Ashok Kaul

Assessing Aggregate Tests of Efficiency for Dynamic Economies

In a seminal contribution Abel, Mankiw, Summers, and Zeckhauser (1989) show that from an aggregate dynamic perspective the US economy is Pareto efficient. We argue that, when applying their test, they implicitly make strong assumptions about the economys future behavior. We show how time series evidence may easily lead to wrong conclusions about the welfare properties of real world economies.We present a test criterion based on Zilcha (1991) and robust evidence that the US economy does not overaccumulate capital. Our contribution highlights that the distinction between efficient capital accumulation and Pareto efficiency is empirically relevant. The latter efficiency benchmark - encompassing also risk sharing issues - cannot be rigorously tested based on available approaches in the literature.

On the interaction between risk sharing and capital accumulation in a stochastic OLG model with production

Abstract We analyze the interaction between risk sharing and capital accumulation in a stochastic OLG model with production. We give a complete characterization of interim Pareto optimal competitive equilibrium allocations. Furthermore, we provide tests of Pareto optimality/suboptimality based on (risky) rates of return only.

The incidence of Cash for Clunkers: an analysis of the 2009 car scrappage scheme in Germany

Governments all over the world have invested tens of billions of dollars in car scrappage programs to fuel the economy in 2009. We investigate the German case using a unique micro transaction dataset covering the years 2007 to 2010. Our focus is on the incidence of the subsidy, i.e., we ask how much of the € 2,500 buyer subsidy is captured by the supply-side through an increase in selling prices. Using regression analysis, we find that average prices in fact decreased for subsidized buyers in comparison to non-subsidized ones, suggesting that eventually subsidized customers benefitted by more than the subsidy amount. However, the incidence was heterogeneous across price segments. Subsidized buyers of cheap cars paid more than comparable buyers who did not receive the subsidy, e.g. for cars of € 12,000 car dealers reaped about 8% of the scrappage prime. The opposite was true for more expensive cars, e.g. subsidized buyers of cars of € 32,000 were granted an extra discount of about € 1,100. For cars priced about € 18,000, we find no price discrimination, i.e., in this price segment consumers fully captured the transfer. Our results can be explained by optimizing behavior on the supply-side both in the lower and upper price segments. The results are extremely robust to extensive sensitivity checks.

The (Possible) Effect of Plain Packaging on the Smoking Prevalence of Minors in Australia: A Trend Analysis

A key stated objective of the Australian Plain Packaging Act 2011 is to influence smoking prevalence, in particular of minors. We use the Roy Morgan Single Source (Australia) data set on minors, (that is, Australians aged 14 to 17 years) over the time period January 2001 to December 2013 to analyze whether there is evidence that this goal has been achieved. We carry out a statistical trend analysis to study the (possible) effect of plain packaging on smoking prevalence of minors in Australia. More specifically, we fit a linear time trend that explains well the fact that observed smoking prevalence has declined steadily over the last 13 years. Two informative analyses help to draw conclusions on the (actual) effect of plain packaging on smoking prevalence of Australian minors. First, we look at the year of data before plain packaging was introduced, which happened in December 2012. Second, we compute confidence intervals around the estimated treatment effects (that is, around the deviations from the fitted trend line) from December 2012 on. Both analyses fail to find any evidence for an actual plain packaging effect on Australians aged 14 to 17 years. Several reasonable variations to our methodology are discussed. All of these would only result in findings even more indicative of an absence of any plain packaging effect.

The (Possible) Effect of Plain Packaging on Smoking Prevalence in Australia: A Trend Analysis

A stated objective of the Australian Plain Packaging Act 2011 is to reduce smoking prevalence. We use the Roy Morgan Single Source (Australia) data set over the time period January 2001 to December 2013 to analyze whether this goal has been achieved in the first year since the implementation. In particular, we carry out a statistical trend analysis to study the (possible) effect of plain packaging on smoking prevalence. Two informative analyses help to draw conclusions on the (actual) effect of plain packaging on smoking prevalence in Australia. First, we look at the year of data before plain packaging was introduced, which happened in December 2012. Second, we compute confidence intervals around the estimated treatment effects. Our main results can be summarized as follows. First, if a statistical significance level of 5% is required, then there is no evidence at all for a plain packaging effect on smoking prevalence. Second, if one is willing to accept a relatively low level of statistical significance (that is, 10%), then there is evidence for a very short-lived plain packaging effect on smoking prevalence, namely in December 2012 only (after which smoking prevalence is statistically indistinguishable from its pre-existing trend). A formal power analysis demonstrates that the power of our inference methods is remarkably high.

Comparative politics and the synthetic control method revisited: a note on Abadie et al. (2015)

Recently, Abadie et al. (Am J Polit Sci 59:495–510, 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. (Am J Polit Sci 59:495–510, 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 USA.

The Performance of Mechanical Earnings Forecasts

We analyze three different mechanical models to forecast earnings and compare their forecasts with those of analysts. Moreover, we evaluate implied cost of capital (ICC) estimates that are based on these forecasts. With our analyses we answer three open questions in the literature. 1) Do model forecasts or analysts’ forecasts perform better? 2) Are ICCs derived from analysts’ forecasts more reliable than ICCs based on model forecasts? And 3) does higher forecast performance also translate into more reliable ICCs? First, we find that analysts’ forecasts are even more accurate than the most accurate model forecasts. However, second, we find that model-based ICCs are always more reliable than analyst-based ICCs. Moreover, model-based ICCs are particularly reliable for a sample of firms for which no analysts’ forecasts are available. While the lack of reliability of analyst-based ICCs seems to indicate a missing link between forecast performance and ICC reliability, in fact, third, we find that ceteris paribus higher forecast performance translates into more reliable ICCs, that is, within one earnings definition the most accurate forecasts also yield the most reliable ICCs.

Standard of proof in WTO dispute settlement proceedings: an applied statistical perspective

ABSTRACT This paper proposes a statistical framework for establishing prima facie evidence in WTO proceedings in which empirical evidence and statistical analysis play an important role. As our main contribution, we suggest a general guideline for the choice of the significance level in a statistical analysis, in order to meet the requirements for establishing prima facie evidence of a claim. We consider a fundamental asymmetry between whether a complainant’s initial claim is positive or negative. Statistically, a positive claim is one in favour of the alternative hypothesis, whereas a negative claim is one in favour of the null hypothesis. To account for this asymmetry, we suggest using different significance levels for positive and negative assertions, respectively. For positive claims, we suggest using the most commonly applied significance level of 5%. For negative claims, we suggest the less strict (but still commonly accepted) significance level of 10%. Choosing a less strict significance level in the case of a negative claim than in the case of a positive claim accounts for another fundamental asymmetry between the two possible outcomes of a statistical hypothesis test: rejecting the null hypothesis or not. Our framework helps to define reasonable hurdles in WTO proceedings to substantiate claims using statistical methods while leaving enough leeway for the panel and the Appellant Body to make adjustments on a case-by-case basis.

Skaleneffekte bei Verwaltungs- und Verfahrenskosten der Unfallversicherungsträger der öffentlichen Hand

Die gesetzliche Unfallversicherung wurde im Rahmen des Unfallversicherungsmodernisierungsgesetzes (UVMG) zum 1.1.2009 grundlegend reformiert. Ein wesentliches Ziel der Reformen war es, durch Fusionen der Versiche-rungstrager die Verwaltungs- und Verfahrenskosten je Versicherten zu reduzieren, also Skaleneffekte zu realisieren. Allerdings gab es zum Zeitpunkt der Reform keinerlei empirische Evidenz, dass bedeutsame Skaleneffekte zu erwar-ten waren. Basierend auf Daten der Jahre 1998-2007 wird die Frage untersucht, ob grose Unfallkassen geringere Verwaltungs- und Verfahrenskosten je Versicherten aufwenden mussen als kleine. Mittels Paneldatenanalyse kann uber mehrere Spezifikationen hinweg robust gezeigt werden, dass die Hohe der Verwaltungs- und Verfahrenskos-ten je Versicherten negativ von der Grose des Tragers (gemessen durch die Anzahl der Versicherten) abhangt. Im statistischen Mittel sinken die Verwaltungs- und Verfahrenskosten je Versicherten um ca. 33 Cent, wenn die Anzahl der Versicherten um 100.000 zunimmt. Bei mittleren Kosten von ca. 6,54 € je Versicherten uber den betrachteten Zeitraum und durchschnittlich 960.000 Versicherten je Unfallkasse ist dieses Einsparpotential erheblich. Bei einer Steigerung der Kassengrose um ca. 600.000 Versicherte (um eine Standardabweichung) besteht damit ein Einspar-potenzial von ca. 30%.

Reformen auf der Suche nach Skaleneffekten: die Struktur der Verwaltungs- und Verfahrenskosten der �ffentlichen Unfallkassen in Deutschland

We analyze the cost savings potential with regard to administration and management costs by merging Public Employment Accident Insurance Funds (i?½ffentliche Unfallkassen) in Germany. Our data base consists of the annual reports of the public statutory accident insurance funds from 1999 to 2005. Results of a multivariate regression analysis help to identify cost drivers and the potential existence of economies of scale. The numbers of employee accidents and pupil accidents as well as expenditures to prevent accidents serve as explanatory variables for administration and management costs. The main results are as follows. First, public accident insurance funds are very heterogeneous with regard to administration and management costs. Second, these cost differences are unrelated to the size of insurance funds (the number of insured employees and pupils). There is no evidence for cost savings of larger insurance funds, i.e., economies of scale. Third, accidents of pupils entail substantially lower management and administration costs than accidents of employees. Furthermore, there is evidence that higher expenditures to prevent accidents such as accident prevention programs also involve higher administration and management costs.