Abstract This paper analyzes private precautions against crime when the value of the property to be protected is private information. In a framework in which potential criminals can choose between different crime opportunities, we establish that decentralized decision-making by potential victims may lead to suboptimal levels of investment in private protection. This outcome is possible when observable precautions inform potential offenders about the value at risk even when the diversion effect due to private safety measures is taken into account.

Abstract This paper experimentally analyzes the cartel coordination challenge induced by the discrimination of cartel ringleaders in leniency policies. Ringleaders often take a leading role in the coordination and formation of a cartel. A leniency policy which grants amnesty to all whistleblowers except for ringleaders may therefore reduce the incentive to become a ringleader and may disrupt cartel formation. We analyze discriminatory and non-discriminatory leniency policies in a multi-stage cartel formation experiment where multiple ringleaders may emerge. Although theory predicts that cartels will always be reported, whistleblowing rarely occurs. Paradoxically the discriminatory leniency policy induces more rms to become ringleaders, which ultimately facilitates coordination in the cartel. JEL Classication numbers: C92, K21, L41.

Abstract We analyze the effects of cartel policies on firm behavior using data from the German powercable cartel. Antitrust authorities affected the cartel under two different legal regimes: penalizing the cartel in some years, and exempting it for ten years from the general cartel prohibition. While penalties did not reduce prices or profits, making collusion legal raised profits by at least 16% each year, compared to the time when the illegal cartel was not prosecuted. The threat of penalties was sufficient to reduce profit from collusion. The intended efficiency gains from rationalization, which was the justification for legalizing the cartel, did not materialize. * We are grateful to a referee for detailed and helpful comments. We are also indebted to John Connor, seminar

Abstract: This paper investigates the existence of a possible media bias by analyzing the impact of automobile manufacturer's advertisements on automobile reviews in a leading German car magazine. By accounting for both endogeneity and sample selection using a two-step procedure, we find a positive impact of advertising volumes on test scores. The main advantage of our study is the measurement of technical characteristics of cars to explain test scores. Due to this kind of measurement, we avoid serious biases in estimating media bias caused by omitted variables. Keywords: Car magazines, Media bias, Selection model, Instrumental variable estimation JEL classification: L150, L820 a We are grateful to Michael Berlemann, Dirk Czarnitzki, Stephen Martin, and the participants of the MSI BrownBag Seminar at KU Leuven. We also thank Stephanie Effern, Hagen Niehaus, Julia Rothbauer and Denise Weber for excellent research assistance. b Ralf Dewenter, Chair of Industrial Economics, Helmut-Schmidt-University of Hamburg, Holstenhofweg 85, 22043 Hamburg, Germany, Mail: dewenter@hsu-hh.de. c Ulrich Heimeshoff, DICE, Heinrich-Heine-University of Duesseldorf, Universitaetsstr. 1, 40225 Duesseldorf, Germany, Mail: heimeshoff@dice.hhu.de. -2- Introduction Media bias has been an important research topic in economics as well as marketing science, political science, and journalism. The most important objective for a free press is unbiased information for citizens. 4 As Hamilton (1994: 7) puts it: "News is a commodity, not a mirror image of reality." There is a large literature on media bias not only from economics, which shows that biased coverage is far from being rare. Consequently, media bias from different sources is studied substantially and it is therefore not surprising that a substantial part of research on media bias deals with political biases. It is well known that many daily newspapers or TV channels are biased either to the left or the right of the political spectrum instead of being politically neutral. It is, e.g., a common complain that the New York Times has an explicit liberal viewpoint. As a result, coverage in the New York Times is not objective and consequently biased. Of course such biases can be found in both directions of the political spectrum, a prominent example is the Washington Post and its apparently conservative bias. These biases are not necessarily profit-oriented, but depend on the political opinions of the owners, the editors, and/or the journalists of the newspaper or magazine. Another important source of media bias is profit-oriented and therefore also connected with the theory of two-sided-markets (see, e.g., The effects of biased coverage in two-sided-markets are significantly different from onesided-markets. In one-sided-markets assuming negligible cost, biased coverage would certainly have at least a non-negative effect on profits. In two-sided-markets predictions about the effects 4 However, it is well known, that some consumers like to read coverage which is not the truth, but in line with their beliefs. See -3-of biased reporting are much more difficult. The effects can be weaker as well as stronger compared to one-sided-markets. The result depends on the size of the network effects from the advertising to the recipients market. Media bias will always have stronger effects than in onesided-markets, if readers like advertising, or in other words if network effects are positive. This effect, which is based on the reinforcing impact of two-sided network effects, shows the interdependency of advertising and recipients markets. Increasing amount of advertising volume also fosters demand for copies which in turn lead to a stronger demand for advertising space. However, readers will not always like advertising in reality. Readers' dislike of advertising will produce a trade-off between advertising and demand for copies. An increase in the demand for advertising space would reduce the demand for copies. As a result, the incentives for biased coverage would be reduced when readers don't like advertising. There is a number of non-economic studies dealing with political biases using data from the U.S., where much evidence on political biases has been collected. Biases have been reported on topics such as the U.S. quality press and climate change (see The problem of measuring product performance is a major advantage of our study. In order to measure a possible bias we use information on car reviews of a leading German car maga- 5 There is now a whole field called "Wine Economics", which started mainly with the work of Ashenfelter. See e.g. Ashenfelter (1989) and Ashenfelter (2008). -5-zine, Auto, Motor und Sport, and at the same time we use information on manufacturers' advertising volumes. By these means a possible link between manufacturers' advertising volumes and the performance of the respective cars can be analyzed. Analyzing car reviews is advantageous for some reasons: We have substantial information on cars' characteristics and as a result of this information, we are able to filter the more or less objective part of a test score. Additionally, we can check where media bias due to advertising volumes steps in and are therefore able to identify which parts of the test scores are either based on technical characteristics or on the impact of advertising customers. Mutual funds and wine are both products which face a major problem which is the evaluation of performance. It is rather easy to evaluate the performance of mutual funds, but it is always some sort of ex post evaluation. Additionally, the performance does not only depend on the fund managers' actions but also on general market conditions. Wine recommendations and reviews are always subjective in some sense. This is where our study adds to the literature. We use data on car reviews and manufacturers' advertising volumes from a leading German car magazine which enables us to control for technical characteristics of cars. As a result, we can explain to which extend technical features of the cars explain the test results and can also test the effect of advertising volumes on test results. Cars have the big advantage that there are several objective criteria as for example horse power, trunk capacity, and mileage we can observe and include into our regressions. This enables us to give a more precise statement on the size of media bias compared to other products whose performance is more difficult to evaluate. In addition we take into account the endogeneity of advertising volumes which result from the interdependency of both advertising and recipients markets. Using instrumental variable techniques, we compare our results to simple OLS estimates to get an impression of the bias. We also take into account possible selection biases, because there are significant differences between manufacturers' cars being reviewed. To solve this problem we estimate Heckman two step models, which estimate the probability of a given car to be reviewed and include this as the inverse Mills ratio into our second step regression, where we estimate the effects of advertising volumes on test results. The paper proceeds as follows: In the next section we describe our data and several steps of aggregation. Section 3 discusses the estimation and identification strategy. In section 4 we present our results, several robustness checks, and provide some interpretations about the stability of media bias over time. Section 5 concludes and gives some suggestions for further research. -6- Data To analyze the impact of car manufacturers' advertising expenditures on car reviews, we collected data on test results as well as a measure (or at least a proxy) for advertising expenditures. We also use several controls in order to prevent misspecification and omitted variable issues. Descriptions of the variables in our dataset as well as descriptive statistics can be found in tables A1, A2, and A3 in the appendix. The data used in this study is extracted from different sources (see 6 To analyze the impact of advertising expenditures on test results, we use the number of advertising pages in an issue placed by a manufacturer as a proxy for advertising expenditures. In 6 Unfortunately, as each issue of AMS includes typically more than one comparative test using more than one model of a specific manufacturer and, furthermore, many models are tested only once over the whole period, we are not able to build a panel but must restrict our analysis to pooled time series. -7-order to prevent cyclical patterns and also in order to control for possible trends in the data we use relative advertising pages instead of absolute values. That is, ADS is the number of advertising pages by a manufacturer divided by all pages of other car manufacturers within an issue of AMS. Controls To account for a possible home bias we include a dummy variable GERMAN which is equal to one if a car is built by a German manufacturer. A home bias would exist in case, that German inhabitants are more likely to buy German cars than vehicles from other countries. We use the issue size in PAGES, as the number of pages increases the potential amount of advertising in a magazine. Furthermore, the more pages a magazines the more tests can be reported in a single issue and the probability of given car to be reviewed might increase. In order to prevent a possible spurious regression we use the technical characteristics of each car as control variables. Without these characteristics our results might be seriously biased. LNHP is the natural logarithm of horsepower, and DIESEL is a dummy variable which takes the value one if a given car is powered by a Diesel engine. Furthermore, LNCAPACITY is the natural logarithm of trunk size and we also include dummy variables for the number of doors (DOORS DUMMIES) and the bodywork a car has (BODYWORK DUMMIES), which might be for example a sedan or a sports car. Additionally, we include dummies for the corporations, which produced the cars. These dummies are group dummies, e.g., we have dummies for the Volkswagen Group or the General Motors Group. Finally, we also have issue dummies in our regressions which provide some control for time and help us to take account of seasonality and other changes in general conditions. LNPRICE is the natural logarithm of the cars' prices. Empirical Analysis Estimation Approach and Identification Strategy Most studies of media bias do not account for the fact that it is by no means random whether a given product will be reviewed in a magazine. The likelihood of being reviewed depends on a number of explanatory variables, which we take into account in our analysis (see also . Furthermore, the advertising variable is very likely to be endogenous which is also very rarely taken into account in earlier studies. The next sections provide an overview on our estimation strategy. Identifying the impact of advertising expenditures on test scores, i.e. a possible media -8-bias, is no trivial task. A number of problems arise which will be discussed in the following: We use the two-step procedure suggested by 1. First step: We estimate the review probability of each manufacturer, using three different models: a. Using ordinary least squares regressions without advertising volumes as an explanatory variable, b. Using ordinary least squares treating advertising volumes as a exogenous explanatory variable, c. Using instrumental variable techniques to account for possible endogeneity of advertising volumes. 2. Second step: Estimating the effects of advertising on test results on a basis of cars being reviewed using instrumental variable techniques taking account of possible endogeneity problems with regard to advertising volumes. We furthermore include the inverse Mills ratios from step a. to c. to control for a possible selection bias in different specifications. First, as only cars of some of the manufacturers available are reviewed in each issue of a magazine, the selection of these models is by no means random. It is rather likely that a number of factors determine the probability of a model being reviewed. It is, for example, more likely that new releases will enter a test than older models. Moreover, also the market share as well as the manufacturers' origin should play a role in the selection process. Ignoring such problems could result in a severe selection bias. In order to prevent a selection bias when analyzing the impact of ad volumes on test scores we first determine the probability for each car manufacturer of being reviewed in a specific issue using probit regressions. For this purpose, we built a new dummy variable which is equal to one for manufacturer i in case that at least one of its models have been reviewed in issue t and zero otherwise. We then are able to calculate the inverse mills' ratio from this first regression. Second, while we assume that advertising volumes are likely to have a positive impact on test scores, it is equally possible that test scores influence (future) advertising volumes as well. Manufacturers are likely to increase ad volumes in order to improve test scores or to decrease advertising expenditures as a response to poor results. In both cases however a problem of reverse causality arises which may result in a bias overestimating the impact of advertising volumes. Similarly, as test scores possibly influence advertising expenditures there might be a -9-connection between advertising volumes and the selection of models being reviewed as well. Therefore, equally when analyzing the review probability reverse causality could occur. A possible solution to problems of endogeneity is to find adequate instruments in order to use instrumental variables techniques (see where REV i,t equals one if a model by manufacturer i is reviewed in issue t. X i,t includes a number of control variables (e.g., a Germany dummy, the number of new releases and time dummies) and εi,t is an error term. We use all of the above mentioned three models to calculate the respective inverse Mills' ratio. Third, following Milgrom and Roberts (1986), a possible spurious regression problem can arise when advertising is used as a signal for quality. In case that manufacturers with higher quality vehicles also spend more on advertising, we would measure a positive impact of ADS too though a media bias would not exist. To account for a possible spurious regression we use characteristics of the reviewed passenger cars such as the type, the motor power, the type of fuel etc., when analyzing the impact of advertising volumes on test scores. 7 This kind of advertising strategy is well known as pulsing campaign in the marketing literature. Usually advertising budgets are not used to have a constant stream of advertising on the relevant market but show some cyclical behavior. See for example Feinberg (1992). -10-Fourth, in case that a media bias exists it is unclear to what extend this bias is also induced by the readers of the car magazines. Customers may be influenced by advertisements or may simply prefer reviews of specific cars. Again, in that case the effect of advertising on test scores may be overestimated. We are not able to control for this problem but assume that it is not too severe. We expect the most important bias due to customers' preferences to occur with respect to German cars. However, as we use a dummy for German manufacturers we do not expect an additional bias. For the second step regressions of our analysis, we get the following general specification: SCORE is the number of points a manufacturers' car received in a given test. ADS is our measure of advertising volume bought by the given manufacturer and the matrix X includes important control variables as the technical characteristics of the cars. As a result, we estimate some sort of hedonic regression to explain the test score including ADS and the inverse Mills ratios from step 1 of our analysis in the regressions. The next section presents the results of both steps of our analysis. Results Review Probability We start with probit regressions of the Heckman selection models. As advertising volumes are likely to affect the probability of a manufacturer's model being reviewed as well as test scores, ADS should be considered as an explanatory variable. However, similarl to the following analysis of test scores, there might also be a problem of reverse causality. Put differently, in case that advertising volumes affect test probabilities, the selection of cars for reviews could also affect manufacturers' advertising expenditures. The second condition which is necessary for valuable instruments is exogeneity. In our case finding completely exogenous instruments is rather difficult. Both variables could also be affected by reviews and are therefore possibly endogenous. However, due to the lead and lag structure of our instruments, these variables should not cause serious endogeneity problems. Furthermore, we use past car registrations not only in Germany but for Europe. This variable does not depend on the German market solely but on all national car markets in Europe. As a result, endogeneity problems should become less likely.

Abstract This paper studies unshrouding decisions in a framework similar to

Die ZBW räumt Ihnen als Nutzerin/Nutzer das unentgeltliche, räumlich unbeschränkte und zeitlich auf die Dauer des Schutzrechts beschränkte einfache Recht ein, das ausgewählte Werk im Rahmen der unter

Abstract This paper analyses the income effect of the participation in elite sports using a unique dataset on former German top-level athletes. To quantify the average treatment effect we use covariate nearestneighbour matching. While our treatment group consists of formerly top-level athletes the control group of non-athletes is drawn from the GSOEP database. On average, former athletes receive higher incomes than similar non-athletes. Moreover, team sports athletes as well as male athletes realise significantly higher incomes. Comparing the income of former female athletes with male non-athletes, we find that participating in elite sports closes the gender-wage gap.

Abstract When Apple entered the ebook market, prices rose. A recent court decision found Apple guilty of colluding with publishers, blaming the price hike, in part, on agency agreements and prohibiting their use. Building a model to compare these with traditional wholesale agreements, we identify a single, pivotal condition that leads prices under agency to be higher than under wholesale with two-part tariffs but lower with linear pricing. Our model shows that the increase in ebook prices can be explained, instead, by heightened competition for reading devices, and it guides our understanding of when restricting agency agreements is advisable.

Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle

Abstract This study investigates the effects of an R&D subsidy scheme on participating firms' net R&D investment. Making use of a specific policy design in Belgium that explicitly distinguishes between research and development grants, we estimate direct and cross-scheme effects on research versus development intensities in recipients firms. We find positive direct effects from research (development) subsidies on net research (development) spending. This direct effect is larger for research grants than for development grants. We also find cross-scheme effects that may arise due to complementarity between research and development activities. Finally, we find that the magnitude of the treatment effects depends on firm size and age and that there is a minimum effective grant size, especially for research projects. The results support the view that public subsidies induce higher additional investment particularly in research where market failures are larger, even when the subsidies are targeting development.