Andreas Behr

Stochastic Data Envelopment Analysis

The deterministic data envelopment analysis is descriptive. The observed firms are compared with a (most often synthetic) benchmark firm which is either observed or constructed as a linear combination of observed firms. The stochastic data envelopment analysis now enriches the analysis by speculating about firms which have not been observed. The basic idea is that there would probably be an even more efficient benchmark firm if only more firms would have been observed. According to this speculation, the deterministic DEA efficiency scores are probably too optimistic. In this chapter, we discuss how stochastic considerations can be added to data envelopment analysis.

Data Envelopment Analysis

Data envelopment analysis is a very popular method to obtain efficiency scores for firms. Its charm is its simplicity. The firms under analysis are compared to the most efficient firm, which most often is a synthetic firm obtained as a linear combination of reference firms. The method is nonparametric as no assumptions on functional relations between inputs and outputs have to be made.

Default Patterns in Seven EU Countries: A Random Forest Approach

Abstract This study uses the relatively new “random forest” (RF) approach, which is based on decision-tree analysis by combining the results of a large set of decision trees. RFs have so far been little used for default prediction but offer an interesting alternative to well-established default prediction techniques. Based on accounting data from 945,062 observed European firms from seven countries in 2010 and 1,019,312 firms in 2011, we provide evidence on the country-specific default patterns. Because of the strong imbalance of the data sets with regard to the solvency status, standard RF implementations have to be modified to allow the estimation of realistic default propensities. We find that by far most accurate out-of-sample default propensities can be obtained for Italy followed by Portugal and Spain and the least accurate for the UK and Finland. The debt ratio, rate of return on sales, dynamic gearing ratio, and the rate of return on assets are found to be the most important variables for default prediction. The variable importance rankings are rather country specific, pointing to heterogeneity in the default patterns across the countries studied.

Default prediction using balance-sheet data: a comparison of models

Purpose - The purpose of this paper is to do a performance comparison of three different data mining techniques. Design/methodology/approach - Logit model, decision tree and random forest are applied in this study on British, French, German, Italian, Portuguese and Spanish balance sheet data from 2006 to 2012, which covers 446,464 firms. Because of the strong imbalance with regard to the solvency status, classification trees and random forests are modified to adapt to this imbalance. All three model specifications are optimized extensively using resampling techniques, relying on the training sample only. Model performance is assessed, strictly, based on out-of-sample predictions. Findings - Random forest is found to strongly outperform the classification tree and the logit model in almost all considered years and countries, according to the quality measure in this study. Originality/value - Obtaining reliable estimates of default propensity scores is of immense importance for potential credit grantors, portfolio managers and regulatory authorities. As the overwhelming majority of firms are not listed on stock exchanges, annual balance sheets still provide the most important source of information. The obtained ranking of the three models according to their predictive performance is relatively robust, due to the consideration of several countries and a relatively long time period.

Health System Efficiency: A Fragmented Picture Based on OECD Data

BackgroundGlobally, health expenditure as a percentage of GDP has increased in recent years, so evaluating the health care systems used in different countries is an important tool for identifying best practices and improving inefficient health care systems.ObjectiveWe investigate health system efficiency at the country level based on OECD health data. We focus on several aspects of health care systems to identify specific inefficiencies within them. This information hints at potential policy interventions that could improve specific parts of a country’s health care system.MethodsA discussion is provided of ideal-typical evaluations of health systems, ignoring data restrictions, which provide the theoretical basis for an analysis performed under factual data restrictions. This investigation includes health care systems in 34 countries and is based on OECD health data. Health care system efficiency scores are obtained using data envelopment analysis (DEA). Relative productivity measures are calculated based on average DEA prices. Given the severe data limitations involved, instead of performing an all-encompassing analysis of each health care system, we focus on several aspects of each system, performing five partial analyses.ResultsFor each country, the efficiencies yielded by the five partial analyses varied considerably, resulting in an ambiguous picture of the efficiencies of the various health care systems considered. A synopsis providing comprehensive rankings of the analyzed countries is provided.ConclusionAnalysis of several aspects of the health care systems considered here highlights potential improvements in specific areas of these systems, thereby providing information for policymakers on where to focus when aiming to improve a country’s health care system.

Stochastic Frontier Analysis

The stochastic frontier analysis is an econometric approach to efficiency measurement. The basic idea is the introduction of two error components, a random error term and an inefficiency term. For both terms, a distributional assumption is made, which facilitates maximum likelihood estimation.

Panel Data Stochastic Frontier Analysis

Stochastic frontier analysis based on cross-sectional data is hampered by the fact that only one observation is available for the estimation of two error components. Panel data containing several observations for each firm considerably improve the situation for estimating firm specific efficiency scores if some assumptions on the time path of inefficiencies are introduced.

Linear Production Model

Linear production models allow a concise representation of the production process in an economy. The interdependence between production units is its main characteristic as each production unit uses the output of other producing units as inputs. Empirical content is provided by analyzing highly aggregated input–output tables for Germany.

Production Functions with Panel Data

In this chapter, we discuss the estimation of production functions using panel data. Panel data sets include data for producing units which have been observed for several periods. Information about differences of production relations between units and of differences between time periods allows to drop some restrictive assumptions on parameter homogeneity which have to be employed in the cases of cross sectional or time series data. We discuss the pooled, the between, the within, and the random effects estimator for static models and hint for some problems when instrumenting in dynamic panel data models. The generation of data according to the different models and their estimation are demonstrated.

Return Distributions and Bootstrap Goodness-of-fit Tests

It is seen as a “stylised fact” that marginal distributions of stock returns have thick tails, are skewed and leptocurtic (see for example Campbell et al. 1997, Eijgenhuijsen and Buckley 1999, Cont 2001 and Behr and Poetter 2007). Therefore, the simple normal distribution is inappropriate to describe empirical return distributions. In response to these “stylised fact” several flexible parametric distributions of returns have been proposed in the literature. The generalised hyperbolic distribution has obtained a fair amount of interest. This five parameter family includes skew leptocurtic densities with thicker tails than the normal while still having moments of all orders (Barndorff Nielsen 1977, Eberlein and Keller 1995 and Kchler et al. 1999)