Gradient boosting for distributional regression: faster tuning and improved variable selection via noncyclical updates

We present a new algorithm for boosting generalized additive models for location, scale and shape (GAMLSS) that allows to incorporate stability selection, an increasingly popular way to obtain stable sets of covariates while controlling the per-family error rate. The model is fitted repeatedly to subsampled data, and variables with high selection frequencies are extracted. To apply stability selection to boosted GAMLSS, we develop a new “noncyclical” fitting algorithm that incorporates an additional selection step of the best-fitting distribution parameter in each iteration. This new algorithm has the additional advantage that optimizing the tuning parameters of boosting is reduced from a multi-dimensional to a one-dimensional problem with vastly decreased complexity. The performance of the novel algorithm is evaluated in an extensive simulation study. We apply this new algorithm to a study to estimate abundance of common eider in Massachusetts, USA, featuring excess zeros, overdispersion, nonlinearity and spatiotemporal structures. Eider abundance is estimated via boosted GAMLSS, allowing both mean and overdispersion to be regressed on covariates. Stability selection is used to obtain a sparse set of stable predictors.     

Innovation and Sectoral Employment: A Trade–off between Compensation Mechanisms

The question whether technological progress displaces employment or whether technological advance is beneficial for the level of employment has been at the core of economic debate for over two centuries. The beneficial effect might be achieved by several compensation mechanisms within the economic system. In this paper we categorize these compensation mechanisms into two basic categories that reflect the different nature of the ideas ruling the compensation. We discriminate the mechanisms of employment despite innovation from employment via innovation. In the context of new innovation economics we model an artificial industry implementing both compensation mechanisms. Simulation analysis is used to examine both the short–run and long–run properties of the model. There we focus on the influence of wage restraint policy on the functioning of the compensation mechanism.     

Co-determination and Personnel Turnover: The German Experience

ABSTRACT: It has often been argued, that within the German system of co-determination the works council has an unparalleled participative role, which goes well beyond any voice function of trade unions in the Anglo Saxon tradition. Nevertheless, there is surprisingly little empirical evidence, whether and to what extent works councils influence employers’dismissal and employees’quit decisions. Analyzing a representative sample of private sector firms that have to obey the Dismissal Protection Act of 1969 and the Works Constitution Act of 1972, the paper finds that firms with a works council have an annual dismissal rate which is 1.6 percentage points lower than the one experienced by firms wthout a plant-level interest representation. At the same time, firms with a works council have a quit rate which is 1.5 percentage points lower than the one in firms without such interest representation. Union density, on the other hand, has no statistically significant influence on either the dismissal or the quit rate. Apparently, it is the works councils that in both cases act as a collective “voice institution”, serving as a direct channel of communication between workers and management. Although it is difficult to reach a safe conclusion about the net economic impact of works councils on the employment behavior of firms, the findings lend some support to the following assumption: since hiring and training costs are usually higher than firing costs, firms on average benefit from the presence of a works council with regard to their user costs of labor: the “savings” due to avoided voluntary quits apparently more than compensate for the additional spendings for severance payments and the costs of co-determination.