Sometimes, you feel you're spoilt for choice: there are so many good predictors that you could use! Why select and focus on just one? I will review the framework of robust online aggregation (also known as prediction of individual sequences or online aggregation of expert advice). This setting explains how to combine base forecasts provided by ensemble methods. No stochastic modeling is needed and the performance achieved is comparable to the one of the best (constant convex combination of) base forecast(s). I will illustrate the technology on various data sets, including electricity consumption and exchange rates. More importantly, I will point out open issues, both on the theoretical and on the practical sides.[-]

In his 1947 essay, Tjalling Koopmans criticized the development of an empirical science that had no theoretical basis, what he referred to as measurement without theory. The controversy over the status of relations based on mere statistical inference has not ceased since then. Instead of looking for the contemporary consequences, however, I will inquire into its early beginnings. As early as the 1900s, Walras, Pareto and Juglar exchanged views on the status of theory and its relation to economic data. These private exchanges acquired the status of scientific controversy in the aftermath of the First World War, with the dissemination of Pareto's work. It is precisely this moment that I will try to grasp, when engineers began to read and write pure economic treatises, questioning the relation between theory and empirical problems, the nature of their project and the expectations that the subsequent development of economics has tried to fulfill.

Cournot Centre session devoted to the transformations that took place in mathematical economics during the interwar period.[-]

In order to explore the advances made on the economic issue of business cycles, I will present the work of the American economist Henry Ludwell Moore, who published four works on the question between the years 1911 and 1923. Within this framework, I will introduce several issues, such as the duality of empirical and theoretical approaches, the causal and semiological interpretations of the correlation, the notion of the ceteris paribus law in economics, the notion of non-probabilistic statistical mathematics, the development of the notion of the dynamic model at the end of the 1920s, the diverse analysis techniques of chronological series and their artefacts. I will also make reference to the contributions of other actors in this field.

Cournot Centre session devoted to the transformations that took place in mathematical economics during the interwar period.[-]

Data mining methods based on finite mixture models are quite common in many areas of applied science, such as marketing, to segment data and to identify subgroups with specific features. Recent work shows that these methods are also useful in micro econometrics to analyze the behavior of workers in labor markets. Since these data are typically available as time series with discrete states, clustering kernels based on Markov chains with group-specific transition matrices are applied to capture both persistence in the individual time series as well as cross-sectional unobserved heterogeneity. Markov chains clustering has been applied to data from the Austrian labor market, (a) to understanding the effect of labor market entry conditions on long-run career developments for male workers (Frühwirth-Schnatter et al., 2012), (b) to study mothers' long-run career patterns after first birth (Frühwirth-Schnatter et al., 2016), and (c) to study the effects of a plant closure on future career developments for male worker (Frühwirth-Schnatter et al., 2018). To capture non- stationary effects for the later study, time-inhomogeneous Markov chains based on time-varying group specific transition matrices are introduced as clustering kernels. For all applications, a mixture-of-experts formulation helps to understand which workers are likely to belong to a particular group. Finally, it will be shown that Markov chain clustering is also useful in a business application in marketing and helps to identify loyal consumers within a customer relationship management (CRM) program.[-]