Selected Papers

Goodbye Smokers' Corner: Health Effects of School Smoking Bans

Forthcoming in The Journal of Human Resources, 2018, downloadable here.

We study the impact of school smoking bans on individual health behavior in Germany. Using a multiple difference-in-differences approach in combination with randomization inference, we find that for individuals affected by a smoking ban during their school time, the propensity toward smoking declines by 14 to 22 percent, while the number of smoked cigarettes per day decreases by 19 to 25 percent. After elaborating on treatment effect heterogeneity and intensity, we evaluate spillovers to smoking behavior of non-treated individuals living in the same household.

The Importance of Tax Adjustments when Evaluating Wage Expectations

The Scandinavian Journal of Economics, 2019, 121:2, 578-605, downloadable here.

Using elicited expectations of future gross salaries, we evaluate characteristics causing German students to make larger or smaller estimation errors. While students seem to underestimate actual salaries by 18 percent, we show that these errors are highly attributable to misconceptions of the progressive income tax. Developing a suitable adjustment procedure, we correct students’ estimates and find that errors decline by 12 percentage points. Conducting regression analyses, we reveal strong connections with students’ age, gender, work experience, secondary school track, and knowledge about student loans. These results change notably if not controlling for students’ misconception of the tax system.

Illuminating the World Cup Effect: Night Lights Evidence from South Africa

Journal of Regional Science, 2018, 58:5, 887-920, downloadable here.

This paper evaluates the economic impact of the $14 billion preparatory infrastructure investments for the 2010 FIFA World Cup in South Africa. We use satellite data on night light luminosity at municipality and electoral district level as a proxy for economic development, applying synthetic control methods for estimation. For the average World Cup municipality, we find significantly positive, short-run effects before the tournament, corresponding to a reduction of unemployment by 1.3 percentage points. At the electoral district level, we reveal distinct effect heterogeneity, where especially investments in transport infrastructure are shown to have long-lasting, positive effects, particularly in more rural areas.

Media Coverage: SciBraai;; Sunday Times - you can also find the corresponding article here:

Click to Download

(SundayTimes_2016_06_19.pdf - 808.02K)

Comparative Politics and the Synthetic Control Method Revisited: A Note on Abadie et al. (2015)

Swiss Journal of Economics and Statistics, 2018, 154:11, downloadable here.

Recently, Abadie et al. (2015) have expanded synthetic control methods by the so-called cross-validation technique. We find that their results are not being reproduced when alternative software packages are used or when the variables’ ordering within the dataset is changed. We show that this failure stems from the cross-validation technique relying on non-uniquely defined predictor weights. While the amount of the resulting ambiguity is negligible for the main application of Abadie et al. (2015), we find it to be substantial for several of their robustness analyses. Applying well-defined, standard synthetic control methods reveals that the authors’ results are particularly driven by a specific control country, the United States.

Cross-Validating Synthetic Controls

Economics Bulletin, 2018, 38:1, 603-609, downloadable here.

While the literature on synthetic control methods mostly abstracts from out-of-sample measures, Abadie et al. (2015) have recently introduced a cross-validation approach. This technique, however, is not well-defined since it hinges on predictor weights which are not uniquely defined. We fix this issue, proposing a new, well-defined cross-validation technique, which we apply to the original Abadie et al. (2015) data. Additionally, we discuss how this new technique can be used for comparing different specifications based on out-of-sample measures, avoiding the danger of cherry-picking.

Outside the Box: Using Synthetic Control Methods as a Forecasting Technique

Applied Economics Letters, 2018, 25:9, 615-618, downloadable here.

We introduce synthetic control methods (SCM) as a forecasting technique. Using i) as economic predictors solely the outcome itself, i.e., lagged values of the dependent variable, and ii) lagged time series of the outcome to build the donor pool, we let SCM choose and weight appropriate values in order to come up with a sensible forecast of U.S. GDP growth. This procedure performs competitively viable compared with alternative forecasting methods.

The Incidence of Cash for Clunkers: Evidence from the 2009 Car Scrappage Scheme in Germany

International Tax and Public Finance, 2016, 23, 1093-1125, downloadable here.

This paper investigates the German car scrappage program, focusing on the incidence of the premium. We ask how much of the € 2,500 ($ 3,500) buyer subsidy is actually captured by the demand side. More precisely, we analyze the program's impact on different car segments, allowing for heterogeneity in incidence at different points in the vehicle-price distribution. Using a unique micro transaction data set, we find that the incidence of the subsidy strongly and significantly varies across price segments. Subsidized buyers of cheap cars paid a little more than comparable buyers who did not receive the subsidy, indicating incidence amounts slightly below 100%. For more expensive vehicles, subsidized buyers were granted large extra discounts on top of the government premium, translating into incidence amounts considerably greater than 100%. Taken together, this results in an aggregate incidence amount of up to plus EUR 350 million, suggesting that the positive effect for expensive cars overcompensates the negative effect for small cars.

Synthesizing Cash for Clunkers: Stabilizing the Car Market, Hurting the Environment?

Discussion Paper

We examine the impact of the €5 billion German Cash for Clunkers program on vehicle registrations and respective CO2 emissions. To construct proper counterfactuals, we develop the multivariate synthetic control method using time series of economic predictors (MSCMT) and show (asymptotic) unbiasedness of the corresponding effect estimator under quite general conditions. Using cross-validation for determining an optimal specification of predictors, we do not find significant effects for CO2 emissions, while the stimulus’ impact on vehicle sales is strongly positive. Modeling different buyer subgroups, we disentangle this effect: 530,000 purchases were simply windfall gains; 550,000 were pulled forward; and 850,000 vehicles would not have been purchased in absence of the subsidy, worth €17 billion.

You can find the paper here:

Click to Download(MPRA_paper_88175.pdf - 667.96K)

You can find the recent version of the MSCMT package (for R) here.

Perceived Wages and the Gender Gap in STEM Fields

Discussion Paper

We estimate gender differences in elicited wage expectations among German University students applying for STEM and non-STEM fields. Descriptively, women expect to earn less than men and also have lower expectations about wages of average graduates across different fields. Using a two-step estimation procedure accounting for self-selection, we find that the gender gap in own expected wages can be explained to the extent of 54-69% by wage expectations for average graduates across different fields. However, gender differences in the wage expectations for average graduates across different fields do not contribute to explaining the gender gap in the choice of STEM majors.

You can find the paper here:

Click to Download(CEPR-DP12719.pdf - 282.84K)

Synthetic Control Methods: Never Use All Pre-Intervention Outcomes Together With Covariates

Discussion Paper

It is becoming increasingly popular in applications of synthetic control methods to include the entire pre-treatment path of the outcome variable as economic predictors. We demonstrate both theoretically and empirically that using all outcome lags as separate predictors renders all other covariates irrelevant. This finding holds irrespective of how important these covariates are for accurately predicting post-treatment values of the outcome, threatening the estimator's unbiasedness. We show that estimation results and corresponding policy conclusions can change considerably when the usage of outcome lags as predictors is restricted, resulting in other covariates obtaining positive weights. Monte Carlo studies examine potential bias.

You can find the paper here:

Click to Download(SCM_Predictors.pdf - 491.69K)

The most current version is here:

Click to Download

(SCM_Predictors_MC.pdf - 465.16K)

More Papers

Click to Download

(Pfeifer.G-CV.pdf - 166.92K)

Contact Information

Department of Economics
University College London
Drayton House, 30 Gordon Street, London, UK