Book

# Mostly Harmless Econometrics: An Empiricist's Companion

Authors:
... 12 Since our variable of interest is at the country×year level, we cluster the standard errors at this level. However, with only 66 clusters, we may have too few clusters to get unbiased standard errors (Angrist and Pischke, 2008). Following Cameron et al. (2008) and Cameron and Trivedi (2010), a solution for this problem is to further bootstrap the standard errors. ...
... They perform Monte-Carlo simulations indicating that the generalized tobit model we use successfully corrects the selection bias. 12 Linear probability and logit models produce almost similar marginal effects when the average probability is around 50% (Angrist and Pischke, 2008). 13 This mainly concerns the biggest ones, called Déclarants Directs Généraux. ...
... A general drawback of the linear probability model is the possibility for predicted probabilities to lie beyond the [0;1] interval. This should not be a concern here since we are not interested in producing any predictions, but focus on the average marginal effects (Angrist and Pischke, 2008). Furthermore, Wooldridge (2001, chap. ...
Thesis
Today’s developed economies are often described as service economies. More than two thirds of employment and value added is generated by the service sector in OECD countries. Services are increasingly important in today’s knowledge based economies, are a crucial component of economic growth and contribute to the competitiveness of the industrial sector (Nord ås and Kim, 2013). Figure IV.9 plots the evolution of value added and employment in France between 1970 and 2007. It shows that the professional service industries (Real estate/renting/business services and ﬁnancial intermediation) are the main contributors to the growth of the French economy. These services (also called “complementary services” by Katouzian (1970)), have been growing much faster than the manufacturing sector, and much faster than the other service sectors (wholesale/retail, hotels and restaurant services). These fast growing services accounted for 33% of the total value added in 2007 (twice as much as the manufacturing sector) and 20% of the overall employment (14% for the manufacturing sector). [...]
... Several reasons lead us to stick with the LPM (Linear Probability Model) as a baseline. Among others, Angrist and Pischke (2009) advocate the use of the LPM. Nonlinear estimation methods may provide an efficiency gain, but at the cost to commit to a precise distributional assumption of the error term and, notably, Probit and Logit average marginal effect estimates, quite often, do not differ much from LPM estimates and the interpretation of the regression coefficients is much more straightforward with the LPM. ...
... We prefer to stick to an IV procedure. By doing so, the validity of our estimates does not rely on any assumption concerning the distribution of the error term, see Angrist and Pischke (2009). 26 Further research may build new instruments developing measures of spatial competition for each degree program, see Bratti et al. (2021). ...
... 27 When using this instrument, one may be prone to suggest to run a Probit in place of an OLS in the first stage. Angrist and Pischke (2009) and Wooldridge (2010) shows that this procedure would be incorrect, namely we would run a kind of forbidden regression. Differently, another feasible alternative would be a bivariate probit. ...
Chapter
Full-text available
Reforming governance in higher education has been a kind of mantra that has characterised governmental policies worldwide. Under the pressure of massification, globalisation and socio-economic demands, governments have continuously intervened to redesign the characteristics of the governance arrangements of their higher education systems as well as institutional governance. This common effort has been characterised by the adoption of a common template (i.e. the ‘steering at a distance’ model), mainly based on the idea of making universities more accountable to the societal goals through the massive use of evaluation, assessment and monitoring. The final results are highly differentiated, owing to the fact that each country has implemented a common template according to its own national characteristics and legacies. In this context, the Italian case shows its own peculiarities, whereas evaluative tools have been significantly adopted in a design highly contradictory of other dimensions such as institutional governance, the rules of careers and academic recruitment and the lack of clear systemic goals to be reached.
... Our empirical identification strategy, regression discontinuity, does not require control variables to ensure the internal validity of the coefficient estimates (Angrist & Pischke, 2009). However, the inclusion of control variables that correlate with the outcomes can increase the statistical efficiency of the estimator. ...
... Intuitively, regression discontinuity design separates subjects with otherwise similar attributes into treatment and control groups based upon a decision rule outside of the control of the subjects. The rule that determines whether a subject receives treatment is based upon an assignment variable that is exogenous to the subject as opposed to the random placement of subjects as in experimental design (Angrist & Pischke, 2009;Murnane & Willett, 2011). ...
... At the bowl eligibility threshold, the probability of bowl participation jumps from 0.18 to 0.76. A fuzzy regression discontinuity design can be employed when an exogenous assignment rule is highly correlated with the actual treatment status but does not fully explain the treatment (Angrist & Pischke, 2009;Murnane & Willett, 2011). The fuzzy regression discontinuity design employs a two-stage procedure to estimate the causal impact of a policy. ...
Article
Full-text available
This study presents the results of a regression discontinuity empirical approach to investigate the effects of postseason bowl game participation on student-athlete academic outcomes and subsequent football team success. The practice expectations for student-athletes on football teams that participate in a bowl game increase by between two and four weeks relative to student-athletes on teams that do not participate in a bowl game. Prior research has been inconclusive on whether this increased practice intensity is associated with academic or athletic outcomes. The sample includes 130 NCAA football bowl subdivision (FBS) teams between the years 2003 through 2018. We apply a fuzzy regression discontinuity design by exploiting the fact that teams in the NCAA FBS become eligible to participate in a bowl game when their regular season winning percentage is greater than 0.50. The results suggest that bowl game participation increased the team’s eligibility rate by 0.8 percentage points, the team’s Academic Progress Rate by 4.6 points, but had no effect on the team’s retention rate. Bowl game participation was not found to affect the subsequent year’s winning percentage or likelihood of bowl game participation. Athletic programs that are undecided about whether the costs, in finances or time, of participating in a bowl game are worthwhile might benefit from these findings. In particular, the results reveal that bowl game participation does not come as a detriment to the academic outcomes of their student-athletes.
... These variables are usually viewed as intermediate variables (mediators), meaning that the slowdown of the economy is supposed to cause an improvement in health conditions exactly by inducing a modification in people's lifestyle. Given their mediating nature, lifestyle factors represent what Angrist and Pischke (2009) call a "bad control" and should accordingly be excluded from the model (see on that also Pearl, 2009). ...
... In the case of the example shown in Fig. 1, the estimation of Eq. (2) through OLS resolves in the estimation of a saturated model. This notwithstanding it can be shown that the parameter is an estimate of the ATT (Angrist & Pischke, 2009;Cunningham, 2021). It should be noted, in this regard, that we cannot exclude the possibility that at least some of the countries in our control group suffered a mild form of recession. ...
Article
Full-text available
Some European countries, such as Greece and Spain, were severely hit by the 2008 economic crisis whereas others, such as Germany, were practically spared by it. This divergence allowed us to implement a difference in differences research design which offered the possibility to observe the long-lasting effects produced by the crisis on European life expectancy. Our analysis-based on Eurostat data from 2001 to 2019-shows that life expectancy increased faster, after the onset of the crisis, in those countries where the rise in unemployment was more intense. Furthermore, our results show that this gain in life expectancy persisted, and sometimes further increased, until 2019 when most macroeconomic variables had returned to their pre-crisis values. Previous research has identified that mortality behaves procycli-cally in developed countries: when the economy slows down mortality decreases and vice versa. Our findings show, by contrast, that life expectancy behaves asymmetrically: it responded to an increase but not to a decrease in unemployment. This calls for a reconsideration of the causal mechanisms linking together the economic cycle and mortality in developed countries.
... First, we compute the ratios of compliers in treatment group and control group, following Angrist and Pischke (2008). For treatment group, the ratio of compliers is computed as ...
... In our sample, compliers account for 34.81% in treatment group, and 11.56% in control group. Compared with references listed in Table 4.4.2 of Angrist and Pischke (2008), the ratios are high for our sample, implying that our instrumental variable is indeed a strong instrument for grandparental childcare. ...
Article
Full-text available
This paper considers the role of grandparental childcare in explaining China’s extraordinarily high female labor market participation rate and low wage penalty. Using a novel and high-quality dataset combined with a creative new identification strategy, we find that grandparental childcare reduces young women’s drop-out from the labor market, especially those with a higher education level living in an urban area, and it also improves mothers’ labor income. We further show that grandparents’ marriage status and home location do not affect the feasibility of grandparental childcare. Our research reveals that grandparental childcare, as a remedy for insufficient supply of public childcare services, supports mothers in the labor market, sustaining high mobility in the labor market in China.
... 6 Removing these observations from the dataset would bias the results, as the districts in which one candidate drops out might be systematically different from others on some variables confounding turnout. Hence, we adopt a fuzzy RDD approach, which is standard in situations where the threshold does not perfectly determine the treatment status (Angrist & Pischke, 2008;Cattaneo et al., 2019). In a series of robustness tests, we also use a wide range of alternative approaches including simple OLS regressions with district and election fixed effects. ...
... In a fuzzy RDD, the threshold creates a discontinuity in the probability of receiving the treatment. Following the standard approach in the literature (Angrist & Pischke, 2008;Cattaneo et al., 2019), the discontinuity then becomes an instrumental variable for the treatment status. In short, fuzzy RDD is a weighted two-stage least squares (2SLS) estimation. ...
Article
Full-text available
A long-lasting question in comparative politics is whether the number of candidates/parties increases turnout. Existing observational studies on the topic find mixed results. We thus apply a regression discontinuity design to data 13,910 legislative and cantonal electoral districts in France since 1978. In the two-round system used in these elections, the candidates who pass a certain vote threshold in the first round can participate in the second round. We use this discontinuity to estimate the causal effect of having a third candidate in the second round: it increases turnout by 3.5% points and the share of valid votes by 7.3% points. We confirm these findings with survey data from the 2012 legislative election. Further, we investigate the mechanism and find evidence supporting the alienation theory, according to which individuals whose preferences do not resonate with the preferences of any of the candidates are likely to abstain.
... coefficients that are easier to interpret than those produced by logit models (see Brands & Fernandez-Mateo, 2017;Sorenson & Waguespack, 2006 for similar procedures). Given that LPM imposes heteroskedasticity on the errors, we used robust estimates of the standard errors and applied two-way clustering for year and D.C. Circuit Judge to account for the non-independence of the observations (Angrist & Pischke, 2009). We also estimated logit models with fixed effects ...
... Finally, we find a negative main effect of Judge Tenure on SCOTUS Clerk ( =.009, p=.001), indicating that candidates recommended by judges with greater tenure are associated with poorer outcomes. x Female Candidate three-way interaction term, which explains more variance and shows improved fit compared to Models 1 and 2. Following best practices, we focus our interpretation and discussion on the highest-order interaction (Angrist & Pischke, 2009;Wooldridge, 2013), Finally, an intriguing and unexpected discovery is that increased tenure is negatively associated with the outcomes of male and female candidates recommended by male judges in comparison to female judges (see three-way interaction plotted in Figure 1). This suggests that not only does the efficacy of male judges' recommendations not improve with increased tenure, but higher levels of judge tenure are associated with negative outcomes for candidates recommended by male judges. ...
Article
The critical role that referrals play in the hiring process, particularly for candidates contending with negative stereotypes and biases, is well documented. However, how those stereotypes and biases impact sponsors, and the effectiveness of the referrals that they provide, is not well understood. Drawing on evidence of reversals of gender bias, we explore the impact of sponsors’ gender and tenure on the effectiveness of their referrals in the context of U.S. Supreme Court law clerk hiring decisions. This is an appropriate setting because success in the application process for these elite early career positions is contingent on having a strong recommendation from a judge with which the candidate has previously worked, making it ideal to study gender differences in the effectiveness of referrals. Analyses show candidates recommended by male sponsors are more likely to be hired compared to those recommended by female sponsors overall, but this dynamic is also dependent on the sponsor’s tenure and the candidate’s gender. For female sponsors, higher levels of tenure are associated with better hiring outcomes for their female candidates only. All other gender combinations do not benefit from sponsor seniority. Possible mechanisms, limitations, and implications for future research directions are discussed.
... DID. To mitigate endogeneity concerns related to reverse causality and to assess the causal relationship between negative media coverage of CSiR and CSP more fully, following Angrist and Pischke (2008), I implemented a DID design to test whether or how the presence of negative news of CSiR may affect an organization's CSP. In other words, to test the causality implied by the first hypothesis, I examined the change in firms' CSP in response to a notable change in media coverage, using a DID estimator. ...
... I coded the variable "Post" as "0" for the years before the shock of negative media coverage of CSiR and "1" for the year when the negative media coverage of CSiR occurred and the subsequent 3 years. Following Angrist and Pischke (2008), I implemented a DID estimator on the following model 4 : ...
Article
Full-text available
Can bad news also be good news? In this study, I explicate why bad news about firms’ corporate social irresponsibility (CSiR) can be good news for firms. Specifically, I address the role of negative media coverage of CSiR in firms’ corporate social performance (CSP). Drawing on signaling theory, I propose that negative media coverage of CSiR is a form of costly yet effective external feedback to firms’ current social signaling. It, therefore, propels firms to undertake organizational changes to send positive response signals through improved CSP. Furthermore, I argue that this effect is augmented by organizational innovation search, which influences firms’ learning capacity required to improve firms’ CSP. Using a multicountry sample of 1,049 firms between 2007 and 2016, I find that negative media coverage of CSiR induces firms to enhance CSP, and this effect is moderated by organizational innovation search.
... weights can lead to well-known issues in efficiency (Angrist & Pischke, 2009) and by using unweighted estimates our sampling variance is reduced and the precision of our estimates is increased. ...
Article
Full-text available
A top priority of prison authorities is maintaining a safe and orderly institutional environment. Gangs are believed to impede this objective, warranting bespoke policies and practices. Drawing on the process-based model of regulation, we depart from orthodox explanations for the gang-misconduct link and argue that gang affiliates are treated less fairly than non-gang affiliates owing to suppression-oriented administrative policies and harsher day-to-day interactions with officers, which, in turn, impact compliance. We use administrative and survey data sources based on a sample of 802 male prisoners and generalized structural equation modeling to examine whether procedural justice and legal orientations mediate the association between official classification of gang affiliation and self-reported misconduct. Our findings reveal partial support for the process-based model: procedural justice and legitimacy are poorer among gang than non-gang respondents but do not mediate the gang affiliation-misconduct link. The traditional pathway between procedural justice, legitimacy, and obligation to obey was observed, none of which were related to misconduct, which stands in sharp contrast to expectations of the process-based model. These findings suggest that factors other than procedural justice and legal orientations may be more relevant for rule violations among gangs, specifically, and within correctional environments, generally.
... Very often this work is done via regression analysis, with various methods for identifying the effect of interest. A good reference for the philosophical approach to identification used by contemporary economics researchers can be found in Angrist and Pischke (2008). Given the prevalence of this approach within economics, we will spend little time on it here. ...
Chapter
Food systems and food networks have gained increased attention among agricultural and applied economists in recent years. This chapter presents a new topic that has not been included in previous Handbook chapters, aiming to provide a broad review of the research related to food systems and food networks to guide and motivate agricultural and applied economists new to this space in conducting research on these topics. Many of our examples come from local and regional food systems and food networks as they are both of wide interest globally, and because they represent a growing volume of expertise among researchers. It would be a daunting task attempting to include all literature in this domain given the complexity of research subjects by agricultural and applied economists and scholars from other fields. However, the frameworks and approaches we discuss can be applied to food systems and food networks at a variety of scales and in a variety of contexts. Unsurprisingly, definitions of food systems and food networks are not fully agreed upon by scholars. In this chapter, we begin with discussing various definitions of food systems and food networks; the types of research questions one might want to answer related to food systems and food networks; and the theories, frameworks, data, and methods used to study and answer these questions. This chapter provides examples of the interdisciplinary nature of this work including: (1) examinations of components of food systems versus food networks, (2) characteristics and interactions of actors in building and maintaining food systems and food networks, and (3) integrated systematic approaches to understanding and enhancing food systems and food networks to address societal problems. Our chapter concludes with a discussion of the opportunities and challenges we see for agricultural and applied economists in this field moving forward. Through this work, we hope to stimulate innovative approaches among agricultural and applied economists, working with other disciplines, to explore and analyze relationships, influences, and impacts of food systems and food networks to enhance social, economic, and environmental sustainability and equity.
... As noted by EEF guidance, in a model that does not account for clustering, when this is a feature introduced by the experimental design, 'the point estimates will be accurate, but the standard errors will be downward biased' (EEF, 2018, p.3). However, we accounted for the potential effects of the experimental design in this respect by calculating standard errors, taking into account clustering (Angrist & Pischke, 2008) at the school level, which allows for the correlation of pupil outcomes within schools. We prefer this to the use of a hierarchical linear model which makes additional assumptions about the school-level effects that may not be justified. ...
... Therefore, this study draws on (Areliano and Bond (1991) method of lagged dependent variable model used in the cross country study by Li et al. (2017a). Fixed effect and lagged dependent variable estimates present a useful bracketing property (Angrist and Pischke, 2009). In case when the lagged dependent variable model is the correct one but mistakenly fixed effects model has been used; estimates of the positive treatment effect will tend to be too big. ...
Article
This study envisages to investigate the hurdles in widespread usage of Electric-Vehicles(EVs) on a global scale by using cross-country panel data. This paper empirically studies the impact of wide range of parameters like policies & incentives, national commitments, socioeconomic factors, charger infrastructure, fuel price and renewables on cross-country EV demand by using two-way Fixed Effect Model. Policies like setting up of EV and charger targets, providing fiscal subsidies and sociodemographic factor of age and Gross Domestic Product/capita, higher renewable energy production and higher number of chargers have a positive impact on EV demand. High Gasoline price has a positive effect on Battery EVs demand while regions of lower population density like non-urban areas are more likely to adopt EVs. Also, national commitments to initiatives like [email protected] is observed to have mostly positive impact across different factors suggesting that green initiatives positively impact the national EV demands. Based on the obtained results, relevant policy implications are discussed.
... As noted by EEF guidance, in a model that does not account for clustering, when this is a feature introduced by the experimental design, 'the point estimates will be accurate, but the standard errors will be downward biased' (EEF, 2018, p.3). However, we accounted for the potential effects of the experimental design in this respect by calculating standard errors, taking into account clustering (Angrist & Pischke, 2008) at the school level, which allows for correlation of pupil outcomes within schools. We prefer this to the use of a hierarchical linear model which makes additional assumptions about the school-level effects that may not be justified. ...
... In the regression analysis, the dependent variable is the forecast share of Leave votes, ranging from 0 to 100. To increase the precision of the estimates (see Angrist & Pischke, 2009;Kam & Trussler, 2017), in Appendix C, we estimate average treatment effects controlling for the same set of covariates used in the observational analysis. The covariate-adjusted estimates are substantially the same as the non-adjusted estimates. ...
Article
Full-text available
The emergent literature on citizen forecasting suggests that the public, in the aggregate, can often accurately predict the outcomes of elections. However, it is not clear how citizens form judgments about election results or what factors influence individual predictions. Drawing on an original survey experiment conducted during the campaign for the United Kingdom’s Brexit referendum, we provide novel evidence of what influences citizen forecasts in a so-far unexplored context of direct democracy. Specifically, we investigate the effect of voting preferences and political sophistication, in addition to three “exogenous factors” that we manipulate experimentally—i.e., social cues, elite cues and campaign arguments. Our findings indicate that citizens are reasonably accurate in their predictions, with the average forecast being close to the actual result of the referendum. However, important individual heterogeneity exists, with politically sophisticated voters being more accurate in their predictions and less prone to wishful thinking than non-sophisticated voters. Experimental findings show that partisan voters adjust their predictions in response to cues provided by their favorite party’s elites and partly in response to campaign arguments, and the effects are larger for low-sophisticated voters. We discuss the mechanisms accounting for the experimental effects, in addition to the implications of our findings for public opinion research and the literature on citizen forecasting.
... A change in statistical significance clearly indicates an issue across waves due to lost participants. A change in the coefficient of 20% or more, which is the same value used for the standardized bias value in propensity score matching (Angrist & Pischke, 2009;Rosenbaum & Rubin, 1985), is a useful proxy for identifying issues. In our empirical demonstration, the primary variables of interest changed in statistical significance when the change in the coefficients were all greater than 20% with the smallest of these changes being a 22% change. ...
Article
Full-text available
Research on missing data in panel datasets has focused on attrition where respondents drop out and do not return. An equally important, but understudied, form of missing data include missed interviews where respondents contribute different number of total interviews to a panel dataset. Because individuals with crime-related characteristics miss more interviews, conditioning one’s sample on different number of waves changes the composition of the sample, and likely the subsequent conclusions. Scholars must weigh the balancing act of wanting a sample with more waves of data to tease out their panel process against the concern that they will lose individuals sensitive to the process under study by requiring too many waves of data. Using the Pathways to Desistance study, a panel dataset of youth who have committed serious offenses, we compared sample characteristics across multiple conditioned samples to unpack how the characteristics of one’s sample changes as more waves are required to be entered into the sample. We further demonstrate the implications of this in an applied setting by examining the relationship between residential mobility and perceptions of informal social costs. Our results indicate that the characteristics of one’s panel sample are sensitive to the number of waves one conditions their sample upon. This is especially prevalent for offending where those who contribute fewer waves of data consistently report higher levels of offending. In our empirical demonstration, substantive conclusions change across conditioned samples. Our study provides unique insight into an understudied phenomenon while also providing practical advice to panel dataset scholars.
... Pertama, penulis menggunakan pendekatan deskriptif murni (pure descriptive research) untuk menjawab pertanyaan riset pertama. Penelitian deskriptif murni memiliki peran penting dalam pengambilan kebijakan-meskipun bagi sebagian peneliti, hal yang mungkin paling menarik dalam ilmu sosial adalah tentang hubungan sebab dan akibat (Angrist & Pischke, 2009). ...
Article
There is a great deal of consensus that in addition to cause adversarial impact on the health sector, Covid-19 pandemic has also caused economic disruption on an unprecendeted scale. However, empirical studies that specifically scrutinise the economic impact of this pandemic in Indonesia are still limited. This study aims to provide a more detailed picture of the economic disruption experienced by Indonesian businesses in the time of the pandemic. Analysing survey data from 12,361 respondents, this paper shows in greater detail the effect of Covid-19 shocks on sales, operating expenses, business capacity, business operational difficulties and status, including marketing and labor-related strategies taken by businesses to maintain their business activities. This study also further analyses the implications of variations in the relationship between business operational difficulties, marketing strategies, and labor-related strategies with the annual turnover and the location of the businesses to show heterogeneity in its impact.
... To employ our instrumental-variable approach, we use a procedure common in Bayesian estimation where we correlate the endogenous equations with the relevant first-stage equations (e.g., Ataman et al., 2008), for a system of nine equations per console generation (one first-stage equation and two endogenous equations per console) in the main model and in the quality and age moderations. Because we have two endogenous variables, we can test for the strength of the instruments with a pooled Angrist-Pischke F-test (Angrist & Pischke, 2009). The results indicate that the instruments are sufficiently strong (p < .05). ...
Article
Full-text available
Steady software supply is a crucial driver of platform sales. While publishers benefit from releasing software across multiple platforms to tap a greater market, platform manufacturers often seek exclusive release to differentiate from competitors.Research has examined such software multihoming across competing platforms of the same technology generation (i.e., the proximal market); however, publishers increasingly multihome software to platforms in distal markets. In the video game console industry, these include previous-generation consoles, handhelds, or mobile devices. This study investigates multihoming to distal markets in the seventh and eighth game console generations. Whereas multihoming to previous-generation consoles cannibalizes focal console sales, multihoming to mobile devices exerts complementary effects. Software quality and console age moderate these relationships, with negative spillovers from multihoming to previous-generation consoles being rooted in lower-quality games and games released later in the console’s lifecycle. By contrast, multihoming to mobile devices is most beneficial early on.
... Eq. (1) below is our baseline model. Within applied econometrics, there is an ongoing debate whether ordinal categorical dependent variables can be estimated through linear regression models (Angrist & Pischke, 2009;Kent, 2020;Winship & Mare, 1984). Employing OLS models on ordered outcomes may lead to a violation of the assumption of independent, identically distributed errors, which can be partly remedied by using robust standard errors (Wooldridge, 2013, p. 294). ...
Article
The livelihoods of rural populations in Africa are closely tied to small-scale farming. In recent years, private investors as well as governments have shown a growing interest in large-scale acquisition of arable land across the continent. While researchers have started to analyze the local economic and envinronmental impacts of such investments, their socio-political as well as psychological consequences remain poorly understood. This paper investigates how changes in land ownership patterns caused by large-scale land acquisitions affect the level of interpersonal trust among rural communities. We main- tain that the transition from community and individual-smallholder land ownership into large-scale investor property has a negative impact on local levels of trust. Furthermore, we assume that the deterioration of trust caused by large-scale land investments is stronger among women than men. To test our claims, we connect circa 71,000 respondents from Afrobarometer surveys to georeferenced information on the location of land deals from 33 African countries. Relying on a difference-in-differences type of empirical strategy as well as an instrumental variable approach, we show that large-scale land investments indeed disrupt local social fabrics by reducing interpersonal trust. Our results suggest that trust in relatives is particularly affected by large-scale land acquisitions. In addition, we find that land deals reduce personalized trust among women but not necessarily among men.
... We choose to use the OLS method to estimate the impact of informal employment on residents' well-being mainly for two reasons. One is that many studies have confirmed that the parameter estimates of OLS regression and Oprobit regression are consistent in direction and significance [39][40][41]; second, the estimated value obtained by OLS has stronger intuitive explanatory power than Oprobit regression. When using the OLS model to estimate the impact of informal employment on residents' well-being, we also provide Oprobit estimation results to verify the reliability of the OLS estimation results. ...
Article
Full-text available
The influence of informal employment on residents’ happiness has gained wide attention around the world. However, few studies focus on this topic in China. Using the 2016 wave of the China Labor Force Dynamics Survey (CLDS) data, we examined the effect of informal employment and its mechanisms on residents’ happiness in China. Our study shows there is a significant negative correlation between informal employment and residents’ happiness in China. Moreover, the correlation between informal employment and residents’ happiness is stronger for residents who are female, migrating, and with a rural household registration. In addition, we investigated possible mechanisms of the effect, including individual income, social respect, unemployment expectations, and social security, and found that informal employment reduces the happiness of residents by widening the gap in unemployment probability and social insurance level among residents.
... T k;i;j ¼ 0� (independence of runner ability and treatment status), thus the selection bias term in Eq (3) cancels. For more discussion on this see [15]. ...
Article
Full-text available
Shorter distance events in track and field are replete with folk tales about which lane assignments on the track are advantageous. Estimating the causal effect of lane assignments on race times is a difficult task as lane assignments are typically non-random. To estimate these effects I exploit a random assignment rule for the first round of races in short distance events. Using twenty years of data from the IAAF world athletic championships and U20 world championships, there is no evidence of lane advantages in the 100m. Contrary to popular belief, the data suggest that outside lanes in the 200m and 400m produce faster race times. In the 800m, which is unique in having a lane break, there is some weak evidence that outside lanes producer slower race times, possibly reflecting the advantage of inside lanes having an established position on the track at the lane break. Given that these results do not support common convictions on lane advantages, they also serve as an interesting case study on false beliefs.
... In fact, a major problem in assessing the impact of any policy, and tracing the causal relationship between a set of government decisions and their outcomes, is that of identifying appropriate counterfactuals and avoiding the risks of selection bias. Selection biases are common to both qualitative and quantitative research designs, because they simply reflect the potentially relevant differences between the compared units receiving and not receiving the treatment, such as implementing or not implementing a certain policy (Angrist and Pischke 2009;King, Keohane, and Verba 1994). ...
... As a real-world application, we consider the National Supported Work (NSW) program. Starting with a seminal contribution by Lalonde [24], studies have used this Randomized Control Trial to benchmark nonexperimental techniques-including an historical 'face-off between regression and propensity-score matching' [25]. Observational methods generally fail to recover the experimental causal effect estimate, except in a smaller handpicked NSW subsample [10,[26][27][28]. ...
Article
Full-text available
We propose a new method to estimate causal effects from nonexperimental data. Each pair of sample units is first associated with a stochastic ‘treatment’—differences in factors between units—and an effect—a resultant outcome difference. It is then proposed that all pairs can be combined to provide more accurate estimates of causal effects in nonexperimental data, provided a statistical model relating combinatorial properties of treatments to the accuracy and unbiasedness of their effects. The article introduces one such model and a Bayesian approach to combine the O(n2)\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$O(n^2)$$\end{document} pairwise observations typically available in nonexperimental data. This also leads to an interpretation of nonexperimental datasets as incomplete, or noisy, versions of ideal factorial experimental designs. This approach to causal effect estimation has several advantages: (1) it expands the number of observations, converting thousands of individuals into millions of observational treatments; (2) starting with treatments closest to the experimental ideal, it identifies noncausal variables that can be ignored in the future, making estimation easier in each subsequent iteration while departing minimally from experiment-like conditions; (3) it recovers individual causal effects in heterogeneous populations. We evaluate the method in simulations and the National Supported Work (NSW) program, an intensively studied program whose effects are known from randomized field experiments. We demonstrate that the proposed approach recovers causal effects in common NSW samples, as well as in arbitrary subpopulations and an order-of-magnitude larger supersample with the entire national program data, outperforming Statistical, Econometrics and Machine Learning estimators in all cases. As a tool, the approach also allows researchers to represent and visualize possible causes, and heterogeneous subpopulations, in their samples.
... A possible implementation of the case-control DID design is illustrated in Fig. 3(c), using panel data. DID models make two key assumptions: the common trends assumption (CT), and the common support assumption (COSU) (Angrist and Pischke, 2009;Keng and Sheu, 2011;Lechner, 2010). CT requires trends in the case and control groups to be roughly parallel prior to the introduction of the policy, and COSU requires that the distributions of other predictors of the outcome variable must remain roughly similar over time (Lechner, 2010). ...
Article
Full-text available
Low fertility is set to worsen economic problems in many developed countries, and maternity, paternity, and parental leave have emerged as key pro-natal policies. Gender inequity in the balance of domestic and formal work has been identified as a key driver of low fertility, and leave can potentially equalise this balance and thereby promote fertility. However, the literature contends that evidence for the effect of leave on fertility is mixed. We conduct the first systematic review on this topic. By applying a rigorous search protocol, we identify and review empirical studies that quantify the impact of leave policies on fertility. We focus on experimental or quasi-experimental studies that can identify causal effects. We identify 11 papers published between 2009 and 2019, evaluating 23 policy changes across Europe and North America from 1977 to 2009. Results are a mixture of positive, negative, and null impacts on fertility. To explain these apparent inconsistencies, we extend the conceptual framework of Lalive and Zweimüller (2009), which decomposes the total effect of leave on fertility into the “current-child” and “future-child” effects. We decompose these into effects on women at different birth orders, and specify types of study design to identify each effect. We classify the 23 studies in terms of the type of effect identified, revealing that all the negative or null studies identify the current-child effect, and all the positive studies identify the future-child or total effect. Since the future-child and total effects are more important for promoting aggregate fertility, our findings show that leave does in fact increase fertility when benefit increases are generous. Furthermore, our extensions to Lalive and Zweimüller’s conceptual framework provide a more sophisticated way of understanding and classifying the effects of pro-natal policies on fertility. Additionally, we propose ways to adapt the ROBINS-I tool for evaluating risk of bias in pro-natal policy studies.
... If we include controls in this context, we create bias in our results. (Angrist & Pischke, 2009, use the term 'bad controls' to describe instances of this sort.) Controls are of course sometimes relevant to a quantitative analysis of patterns in life satisfaction (and other forms of subjective well-being) -but a more explicit understanding of their function is required, so that we can identify clear criteria for selecting specific controls. ...
Article
Full-text available
Previous research has suggested that there is a ‘gender paradox’ associated with life satisfaction and depression: women are said to have higher levels of life satisfaction on average but also a higher likelihood of experiencing depression. That finding comes from quantitative analyses that include socio-demographic control variables. In this article I show that the inclusion of these control variables leads to biased results. In general, controls are to be selected on the basis that they are antecedents of the focal independent variable (as well as the dependent variable). When the focal independent variable is gender (or, more precisely, sex), no controls are required: there are no determinants of life satisfaction that also determine someone’s sex. If we include socio-demographic controls, we get biased results – because the controls themselves are affected by sex. More precisely: if we include controls (e.g. for income) to discern the difference between women’s and men’s life satisfaction, we get a result that fails to reflect the way women experience specific disadvantages (e.g. lower income) that contribute to lower life satisfaction. The same points apply to an analysis of depression. In a properly specified model (using data from the European Social Survey), there is no difference between women’s and men’s life satisfaction – so, there is no paradox with respect to depression.
... Benchmark results' analysis When using econometric models to estimate unknown parameters in linear regression models, the ordinary least square (OLS) method is the most common and basic estimation method in regression models (Angrist and Pischke, 2008). It minimizes the sum of squares of the differences between the real dependent variables and the predicted dependent variables to obtain the best linear unbiased estimator. ...
Article
Full-text available
As cyberspace has become an important factor in modern-day life, the impact of the Internet on residents has also attracted more attention. Based on the data of China Family Panel Studies (CFPS), this study empirically examines the impact of Internet use on Chinese residents’ subjective well-being (SWB) from a multi-dimensional perspective. The research found that Internet use had a significant impact on residents’ SWB, which was mainly reflected in job satisfaction, happiness, social ties, and future confidence. The impacts of the Internet’s different application fields are not consistent. Applying its use more in studying, working, socialize and commercial activities has a stronger effect, but has no significant impact on entertainment. Further heterogeneity tests also found that the marginal effect of Internet use increases with age, male and low-income groups can obtain greater benefit from the Internet, and there is almost no difference between urban and rural areas. This research provides micro evidence of the social effects of Internet use and provides enlightenment for how to further promote the quality of Internet use to better benefit people.
... Dafoe et al. (2018) argue that causal inference in these designs depends on the assumption of information equivalence, which has been referred to variously as "information leakage," "confounding," "masking," and "excludability" (Butler & Homola, 2017;Hainmueller et al., 2014;Sher & McKenzie, 2006;Tomz & Weeks, 2013). The excludability assumption refers to the notion that the causal effect runs only through the treatment, and not some other factor (Angrist & Pischke, 2009). Similarly, information equivalence refers to the assumption that manipulating the variable of interest updates their beliefs about the given attribute, but not beliefs about the background characteristics (i.e., the treatment effect runs only through the manipulation). ...
Article
Full-text available
Objectives: The current review assesses the methodological characteristics of between-subjects experiments, in particular documenting the scenarios and treat- ments described in each vignette, the extent to which confounds are embedded or accounted for in the design, and the analytic approach to estimating direct and inter- action effects. Methods: We conducted a pre-registered systematic review of 20 publications con- taining 20 independent studies and 23 vignette scenarios. Results: We find that the majority of studies rely on non-probability convenience sampling, manipulate a combination of procedural justice elements at positive and negative extremes, but often do not address potential confounds or threats to internal validity. The procedural justice manipulations that combine different elements show relatively consistent associations with a range of attitudinal outcomes, whereas the results for manipulations that test individual components of procedural justice (e.g., voice) are more mixed. Conclusions: Based on our review, we recommend that future studies using text- based vignettes disaggregate different elements of procedural justice in manipula- tions, and include a gradient of treatment or behavior (including control) to avoid comparing extremes, to incorporate potential confounders as either fixed covariates or manipulations, and to formally assess the information equivalence assumption using placebo tests.
Article
Full-text available
This paper investigates the productivity–wage relation using a novel and integrated employer-employee database covering the entire population of non-financial firms’ plants in one of the most developed regions in Europe, i.e., the Italian Lombardy region. We suggest that although a growing literature shows that locating in urban areas yields substantial productivity gains due to agglomeration economies, the interaction between productivity and wages is ultimately the key to ascertaining the true advantage of the high densely populated areas. By adopting an empirical specification that allows us to explore interaction effects between localization and the sector of activity at the establishment level, we find that agglomeration economies play a significant but conditional role in affecting productivity and wage differentials while also controlling for firm-specific factors (in particular, job-related characteristics) and selection effects. The estimated impacts are heterogeneous across sectors, depending on their technological features. The effect of locating in High-density urban areas on the productive-wage gap is significantly positive only in highly knowledge-intensive services sectors; for firmssupplyinglesstechnologicallysophisticatedservicesandfor manufacturingplants,theimpactiseithernotsignificantornegative. LocatinginRural areas generally exerts adownward(ornotsignificant)impactontheproductivity-wage gap.
Article
Full-text available
There is public pressure to reduce antimicrobial use in livestock production. Metaphylaxis usage raises special concern as it is given to a whole group of animals. The objective of this research was to determine the difference in cattle productivity and health (average daily gain, death loss, etc.) between cattle given metaphylaxis and those to which it was not given. Observational data were provided by a commercial feedlot in the Southern Great Plains region of the U.S.A. with an operating capacity >50,000 head. Cattle that received metaphylaxis treatment had substantially poorer health outcomes than those that did not. Cattle were more likely to have been given metaphylaxis treatment if they had a lower weight, were from a sale barn, or had been shipped long distances. Propensity score matching was used in an attempt to estimate the effect of metaphylaxis treatment on feedlot cattle. Propensity score matching was unable to overcome the endogeneity issues present in the data (endogeneity results from the animals being more likely to benefit from the treatment being the ones who received it). The dataset had information on cattle weight, state of purchase, and whether or not the cattle were from a sale barn, and so the feedlot must have based the treatment decision on information that was not recorded and therefore not included in the dataset. As an observational study, there are limitations in addition to data limitations, such as the possibility that the feedlot studied might not be representative of others. Even though the effect of metaphylaxis was not identified, the fact that it was unidentifiable supports the argument that the feedlot did treat the animals most likely to need metaphylaxis treatment. This should temper some fear of metaphylaxis treatment being overused and of antimicrobials being given needlessly.
Preprint
Full-text available
The purpose of this paper is to investigate if a country quality of governance moderates the effect of natural disasters on startup activity within that country. We test our hypotheses using a panel of 95 countries from 2006 to 2016. Our findings suggest that natural disasters discourage startup activity in countries that have low quality governance but encourage startup activity in countries that have high quality governance. Moreover, our estimates reveal that natural disasters effects on startup activity persist for the short term (1-3 years) but not the long term. Our findings provide new insights into how natural disasters affect entrepreneurship activity and highlight the importance of country governance during these events.
Article
The effect of the COVID shock on European economies has been severe and also unequal, with some firms being affected much more strongly than others. To improve the effectiveness of policy interventions, policymakers need to understand which types of vulnerable firms have been suddenly pushed into dire circumstances. We seek to fill this important gap in our knowledge by providing evidence from the European Investment Bank Investment Survey 2016–2020 on how the COVID shock has affected the investment activity and investment-related framework conditions of vulnerable firms. While data on actual investment activity post-COVID is not yet available to us, we focus on investment expectations. We exploit the fact that the same questions relating to investment expectations have been asked in several previous survey waves, which enables a difference-in-differences approach to investigate how investment expectations might have suddenly changed, for vulnerable groups of firms, immediately after the onset of the COVID crisis. We focus on 4 groups of vulnerable firms: High-Growth Enterprises (HGEs), young and small firms, R&D investors and non-subsidiary firms. R&D investors are more likely to be pessimistic about investment plans as a consequence of the COVID shock, and (similarly) HGEs are less likely to be optimistic about investment plans. R&D investors are less likely to be optimistic about the availability of internal finance, while HGEs and R&D investors are more likely to be pessimistic about the availability of external finance. Subsidiary firms, interestingly, are more likely to report a decrease in expected investment, which is not necessarily evidence of financial constraints, because it could instead be part of a conservative group-level strategy and coordinated group-level reduction in investment. Event study graphs generally confirm our regression results.
Article
Agricultural mechanization services (AMS) have emerged as a viable and effective solution for helping farmers gain access to machinery equipment in developing countries. This study investigates the simultaneous decision‐making regarding multiple mechanization services and the causal impacts of the AMS on land productivity among Chinese smallholders. We find that: (i) The ratio of off‐farm wage to AMS price, or the wage‐rent ratio, has a significant positive effect on AMS adoption; the sequential adoption of AMS starts with power‐intensive followed by control‐intensive production tasks, (ii) Switching to AMS in plowing, transplanting, and harvesting increases rice yield by 48.0 kg, 23.7 kg, and 7.9 kg, respectively, and (iii) The AMS of pesticide spraying significantly decreases yield but the negative effect can be mitigated with professional service providers and effective labor monitoring, highlighting potential moral hazards associated with AMS when monitoring is costly. Simulation illustrates that increased AMS adoption can significantly enhance food security in China.
Article
Inference for the parameters indexing generalised linear models is routinely based on the assumption that the model is correct and a priori specified. This is unsatisfactory because the chosen model is usually the result of a data‐adaptive model selection process, which may induce excess uncertainty that is not usually acknowledged. Moreover, the assumptions encoded in the chosen model rarely represent some a priori known, ground truth, making standard inferences prone to bias, but also failing to give a pure reflection of the information that is contained in the data. Inspired by developments on assumption‐free inference for so‐called projection parameters, we here propose novel nonparametric definitions of main effect estimands and effect modification estimands. These reduce to standard main effect and effect modification parameters in generalised linear models when these models are correctly specified, but have the advantage that they continue to capture respectively the (conditional) association between two variables, or the degree to which two variables interact in their association with outcome, even when these models are misspecified. We achieve an assumption‐lean inference for these estimands on the basis of their efficient influence function under the nonparametric model while invoking flexible data‐adaptive (e.g. machine learning) procedures.
Article
Recent advances in logistics tracking technologies have enabled e‐commerce firms to both accurately track the shipments and obtain accurate estimates of delivery times. With customers constantly tracking the status of their orders, the effects of logistics processes on customers' evaluation of their online shopping experience are relatively unknown as academic literature provides little guidance. Drawing upon the expectation‐disconfirmation theory and attribution theory, and using a unique dataset assembled in collaboration with an e‐commerce firm, we empirically investigate the effects of order processing and delivery times on online customer ratings. We also estimate the impact of positive deviation (i.e., early delivery) and negative deviation (i.e., late delivery) from the promised delivery date on online customer ratings and whether the impact of delivery status (early or late) depends on order cost and freight cost. Our empirical findings demonstrate that longer order processing and delivery times are associated with lower ratings. Our results also highlight that the late delivery of an order is negatively associated with ratings and that the order cost amplifies this effect. Furthermore, freight cost reduces both the negative effect of late delivery and the positive effect of early delivery on online ratings. Our results also indicate that there exist curvilinear relationships between online ratings and the number of days an online order is late or early. Specifically, the negative effect of a late delivery follows a convex‐shape curve such that its impact decreases as the number of days an online order is late increases. Likewise, the positive effect of an early delivery increases at a decreasing rate as the number of days an online order is early increases. Our supplementary analyses further account for potential endogeneity issues and corroborate our main results with additional empirical evidence obtained through alternative model specifications and estimation strategies. We present important theoretical implications and managerial takeaways from our findings.
Article
Background: Although positive youth development (PYD) programs have demonstrated effectiveness in improving adolescent reproductive health outcomes, there is a lack of evidence on effective school-based interventions designed especially for high school settings. This study examined the efficacy of Peer Group Connection (PGC-HS), a school-based PYD program, in improving sexual health outcomes for high school participants. Methods: A total of 1523 ninth-grade students at 18 schools were randomly assigned to be offered PGC-HS or a classes-as-usual control condition during 2016 to 2017 and 2017 to 2018 school years. Impacts were assessed on three confirmatory and 6 exploratory outcomes via self-reported participant questionnaire data collected at the beginning of 10th grade. Results: Although the offer of PGC-HS had no statistically detectable effect on confirmatory behavioral outcomes (sexual initiation, frequency of sex, and number of sexual partners) at 10th grade follow-up, causal impact estimates indicate that PGC-HS participants were less likely than control participants to ever have had vaginal sex. PGC-HS participants also scored higher on decision-making skills and perceived connection to peer connectedness. Conclusions: Results suggest that by building social and emotional skills and helping students form supportive peer relationships, PGC-HS may encourage students to make healthier choices and avoid risky behaviors during a critical period in high school, thus, reducing the risk of pregnancy.
Article
Full-text available
This article investigates how class of origin and intergenerational social mobility impact left-wing party support among new and old core left-wing electorates in the context of post-industrial electoral realignment and occupational transformation. We investigate the remaining legacy of political socialization in class of origin across generations of voters in the UK, Germany and Switzerland. We demonstrate that part of the contemporary middle-class left-wing support is a legacy of socialization under industrial class–party alignments, as many individuals from working-class backgrounds – traditional left-wing constituencies – have a different (post-industrial) class location than their parents. These enduring effects of production worker roots are weaker among younger generations and in more realigned contexts. Our findings imply that exclusively considering respondents' destination class underestimates the relevance of political socialization in class of origin, thereby overestimating electoral realignment. However, these past industrial alignments are currently unparalleled, as newer left-wing constituencies do not (yet) demonstrate similar legacies.
Article
Health information exchanges (HIEs) are designed to improve the quality and efficiency of healthcare by facilitating improved information sharing between health entities. This study systematically examines the impact of HIE use in emergency departments (EDs) on the quality and efficiency of medical care. We focus on the length of stay (LOS) and the 30-day readmission rate to capture healthcare efficiency and quality, respectively. We also examine whether the breadth of patient health information and physicians’ experience with the HIE moderates these effects. We leverage a unique panel data that tracks actual HIE access by physicians who practice in a set of hospitals that participate in the focal HIE. The patient-level encounter data set—which involves more than 80,000 ED encounters attended by more than 300 physicians over a 19-month period—comprises detailed medical provider information, patient-level medical information, and various other information related to procedures that were performed. After controlling for a battery of patient-specific, physician-specific, disease-specific, and ED visit-specific variables, our results show that HIE access in information-intensive environments (such as EDs) reduces LOS and 30-day readmission rate. We find that breadth of patient health information and physicians’ HIE experience amplify these benefits. We account for endogeneity issues and perform additional falsification tests and robustness checks. We document that benefits of HIE access are amplified for noninjury, chronic condition, and uncommon diagnoses related patient visits. Based on our results, we offer insights to practitioners and academicians alike on how HIEs can yield better patient-level and provider-level outcomes. This paper was accepted by Anandhi Bharadwaj, information systems.
Article
Interregional housing price comovement is a stylized fact worldwide. This study explores how it is affected by administrative division adjustment. We exploit city‐county mergers in China as a quasi‐natural experiment to construct a difference‐in‐differences strategy for causal identification. Based on monthly housing price data for districts (counties) in China from 2010 to 2019, we find that city‐county mergers significantly improve correlations in housing prices between the merged county and the urban district. This effect is more obvious in cities with a large economic gap between merged counties and urban districts, located in the central and western regions, and with lower administrative hierarchies (non‐provincial‐capital cities). The mechanism test shows that the impact of city‐county mergers on housing price comovement results mainly from integrating housing demand rather than integrating housing supply, like the unified land supply policy that local government implements in the new administrative scope after mergers. The results are helpful in understanding housing price comovement from the view of regional integration and provide clear policy implications for housing market regulation in China.
Article
Full-text available