relocating firms with one workplace and more than 10 employees. Focusing on this sample allows us to use plausibly exogenous variation in the commuting distance arising from the relocation. We extend the literature on the effect of commuting on relationship stability by reducing the possibility for unobserved time-variant factors to bias our estimates. While previous literature has focused on the difference between short- and long-distance commuting, we focus on changes in the commuting distance that are externally induced by firm management. We find a small but statistically significant negative effect of increased firm relocation distance on family stability. A 10 km change in commuting distance leads to a 0.09 percentage point higher probability of separation if the commuter remains with the firm for the next five years.

In March 2018, an Uber-pedestrian crash and a Tesla's Model X crash attracted a lot of media attention because the vehicles were operating under self-driving and autopilot mode respectively at the time of the crash. This study aims to conduct before-and-after sentiment analysis to examine how these two fatal crashes have affected people's perceptions of self-driving and autonomous vehicle technology using Twitter data. Five different and relevant keywords were used to extract tweets. Over 1.7 million tweets were found within 15 days before and after the incidents with the specific keywords, which were eventually analyzed in this study. The results indicate that after the two incidents, the negative tweets on “self-driving/autonomous” technology increased by 32 percentage points (from 14% to 46%). The compound scores of “pedestrian crash”, “Uber”, and “Tesla” keywords saw a 6% decrease while “self-driving/autonomous” recorded the highest change with an 11% decrease. Before the Uber-incident, 19% of the tweets on Uber were negative and 27% were positive. With the Uber-pedestrian crash, these percentages have changed to 30% negative and 23% positive. Overall, the negativity in the tweets and the percentage of negative tweets on self-driving/autonomous technology have increased after their involvement in fatal crashes. Providing opportunities to interact with this developing technology has shown to positively influence peoples' perception.

Background The aim of the current study is to investigate the variations of anatomical (LVRem%) and functional remnant volumes (fLVRem%) and the dynamic uptake of Technetium-Mebrofinate (FRLF) measured from 99m Technetium-Mebrofinate SPECT-CT scan (TMSCT) in patients at high risk of post-hepatectomy liver failure (PHLF). Methods Variations in the measures of LVRem% and fLVRem% were assessed. The predictive accuracies of LVRem%, fLVRem% and FRLF with respect to PHLF were reported. Results From the N = 92 scans performed, LVRem% and fLVRem% returned identical results in 15% of cases, and ±10 percentage points in 79% of cases. Some patients had larger discrepancies, with difference of >10 percentage points in 21% of cases. The difference was significant in those with primary liver cancers (−4.4 ± 9.2, p = 0.002). For the N = 29 patients that underwent surgery as planned on TMSCT, FRLF was a strong predictor of PHLF, with an AUROC of 0.83 (p = 0.005). Conclusion TMSCT is emerging as a useful modality in pre-operative assessment of patients undergoing major liver resection. For those with primary liver cancer, there is a significant variation in the anatomical and functional distributions that needs considered in surgical planning. Reduced FRLF, measured as the dynamic uptake in the future liver remnant, is a strong predictor of PHLF.

Introduction This study examines whether smoke-free laws are differentially associated with youth smoking outcomes by parental education, race/ethnicity, sex, and college plans in a U.S. sample. Methods This study assessed the relationships between smoke-free laws in workplaces and hospitality venues (restaurants/bars) and past 30-day smoking participation, first cigarette initiation, and daily smoking initiation within a repeated cross-sectional sample of 8th, 10th, and 12th graders from the Monitoring the Future study. Data were collected between 2001 and 2018 and were analyzed in 2020–2021. Grade-stratified Poisson models were used to calculate prevalence ratios and average marginal effects, incorporating interaction terms to examine differential associations across groups. Results Hospitality smoke-free laws were significantly associated with lower probabilities of smoking participation in all grades as well as with first cigarette and daily smoking initiation in 8th and 10th grade. Workplace smoke-free laws were associated with lower probabilities of smoking participation among 10th and 12th graders as well as with first cigarette and daily smoking initiation among 10th graders. Average marginal effects ranged from −0.4 percentage points (hospitality laws and daily smoking initiation in 8th and 10th grades) to −2.2 percentage points (workplace laws and smoking participation in 10th grade). Associations between smoke-free laws and a lower probability of smoking participation were most pronounced among students who definitely planned to attend college. Other instances of effect modification suggested more pronounced associations for students who were female and from high-SES households; however, relationships varied by grade. Conclusions Smoke-free laws were associated with reduced smoking among youth; however, associations varied by grade, sex, parental education, and college plans.

How do barriers to the diffusion of academic research affect innovation? In 2008, the National Institutes of Health (NIH) mandated free online availability of funded research. This policy caused a 50 percentage point increase in free access to funded articles. We introduce a novel measure, in-text patent citations, to study how this mandate affected industry use of academic science. After 2008, patents cite NIH-funded research 12% to 27% more often. Nonfunded research, funded research in journals unaffected by the mandate, and academic citations see no change. These estimates are consistent with a model of search for useful knowledge. Inefficiency caused by academic publishing may be substantial.

This study shows that skilled-labor-intensive firms have lower dividend payouts. A one-standard-deviation increase in skilled labor intensity on average results in a 0.74 percentage point decline in the propensity to pay dividends and a 0.72 percentage point decrease in dividend payout ratio. Our evidence suggests that greater labor adjustment costs (LACs) of skilled labor increase future cash flow uncertainty, which discourages managers from paying dividends. Consistently, we find that an exogenous increase (decrease) in LACs following the shock of 2017 H-1B policy changes (the shock of 2005 Hurricane Katrina) leads treated firms to lower (heighten) their dividend payouts. Collectively, these findings indicate a strong dampening effect of LACs on corporate dividend payouts.

We study how the public changes their mobility and retail spending patterns as precautionary responses to the disclosed location of COVID-19 cases. To look into the underlying mechanisms, we investigate how such change varies spatially and whether there is any spatial spillover or substitution. We use the daily data of cell phone?based mobility and credit card transactions between February 10 and May 31 in both 2019 and 2020 in Seoul, South Korea, and employ the empirical approach analyzing the year-over-year percent change for the mobility and consumption outcomes. Results report that one additional COVID-19 case within the last 14 days decreased non-resident inflow and retail spending by 0.40 and 0.65 percentage points, respectively. Then, we also find evidence of spatial heterogeneity: the mobility and retail performances of neighborhoods with higher residential population density were more resilient to COVID-19 case information while neighborhoods with higher levels of land use diversity and retail agglomeration experienced a greater localized demand shock. This heterogeneity is not negligible. For example, one additional COVID-19 case in neighborhoods in the bottom 20% for population density led to a decline of 1.2 percentage points in retail spending, while other neighborhoods experienced a less negative impact. Finally, we find a significant spatial spillover effect of disclosed COVID-19 information instead of spatial substitution. One additional COVID-19 case in geographically adjacent areas within the last 14 days reduced non-resident inflow and retail spending in the subject neighborhood by 0.06 and 0.09 percentage points, respectively. This article is protected by copyright. All rights reserved.

OBJECTIVE We investigated asthma quality measures to understand patient characteristics associated with non-attainment of quality care and measure the association with asthma-related emergency department (ED) visits or inpatient hospitalizations (IPs). METHODS Using administrative data from ALL Kids, Alabama's Children's Health Insurance Program, from 2013 to 2019 we calculated non-attainment of the Medication Management for Asthma (MMA) and Asthma Medication Ratio (AMR) quality measures. Patient characteristics and asthma-related ED visits and IPs associated with non-attainment of the MMA and AMR measures were assessed using logit regression models and Marginal effects at the mean. RESULTS Among 2528 children with asthma, 53.2% failed to attain the MMA measure and 8.5% the AMR measure. Prior asthma-related ED visits or IP stays increased likelihood of non-attainment by 14.8 percentage points (95% CI 8.6-20.9) for MMA and 7.3 percentage points (95% CI 2.8-11.8) for AMR. Among 868 children (34.3%) with three years of continuous enrollment, AMR non-attainment was associated with a 6.1 percentage point increase in ED or IP utilization (95% CI 1.3-10.9), however MMA non-attainment was not associated with either outcome. Prior ED visit/IP stay was associated with a 17.2 percentage point (95% CI 8.3-26.1) increase in the likelihood of a subsequent ED visit/IP stay among those with non-attainment MMA and a 15.5 percentage point increase (95% CI 6.9-24.2) for non-attainment AMR. CONCLUSIONS Patient characteristics associated with non-attainment of asthma quality measures presents actionable evidence to guide improvement efforts as non-attainment AMR increases the risk of subsequent ED visits and IP stays.

We examine the relationship between parental smoking in childhood and the probability of being employed in adulthood. To do so, we use 18 waves of the nationally representative longitudinal data from the Household, Income, and Labour Dynamics in Australia (HILDA) survey. We find that parental smoking in childhood is an important factor contributing to the likelihood of individuals being unemployed in adulthood. Our estimate suggests that individuals whose parents smoked during their childhood have 1.7 percentage points lower probability of being employed in adulthood than individuals of non-smoking parents. This finding is robust to the use of bounding approach that tests for coefficient stability and matching methods that attribute causal interpretation. We find that self-health and mental health, along with non-cognitive skills in the form of emotional stability and conscientiousness in adulthood are important channels through which exposure to parental smoking in childhood influences the probability of being employed in adulthood.

The fundamental requirement for real-time operation of a speech-processing algorithm is causality—that it operate without utilizing future time frames. In the present study, the performance of a fully causal deep computational auditory scene analysis algorithm was assessed. Target sentences were isolated from complex interference consisting of an interfering talker and concurrent room reverberation. The talker- and corpus/channel-independent model used Dense-UNet and temporal convolutional networks and estimated both magnitude and phase of the target speech. It was found that mean algorithm benefit was significant in every condition. Mean benefit for hearing-impaired (HI) listeners across all conditions was 46.4 percentage points. The cost of converting the algorithm to causal processing was also assessed by comparing to a prior non-causal version. Intelligibility decrements for HI and normal-hearing listeners from non-causal to causal processing were present in most but not all conditions, and these decrements were statistically significant in half of the conditions tested—those representing the greater levels of complex interference. Although a cost associated with causal processing was present in most conditions, it may be considered modest relative to the overall level of benefit.

In this paper, we determine treatment effects when the treatment assignment is based on two or more cut-off points of covariates rather than on one cut-off point of one assignment variable. using methods that are referred to as multivariate regression discontinuity designs (MRDD). One major finding of this paper is the discovery of new evidence that both matric points and household income have a huge impact on the probability of eligibility for funding from the National Student Financial Aid Scheme (NSFAS) to study for a bachelor’s degree program at universities in South Africa. This evidence will inform policymakers and educational practitioners on the effects of matric points and household income on the eligibility for NSFAS funding. The availability of the NSFAS grant impacts greatly students’ decisions to attend university or seek other opportunities elsewhere. Using the frontier MRDD analytical results, barely scoring matric points greater than or equal to 25 points compared to scoring matric points less than 25 for students whose household income is less than R350,000 (≈US$2500) increases the probability of eligibility for NSFAS funding by a significant 3.75 ( p-value = 0.0001 &lt, 0.05) percentage points. Therefore, we have shown that the frontier MRDD can be employed to determine the causal effects of barely meeting the requirements of one assignment variable, among the subjects that either meet or fail to meet the requirements of the other assignment variable.

The purposes of this study are to measure the earnings difference and the factors that influence earnings difference between 2007 and 2014 using data sourced from the Indonesian Family Life Survey (IFLS) in 2007 and 2014. The income determinant analysis results found that the longer a person's education year and work experience, the higher the income. Income will be even greater if someone is a man, lives in urban areas, and works in the non-agriculture sector. Moreover, there is no evidence that religion and ethnicity affect income. Then, I used the Blinder-Oaxaca decomposition method to distinguish the factors contributing to the difference in income to be explained factors and unexplained factors. It was found that that the income gap between 2007 and 2014 was 13.2 percentage points. Endowment factor contribution is more significant than unexplained factors. Furthermore, decomposition at different income levels shows that the endowment factor's effect on earning difference is getting smaller at higher income levels.

We estimate the impact of the Social Security Annual Earnings Test (AET) on older workers’ employment. The AET reduces social security claimants’ current benefits in proportion to their earnings in excess of an exempt amount. Using a regression kink design and Social Security Administration data, we document that the discontinuous change in the benefit reduction rate at the exempt amount causes a corresponding change in the slope of the employment rate, suggesting that the extensive margin of labor supply is more sensitive to this policy than commonly thought. We develop a model and method that allow us to translate the behavioral responses into a lower bound estimate of 0.49 for the extensive margin elasticity, which implies more than a 1 percentage point increase in work in the absence of the AET. (JEL H55, J14, J22, J26, J31)

Background Characterizing phage–host interactions is critical to understanding the ecological role of both partners and effective isolation of phage therapeuticals. Unfortunately, experimental methods for studying these interactions are markedly slow, low-throughput, and unsuitable for phages or hosts difficult to maintain in laboratory conditions. Therefore, a number of in silico methods emerged to predict prokaryotic hosts based on viral sequences. One of the leading approaches is the application of the BLAST tool that searches for local similarities between viral and microbial genomes. However, this prediction method has three major limitations: (i) top-scoring sequences do not always point to the actual host; (ii) mosaic virus genomes may match to many, typically related, bacteria; and (iii) viral and host sequences may diverge beyond the point where their relationship can be detected by a BLAST alignment. Results We created an extension to BLAST, named Phirbo, that improves host prediction quality beyond what is obtainable from standard BLAST searches. The tool harnesses information concerning sequence similarity and bacteria relatedness to predict phage–host interactions. Phirbo was evaluated on three benchmark sets of known virus–host pairs, and it improved precision and recall by 11–40 percentage points over currently available, state-of-the-art, alignment-based, alignment-free, and machine-learning host prediction tools. Moreover, the discriminatory power of Phirbo for the recognition of virus–host relationships surpassed the results of other tools by at least 10 percentage points (area under the curve = 0.95), yielding a mean host prediction accuracy of 57% and 68% at the genus and family levels, respectively, and drops by 12 percentage points when using only a fraction of viral genome sequences (3 kb). Finally, we provide insights into a repertoire of protein and ncRNA genes that are shared between phages and hosts and may be prone to horizontal transfer during infection. Conclusions Our results suggest that Phirbo is a simple and effective tool for predicting phage–host relationships.

Fire sales induced by investor redemptions have powerful spillover effects among funds that hold the same assets, hurting peer funds' performance and flows, and leading to further asset sales with negative bond price impact. A one‐standard‐deviation increase in our fire‐sale spillover measure leads to a 45 (90) bp decrease in peer fund returns (flows) and a two percentage point increase in the likelihood of a large bond price drop. The results hold in a regression‐discontinuity design addressing identification concerns. Timing, heterogeneity, instrumental‐variable, and placebo tests further support the price‐impact mechanism. Model‐based counterfactual and stress‐test analyses quantify the financial stability implications.

Background For the recruitment and monitoring of subjects for therapy studies, it is important to predict whether mild cognitive impaired (MCI) subjects will prospectively develop Alzheimer’s disease (AD). Machine learning (ML) is suitable to improve early AD prediction. The etiology of AD is heterogeneous, which leads to high variability in disease patterns. Further variability originates from multicentric study designs, varying acquisition protocols, and errors in the preprocessing of magnetic resonance imaging (MRI) scans. The high variability makes the differentiation between signal and noise difficult and may lead to overfitting. This article examines whether an automatic and fair data valuation method based on Shapley values can identify the most informative subjects to improve ML classification. Methods An ML workflow was developed and trained for a subset of the Alzheimer’s Disease Neuroimaging Initiative (ADNI) cohort. The validation was executed for an independent ADNI test set and for the Australian Imaging, Biomarker and Lifestyle Flagship Study of Ageing (AIBL) cohort. The workflow included volumetric MRI feature extraction, feature selection, sample selection using Data Shapley, random forest (RF), and eXtreme Gradient Boosting (XGBoost) for model training as well as Kernel SHapley Additive exPlanations (SHAP) values for model interpretation. Results The RF models, which excluded 134 of the 467 training subjects based on their RF Data Shapley values, outperformed the base models that reached a mean accuracy of 62.64% by 5.76% (3.61 percentage points) for the independent ADNI test set. The XGBoost base models reached a mean accuracy of 60.00% for the AIBL data set. The exclusion of those 133 subjects with the smallest RF Data Shapley values could improve the classification accuracy by 2.98% (1.79 percentage points). The cutoff values were calculated using an independent validation set. Conclusion The Data Shapley method was able to improve the mean accuracies for the test sets. The most informative subjects were associated with the number of ApolipoproteinE ε4 (ApoE ε4) alleles, cognitive test results, and volumetric MRI measurements.

BackgroundOngoing shifts in economic structure from automation and globalisation can affect employment and mortality, yet these relations are not well described.ObjectiveWe assess whether long-term employment and health outcomes relate systematically to structural change in the labour market, using the occupational Routine Task Intensity (RTI) score as indicator of exposure is to risks of outsourcing and technology-induced job loss.MethodsUsing a cohort design and administrative data with national population coverage, we categorise all Norwegian employees in 2003 by the RTI score of their occupation and examine how this score correlates with employment and health outcomes measured in 2018 and 2019. The study sample counts 416 003 men and 376 413 women aged 33–52 in 2003.ResultsThe occupational RTI score at baseline is robustly associated with long-term employment, disability and mortality outcomes. Raw correlations are reduced after adjustment for potential confounders, but associations remain substantial in models controlling for individual covariates and in sibling comparisons. Working in an occupation with RTI score 1 SD above the mean in 2003 is associated with a raised probability of being deceased in 2019 of 0.24 percentage points (95% CI: 0.18 to 0.30) for men and 0.13 percentage points (95% CI: 0.02 to 0.24) for women, corresponding to raised mortality rates of 6.7% and 5.5%.ConclusionsIndividuals in occupations characterised by high routine intensity are less likely to remain employed in the long term, and have higher rates of disability and mortality.

How do early-life experiences shape political identity? We examine the end of race-based busing in Charlotte-Mecklenburg Schools, an event that led to large changes in school racial composition. Using administrative data, we compare party affiliation in adulthood for students who had lived on opposite sides of newly drawn school boundaries. Consistent with the contact hypothesis, we find that a 10 percentage point increase in the share of minorities in a White student’s assigned school decreased their likelihood of registering as a Republican by 2 percentage points (12 percent). Our results suggest that schools in childhood play an important role in shaping partisanship. (JEL D72, H75, I21, I28, J15)

Background In people with a Fontan circulation, serial cardiopulmonary exercise testing (CPET) to evaluate change in peak exercise capacity has been increasingly recognised as a useful prognostic tool; a decline is associated with adverse clinical outcomes. The aim of this study is to describe the “natural” history of exercise capacity in the Australian and New Zealand (ANZ) Fontan cohort and to identify factors associated with a decline. Methods The ANZ Fontan registry was retrospectively reviewed for adolescent and adult patients (≥16 years) with serial CPET results performed on a cycle ergometer ≥6 months apart. Patients were excluded if they underwent a surgical procedure or fenestration closure in-between tests or if the tests were considered as submaximal effort. Exercise capacity trajectory was defined as the change in percentage of predicted peak oxygen uptake (% pred VO2peak) points per year. Results Thirty-seven (37) patients (59.5% male, mean age 24±7 years) were eligible. Average duration between CPET was 5.3±3.9 years. At baseline, % pred VO2peak was 61.3±14.5%. Thirteen (13) (35%) had a systemic right ventricle, and 14 (38%) had an atriopulmonary type Fontan circulation. Average change in % pred VO2peak overall was +1.3±6.4 percentage points per year. Sixteen (16) had a negative exercise capacity trajectory, and the average decline in that group was −2.7±3.4 percentage points per year. There was no association between exercise capacity trajectory and clinical characteristics. Of the 18 patients with physical activity levels recorded, 12 (67%) were physically active and % pred VO2peak in that group increased by 2.7±4.0 percentage points per year compared with the physically inactive group who fell by 0.5±0.8 percentage points per year. Conclusions In this ANZ series of Fontan patients, over half of our cohort had stable, or an increase, in peak exercise capacity. Regular participation in physical activity was common in patients with a positive exercise capacity trajectory. Clinical characteristics were not associated with exercise capacity trajectory.

A basic aim of marketing research is to predict consumers’ preferences and the success of marketing campaigns at the population-level. However, traditional marketing tools have various limitations, calling for novel measures to improve predictive power. In this study, we use multiple types of measures extracted from electroencephalography (EEG) recordings and machine learning (ML) algorithms to improve preference prediction based on self-reports alone. Subjects watched video commercials of six food products as we recorded their EEG activity, after which they responded to a questionnaire that served as a self-report benchmark measure. Thereafter, subjects made binary choices over the food products. We attempted to predict within-sample and population level preferences, based on subjects’ questionnaire responses and EEG measures extracted during the commercial viewings. We reached 68.5% accuracy in predicting between subjects’ most and least preferred products, improving accuracy by 4.07 percentage points compared to prediction based on self-reports alone. Additionally, EEG measures improved within-sample prediction of all six products by 20%, resulting in only a 1.91 root mean squared error (RMSE) compared to 2.39 RMSE with questionnaire-based prediction alone. Moreover, at the population level, assessed using YouTube metrics and an online questionnaire, EEG measures increased prediction by 12.7% and 12.6% respectively, compared to only a questionnaire-based prediction. We found that the most predictive EEG measures were frontal powers in the alpha band, hemispheric asymmetry in the beta band, and inter-subject correlation in delta and alpha bands. In summary, our novel approach, employing multiple types of EEG measures and ML models, offers marketing practitioners and researchers a valuable tool for predicting individual preferences and commercials’ success in the real world.

We perform a meta analysis of gender differences in the standard windfall gains dictator game (DG) by collecting raw data from 53 studies with 117 conditions, giving us 15,016 unique individual observations. We find that women on average give 4 percentage points more than men (Cohen’s $$d=0.16$$ d = 0.16 ), and that this difference decreases to $$3.1\%$$ 3.1 % points (Cohen’s $$d=0.13$$ d = 0.13 ) if we exclude studies where dictators can only give all or nothing. The gender difference is larger if the recipient in the DG is a charity, compared to the standard DG with an anonymous individual as the recipient (a 10.9 versus a $$2.3\%$$ 2.3 % points gender difference). These effect sizes imply that many individual studies on gender differences are underpowered; the median power in our sample of standard DG studies is only $$9\%$$ 9 % to detect the meta-analytic gender difference at the $$5\%$$ 5 % significance level. Moving forward on this topic, sample sizes should thus be substantially larger than what has been the norm in the past.

With the implementation of five remediation strategies immediately after the first failed exam in a pharmacology course, this quality improvement study demonstrated an increase in retention of at-risk associate degree nursing students. Unit exam scores and course failure rates were measured before and after implementation. Twelve of 14 students who completed the remediation demonstrated a statistically significant increase in scores from the first failed exam to the last exam taken. The mean increase was 17.3 percentage points (p < .0001, 95 percent confidence interval [13.3, 20.9]). Cohen's d, which equaled 0.604, indicated a significant increase in course pass rates.

Strong empirical evidence points towards a significantly higher prevalence of opioid consumption among people receiving disability benefits (DB) than in the general population of the United States. However, no previous research established a causal relationship between the decision to award DB to applicants and their subsequent opioid use. We aim to contribute towards filling this gap. There are channels through which awarding DB may both increase and depress opioid consumption, and thus, ex ante, the sign of a potential causal relationship is ambiguous. To correct for the treatment endogeneity, since an individual's age at the time of the decision on an application impacts discontinuously at certain age cutoffs the award decision, we employ a fuzzy Regression Discontinuity model with three age cutoffs used for identification. We find that awarding DB increases the likelihood of using opioids by about 27-30 percentage points. This suggests that the positive association between DB receipt and opioid consumption is likely to be causal.

The BIS indicated in July 2020 an unprecedented rise in default risk correlation as a result of pandemics-induced credit risks’ accumulation. A third of the world banking assets credit risk measurement depends on the Basel internal-ratings-based (IRB) models. To ensure financial stability, we wish IRB models to be accurate in default probability (PD) forecasting. There naturally arises a question of which model may be deemed accurate if the data demonstrates the presence of the default correlation. The existing prudential IRB validation guidelines suggest a confidence interval of up to 100 percentage points’ length for such a case. Such an interval is useless as any model and any PD forecast seem accurate. The novelty of this paper is the justification for the use of twin confidence intervals to validate PD model accuracy. Those intervals more concentrate around the two extremes (default and its absence), the higher the default correlation is.

This study presents a methodology to predict the child poverty impact of COVID-19 that can be readily applied in other country contexts where similar household data are available-and illustrates this case using data from Turkey. Using Household Budget Survey 2018, the microsimulation model estimates the impact of labour income loss on household expenditures, considering that some types of jobs/sectors may be more vulnerable than others to the COVID-19 shock. Labour income loss is estimated to lead to reductions in monthly household expenditure using an income elasticity model, and expenditure-based child poverty is found to increase in Turkey by 4.9-9.3 percentage points (depending on shock severity) from a base level of 15.4%. Among the hypothetical cash transfer scenarios considered, the universal child grant for 0-17 years old children was found to have the highest child poverty reduction impact overall, while schemes targeting the bottom 20-30% of households are more cost-effective in terms of poverty reduction. The microsimulation model set out in this paper can be readily replicated in countries where similar Household Budget Surveys are available.Cette étude présente une méthodologie pour prédire l'impact de la COVID-19 sur la pauvreté infantile et que l’on peut aisément appliquer dans d'autres pays où des données similaires sur les ménages sont disponibles. L’étude illustre ce point en utilisant des données venant de Turquie. À l'aide de l'Enquête sur le budget des ménages de 2018, le modèle de microsimulation estime l'impact de la perte de revenus professionnels sur les dépenses des ménages, en prenant en compte le fait que certains types d'emplois/secteurs peuvent être plus vulnérables que d'autres au choc provoqué par la COVID-19. On estime que la perte de revenus professionnels entraîne des réductions au niveau des dépenses mensuelles des ménages, selon un modèle d'élasticité du revenu, et que la pauvreté infantile – sur la base des dépenses - augmente en Turquie entre 4,9 à 9,3 points de pourcentage (en fonction de la gravité du choc) à partir d'un niveau de base de 15,4 %. Parmi les scénarios hypothétiques de transferts monétaires envisagés, l'allocation universelle pour les enfants de 0 à 17 ans s'est avérée être la mesure qui a l'impact global le plus important sur la réduction de la pauvreté des enfants, tandis que les programmes ciblant les 20 à 30 % des ménages les plus pauvres sont les plus coût-efficaces en terme de réduction de la pauvreté. Le modèle de microsimulation présenté dans cet article peut être facilement reproduit dans les pays où des enquêtes similaires sur le budget des ménages sont disponibles.

Professional athletes are both highly paid and highly mobile workers. Previous research has shown that athletes respond to state income taxes differentials through bargaining and migration. If athletes are compensated for state tax burden, teams located in higher taxed states may be at a competitive disadvantage. I examine the effect of state income taxes on professional sports team performance. Using within-team variation in state top marginal income tax rates, I show that, only after the availability of free agency, did state income tax increases lower team winning percentages. I find that for each percentage point increase in state income tax rates, team winning declines by 0.70 percentage points.

OBJECTIVE: To measure the impact of Medicaid managed long‐term services and supports (MLTSS) on nursing home (NH) quality and rebalancing. DATA SOURCES/STUDY SETTING: This study analyzes secondary data from annual NH recertification surveys and the minimum dataset (MDS) in three states that implemented MLTSS: Massachusetts (2001–2007), Kansas and Ohio (2011–2017). STUDY DESIGN: We utilized a difference‐in‐difference approach comparing NHs in border counties of states that implemented MLTSS with a control group of NHs in neighboring border counties in states that did not implement MLTSS. Sensitivity analyses included a triple‐difference model (stratified by Medicaid payer mix) and a within‐state comparison. We examined changes in six NH‐level outcomes (percentage of low‐care NH residents, facility occupancy, and four NH quality measures) after MLTSS implementation. DATA COLLECTION/EXTRACTION METHODS: For each state, all freestanding NHs in border counties were included, as were NHs in neighboring counties located in other states. Information on low‐care residents was aggregated to the NH level from MDS data, then combined with Online Survey Certification and Reporting (OSCAR) and Certification and Survey Provider Enhanced Reporting (CASPER) data. PRINCIPAL FINDINGS: MLTSS had no statistically significant effects on NH quality outcomes in Massachusetts or Kansas. In Ohio, MLTSS led to an increase of 0.21 nursing hours per resident day [95% CI: 0.03, 0.40], and a decrease of 1.47 deficiencies [95% CI: −2.52, −0.42] and 9.38 deficiency points [95% CI: −18.53, −0.24] per certification survey. After MLTSS, occupancy decreased by 1.52 percentage points [95% CI: −2.92, −0.12] in Massachusetts, but increased by 3.17 percentage points [95% CI: 0.36, 5.99] in Ohio. We found no effect on low‐care residents in any state. Findings were moderately sensitive to the choice of comparator group. CONCLUSION: The study provides little evidence that MLTSS reduces quality of care, occupancy, or the percentage of low‐care residents in NHs.

The National Immunization Survey (NIS) is the United States' primary tool for assessing immunization coverage among 19- to 35-month- old children. Although annual estimates from the NIS are quite precise at the national level, US State-level estimates have much larger sampling error than national-level estimates. We combined two independent unbiased estimates of US State-level coverages within a given year to obtain new estimates which are more precise than previously published estimates. We first calculated a model-based estimate for each State for 2001 using multiple years of NIS data. Next, we combined each model-based estimate with the corresponding, previously reported NIS estimate for 2001. Our resulting estimates of State-level immunization coverage had smaller standard errors than the previously published estimates. To make similar improvements in precision by increasing sample size would, depending on State, require an increase in sample size of 30% - 120%. 1. Background Childhood immunizations in the United States have made rare many diseases that once caused enormous morbidity and mortality. Maintaining high rates of immunization in each new birth cohort is the key to continued success; coverage among infants and young children is monitored by the National Immunization Survey (NIS), which estimates coverage nationally, for the 51 US States (we treat the District of Columbia as a State, resulting in one more State than the usual 50), and for selected urban areas. The NIS produces annual estimates of immunization coverage for each geographic area considered. At the national level, annual NIS coverage estimates are quite precise, with standard errors smaller than 0.5 percentage points. At the level of US States and urban areas, which have smaller sample sizes than the national estimate, standard errors are of size up to 3 percentage points. However, precise estimates

Using data from Major League Baseball, I compute an objective measure of the home plate umpire's work quality- the accuracy of his ball and strike calls during a game- and measure how it varies with temperature. I find that an increase in game-time temperature from between 70F and 80F to above 95F decreases an umpire's accuracy by a little less than a percentage point, which is a 5.5% increase in the pitch-calling error rate when evaluated at the mean error rate of 13.3%. Restricting the sample to borderline pitches increases the magnitude of the hot-weather effect on accuracy to over a percentage point. My results indicate that very hot temperatures have a nontrivial, negative effect on the labor supply quality of a highly trained and highly skilled workforce in an important, high-revenue, and high-stakes industry, and suggest that protecting workers from daily variation in temperature can improve labor productivity.

BACKGROUND Medicare requires that hospitals report on their adherence to the Severe Sepsis and Septic Shock Early Management Bundle (SEP-1). OBJECTIVE To evaluate the effect of SEP-1 on treatment patterns and patient outcomes. DESIGN Longitudinal study of hospitals using repeated cross-sectional cohorts of patients. SETTING 11 hospitals within an integrated health system. PATIENTS 54 225 encounters between January 2013 and December 2017 for adults with sepsis who were hospitalized through the emergency department. INTERVENTION Onset of the SEP-1 reporting requirement in October 2015. MEASUREMENTS Changes in SEP-1-targeted processes, including antibiotic administration, lactate measurement, and fluid administration at 3 hours from sepsis onset; repeated lactate and vasopressor administration for hypotension within 6 hours of sepsis onset; and sepsis outcomes, including risk-adjusted intensive care unit (ICU) admission, in-hospital mortality, and home discharge among survivors. RESULTS Two years after its implementation, SEP-1 was associated with variable changes in process measures, with the greatest effect being an increase in lactate measurement within 3 hours of sepsis onset (absolute increase, 23.7 percentage points [95% CI, 20.7 to 26.7 percentage points]; P < 0.001). There were small increases in antibiotic administration (absolute increase, 4.7 percentage points [CI, 1.9 to 7.6 percentage points]; P = 0.001) and fluid administration of 30 mL/kg of body weight within 3 hours of sepsis onset (absolute increase, 3.4 percentage points [CI, 1.5 to 5.2 percentage points]; P < 0.001). There was no change in vasopressor administration. There was a small increase in ICU admissions (absolute increase, 2.0 percentage points [CI, 0 to 4.0 percentage points]; P = 0.055) and no changes in mortality (absolute change, 0.1 percentage points [CI, -0.9 to 1.1 percentage points]; P = 0.87) or discharge to home. LIMITATION Data are from a single health system. CONCLUSION Implementation of the SEP-1 mandatory reporting program was associated with variable changes in process measures, without improvements in clinical outcomes. Revising the measure may optimize its future effect. PRIMARY FUNDING SOURCE Agency for Healthcare Research and Quality.

Opioid misuse, overprescribing, dependency, and overdose remains a significant concern in the United States. A quality improvement study was conducted at the University of Illinois HospitalHealth Sciences System to determine the effect of standardizing the default orders for hydrocodone-acetaminophen products implemented on June 22, 2016.Prior to the intervention, default orders had variable dose tablet numbers (1 or 2) and dosing frequencies (every 4 or 6 hours), and no default dispense quantity. Defaults were modified to 1 tablet every 6 hours as needed for pain and dispense quantities of 3 and 5 days' supply were added. Number of tablets per order, dosing frequency, and days' supply prescribed between January 1, 2016, and June 21, 2016, were compared to those placed between June 22, 2016, and December 31, 2016. Opioid doses were converted into morphine milligram equivalents (MME). Analyses were performed to determine the effect of the intervention on daily opioid dose and number of days' supply prescribed.22,052 orders were included in this study. Following the intervention, the number of tablets prescribed was reduced by an average of 19,832 tablets per month. Every 6 hours dosing (as opposed to every 4 hours) increased by 21.52 percentage points. Prescriptions with ≥ 50 MME/day dropped by 5.8 percentage points, and3 days' supply decreased by 2.54 percentage points. Linear regression demonstrated an increase in opioid prescriptions with daily50 MME (odds ratio [OR] = 1.72, p0.001) and ≤ 3 days' supply (OR = 1.27, p0.001).Default electronic health record settings strongly influence prescribing patterns.

SummaryBackgroundCountries have adopted different approaches, at different times, to reduce the transmission of coronavirus disease 2019 (COVID-19). Cross-country comparison could indicate the relative efficacy of these approaches. We assess various non-pharmaceutical interventions (NPIs) over time, comparing the effects of self-imposed (i.e. voluntary) behavior change and of changes enforced via official regulations, by statistically examining their impacts on subsequent death rates in 13 European countries.Methods and findingsWe examine two types of NPI: the introduction of government-enforced closure policies over time; and self-imposed alteration of individual behaviors in response to awareness of the epidemic, in the period prior to regulations. Our proxy for the latter is Google mobility data, which captures voluntary behavior change when disease salience is sufficiently high. The primary outcome variable is the rate of change in COVID-19 fatalities per day, 16-20 days after interventions take place. Linear multivariate regression analysis is used to evaluate impacts. Voluntarily reduced mobility, occurring prior to government policies, decreases the percent change in deaths per day by 9.2 percentage points (95% CI 4.5-14.0 pp). Government closure policies decrease the percent change in deaths per day by 14.0 percentage points (95% CI 10.8-17.2 pp). Disaggregating government policies, the most beneficial are intercity travel restrictions, cancelling public events, and closing non-essential workplaces. Other sub-components, such as closing schools and imposing stay-at-home rules, show smaller and statistically insignificant impacts.ConclusionsThis study shows that NPIs have substantially reduced fatalities arising from COVID-19. Importantly, the effect of voluntary behavior change is of the same order of magnitude as government-mandated regulations. These findings, including the substantial variation across dimensions of closure, have implications for the phased withdrawal of government policies as the epidemic recedes, and for the possible reimposition of regulations if a second wave occurs, especially given the substantial economic and human welfare consequences of maintaining lockdowns.

OBJECTIVE: Between January 2005 and July 2020, 171 rural hospitals closed across the United States. Little is known about the extent that other providers step in to fill the potential reduction in access from a rural hospital closure. The objective of this analysis is to evaluate the trends of Federally Qualified Health Centers (FQHCs) and Rural Health Clinics (RHCs) in rural areas prior to and following hospital closure. DATA SOURCES/STUDY SETTING: We used publicly available data from Centers for Medicare and Medicaid Provider of Services files, Cecil G. Sheps Center rural hospital closures list, and Small Area Income and Poverty Estimates. STUDY DESIGN: We described the trends over time in the number of hospitals, hospital closures, FQHC sites, and RHCs in rural and urban ZIP codes, 2006–2018. We used two‐way fixed effects and pooled generalized linear models with a logit link to estimate the probabilities of having any RHC and any FQHC within 10 straight‐line miles. DATA COLLECTION/EXTRACTION METHODS: Not applicable. PRINCIPAL FINDINGS: Compared to hospitals that never closed, the predicted probability of having any FQHC within 10 miles increased post closure by 5.95 and 11.57 percentage points at 1 year and 5 years, respectively (p

Using daily data, we estimate a vector autoregression model to characterise the dynamic relationship between COVID‐19 infections in Australia and the performance of the Australian stock market, specifically the ASX‐200. Impulse response functions show that COVID‐19 infections in Australia have a significant positive effect on the performance of the stock market: a one standard deviation increase in new registered cases of COVID‐19 infections in Australia increases the daily growth rate of the ASX‐200 by around half a percentage point. This result is robust to alternative lag selections of the VAR model as suggested by alternative information criteria, including in the model control variables for stock market volatility, that is the ASX‐200 VIX; the USD‐AUD exchange rate and the international oil price; news by the World Health Organization regarding a COVID‐19 pandemic and public health emergency; and the government‐imposed shutdown of parts of the Australian economy. We also present estimates of the dynamic relationship between the daily growth rate of the Dow Jones and daily new cases of COVID‐19 infections in the United States. The US data show, similar to the Australian data, that there is a significant positive effect of COVID‐19 infections on the performance of the stock market.

This paper uses an econometric approach to examine the inflation consequences of the American Rescue Plan Act of 2021. Price equations are estimated and used to forecast future inflation. The main results are: (1) The data suggest that price equations should be specified in level form rather than in first or second difference form. (2) There is some slight evidence of nonlinear demand effects on prices. (3) There is no evidence that demand effects have gotten smaller over time. 4) The stimulus from the act combined with large wealth effects from past household saving, rising stock prices, and rising housing prices is large and is forecast to drive the unemployment rate down to below 3.5 percent by the middle of 2022. 5) Given this stimulus, the inflation rate is forecast to rise to slightly under 5 percent by the middle of 2022 and then comes down slowly. 6) There is considerable uncertainty in the point forecasts, especially two years out. The probability that inflation will be larger than 6 percent next year is estimated to be 31.6 percent. 7) If the Fed were behaving as historically estimated, it would raise the interest rate to about 3 percent by the end of 2021 and 3.5 percent by the end of 2022 according to the forecast. This would lower inflation, although slowly. By the middle of 2022 inflation would be about 1 percentage point lower. The unemployment rate would be 0.5 percentage points higher.

Objectives. To understand how stay-at-home orders changed mobility patterns and influenced the spread of COVID-19. Methods. I merged 2020 data from the Virginia Department of Health, Google Mobility Reports, and the US Census to estimate a series of 2-way fixed-effect event-study regression models. Results. A stay-at-home order caused people to increase the amount of time spent at home by 12 percentage points and decrease the time the spent at work by 30 percentage points, retail and recreation venues by 40 percentage points, and grocery stores and pharmacies by 10 percentage points. People did not sustain changes in mobility and gradually returned to prepandemic levels before the stay-at-home order was lifted. In areas where people spent the most time at indoor locations, there was a large increase in COVID-19. Conclusions. A more robust and stricter policy response coordinated at the national level combined with a strong economic response from policymakers could have increased the effectiveness of the stay-at-home order.

We study how changes in return migration patterns over the last decade impacted self-employment rates in Mexico. Using state historical migration rates as instruments, we calculate elasticities of self-employment to return migration of 0.093, 0.156, and 0.236 for all workers, males, and low-educated male workers, respectively. Back-of-the-envelope calculations indicate that the 0.18 percentage point decline in the return migration rates observed between 2006 and 2016 reduced the self-employment rate by 5.8 percent relative to the level seen in 2006. Additionally, we estimate the gains from migration and self-employment among Mexican return migrants controlling for observable and unobservable characteristics. While return migrants have earnings 1.72 log points higher than non-migrant Mexican workers, we do not find differences in the earnings of self-employed and wage-employed return migrants once we control for unobservable skills.

This study estimates the overall effect of two influenza vaccination programs consecutively administered in a cluster-randomized trial in western Senegal over the course of two influenza seasons from 2009 to 2011. We apply cutting-edge methodology combining social contact data with infection data to reduce bias in estimation arising from contamination between clusters. Our time-varying estimates reveal a reduction in seasonal influenza from the intervention and a non-significant increase in H1N1 pandemic influenza. We estimate an additive change in overall cumulative incidence (which was 6.13% in the control arm) of -0.68 percentage points during Year 1 of the study (95% CI: −2.53, 1.18). When H1N1 pandemic infections were excluded from analysis, the estimated change was −1.45 percentage points and was significant (95% CI, −2.81, −0.08). Because cross-cluster contamination was low (0–3% of contacts for most villages), an estimator assuming no contamination was only slightly attenuated (−0.65 percentage points). These findings are encouraging for studies carefully designed to minimize spillover. Further work is needed to estimate contamination – and its effect on estimation – in a variety of settings.

BACKGROUND Each year, about 10% of Medicare Advantage (MA) enrollees voluntarily switch to another MA contract, while another 2% voluntarily switch from MA to fee-for-service Medicare. Voluntary disenrollment from MA plans is related to beneficiaries' negative experiences with their plan, disrupts the continuity of care, and conflicts with goals to reduce Medicare costs. Little is known about racial/ethnic disparities in voluntary disenrollment from MA plans. OBJECTIVE The objective of this study was to investigate differences in rates of voluntary disenrollment from MA plans by race/ethnicity. SUBJECTS A total of 116,770,319 beneficiaries enrolled in 736 MA plans in 2015. METHODS Differences in rates of disenrollment across racial/ethnic groups [Asian or Pacific Islander (API), Black, Hispanic, and White] were summarized using 4 types of logistic regression models: adjusted and unadjusted models estimating overall differences and adjusted and unadjusted models estimating within-plan differences. Unadjusted overall models included only racial/ethnic group probabilities as predictors. Adjusted overall models added age, sex, dual eligibility, disability, and state of residence as control variables. Between-plan differences were estimated by subtracting within-plan differences from overall differences. RESULTS Adjusted rates of disenrollment were significantly (P