Forecasting Macroeconomic Labour Market Flows: What Can We Learn from Micro‐level Analysis?
| Published date | 01 August 2018 |
| DOI | http://doi.org/10.1111/obes.12222 |
| Date | 01 August 2018 |
| Author | Ralf A. Wilke |
822
©2017 The Department of Economics, University of Oxford and JohnWiley & Sons Ltd.
OXFORD BULLETIN OF ECONOMICSAND STATISTICS, 80, 4 (2018) 0305–9049
doi: 10.1111/obes.12222
Forecasting Macroeconomic Labour Market Flows:
What Can We Learn from Micro-levelAnalysis?*
Ralf A. Wilke†
†Department of Economics, Copenhagen Business School, Porcelænshaven 16A, 2000
Frederiksberg, Denmark (e-mail: rw.eco@cbs.dk)
Abstract
Forecasting labour market flows is important for budgeting and decision-making in gov-
ernment departments and public administration. Macroeconomic forecasts are normally
obtained from time series data. In this article, we follow another approach that uses
individual-level statistical analysis to predict the number of exits out of unemployment
insurance claims. We present a comparative study of econometric, actuarial and statistical
methodologies that base on different data structures. The results with records of the Ger-
man unemployment insurance suggest that prediction based on individual-level statistical
duration analysis constitutes an interesting alternative to aggregate data-based forecasting.
In particular, forecasts of up to six months ahead are surprisingly precise and are found to
be more precise than considered time series forecasts.
I. Introduction
Labour market analysis on the individual level normally uses individual-level data and
macro-level analysis bases on aggregate data (see e.g. Barnichon and Nekarda, 2012).
Until recently, such a separation was natural due to restrictions on data availability and
computing performance. Although surveys, such as the labour force survey, are frequently
used to validate or construct macroeconomic figures such as the unemployment rate or
labour market flows (see e.g. Elsby, Hobijn and Sahin, 2015; Hutter and Weber, 2015),
econometric forecasting of macroeconomic figures is typically based on aggregate data
(examples include Brown and Moshiri, 2004; Sermpinis et al., 2014; Hutter and Weber,
2017). The increased availability of large linked administrative individual data opens up
new opportunities for micro-level statistical analysis as information on individual level is
becoming availablefor the population. While these data have become the industry standard
for applied analysis on individual level,little efforts have been devoted to link the statistical
analysis on individual level with forecasts for the macro level.
JEL Classification numbers: C53, C55, J60.
*Thanks are due to Jens P. Nielsen, Lola Mart´ınez, H´ector C. Villanova and EnzoWeber for helpful discussions
and to the former two for making their R code for the chain ladder model (-to be released as package ‘Double Chain
Ladder-) available.The comments of two reviewers have been also gratefully acknowledged.
Micro data based labour market forecasts 823
Economic literature on linking microeconometric analysis with macro models is sparse.
Partly, this is done in microsimulation models. However, these models are less a statistical
approach but more a computational tool in applied economics.Their core is simulating par ts
of the economy using models based on economic theory. More related to the goals of this
study, new research in actuarial sciences has developed forecasts for aggregate figures on
the grounds of micro-data-level analysis (compareAntonio and Plat, 2014). This approach
has also been shown to be useful in estimating outstanding claim liabilities in the disability
insurance (Spierdijk and Koning,2014) and forecasting mesothelioma mor tality (Mart´ınez-
Miranda, Nielsen and Nielsen, 2015). This article adopts the idea of using individual-level
data to construct forecasts for macroeconomic figures. The considered models exploit the
richness of the data and avoidparametric restrictions on the micro level as much as possible.
An empirical study is presented which compares forecasts obtained by differentapproaches
including a classical time series forecast. Extensive unemploymentinsurance claim records
from Germany are used to predict the number of unemployment benefits leavers. Within
this application, it is shown how macroeconomic labour market forecasts are obtained
from estimated individual-level transition probabilities.The latter are estimated using data
about individual employment biographies in past periods. By focusing on exits out of
the unemployment insurance, we consider a standard problem in actuarial sciences, the
so-called reserving problem. There the insurer uses data on past claims and contract signing
dates to produce a forecast for future insurance claims or outstanding liabilities. Using a
simple data structure in form of a triangle, these so-called chain ladder method (CLM)
(compare Weindorfer, 2012) are used by most if not all insurers to estimate outstanding
liabilities. Thus, these estimates are very important for the decision about the adequate size
of financial reserves.The statistical proper ties of these actuarial methods are welldeveloped
(Kuang, Nielsen and Nielsen, 2009; Pigeon,Antonio and Denuit, 2013). Str uctured density
forecasting is adopted in this study as it uses individual-level data to construct in-sample
forecasts for the aggregate. In contrast to time series models, these forecasts do not simply
extrapolate an aggregate figure but use as much as possible individual-level information at
the edge of the observation period to construct the forecast. As third contender, statistical
duration models are used to construct macro-level forecasts. The idea here is to estimate
individual-level probabilities for existingunemployment insurance claims. Thus, this work
puts an interdisciplinary method mix to data to analyse the same problem and to explore
how estimation results compare. A rather non-technical presentation of the material is
chosen to make the material attractive for practitioners.
The paper is structured as follows. Section II briefly presents the economic and institu-
tional framework for the empirical analysis. Section III provides details about the admin-
istrative data used in the analysis. Section IV describes the methodologies and section V
presents the forecasting results. Section VI summarizes the main findings and outlines
trajectories for further improvements.
II. Unemployment insurance in Germany
Precise prediction of the number of unemployment benefit leavers is not only impor-
tant for policy makers to be able to build reliable expectations about state of the labour
market but also for financial planning units within the unemployment insurance. In
©2017 The Department of Economics, University of Oxford and JohnWiley & Sons Ltd
To continue reading
Request your trial