New
Approaches to the Study of Economic Fluctuations
Training and Mobility of Researchers
Network
Workshops and Conferences
A third network conference was co-organised by Graziella Bertocchi (Modena) and Helmut Luetkepohl (Humboldt) and took place in Hydra (Greece) on 11-12 May 2001. Fifteen network team-members participated as presenters or discussants, of which five were current or former Young Researchers hired by the network.
Hydra 1999
The first major Network meeting took place in May 1999 at Hydra, hosted by the Greek partner and organised by Harald Uhlig (Tilburg University and CEPR) and Vanghelis Vassilatos (IMOP). The workshop focussed on new approaches to the study of the business cycle with an emphasis on aggregate and disaggregate analysis, the analysis of large cross-sections, estimation and calibration techniques, and also monetary, public finance, growth and income distribution issues. Five papers were presented by Network members:
Macroeconomists typically write down fully specified models in which a few shocks drive the whole economy. These shocks should then be able to explain most features of the data. However, the number of shocks, as well as their economic interpretation depends upon the particular economic theory under consideration. The paper presented by Harald Uhlig (Tilburg University and CEPR), and co-authored by Lucrezia Reichlin (ECARES, Université Libre de Bruxelles and CEPR), entitled ‘What are the Main Macroeconomic Forces’, proceeds the other way around as it tries to infer, directly from the data, both the relevant number of shocks and an economic interpretation for these shocks. It devises a new methodology and applies it to a large macroeconomic dataset for the US.
First, dynamic principal component analysis reveals that only two shocks are needed to explain 95% of the total variance at any frequency. Then, three different techniques are suggested in order to identify these shocks. The first two techniques are conceptually similar: the first seeks to explain all of the variance of a single variable while the second seeks to explain as much as possible of the variance of all the variables. Both begin by running a VAR and then selecting the two shocks that explain the largest part of the prediction error variance at some predetermined horizon. The impulse-response functions for these shocks are then computed and their shape drives the interpretation of the underlying shocks. These two methods yield qualitatively similar and rather disappointing results: except for the particular horizon at choice, the two shocks only explain half, or less, of the prediction error variance. The third method takes a different approach: it attempts to identify two shocks that explain most of the variations in all the variables when averaging over a specific frequency band. These results are much more encouraging since the two shocks seem to explain practically all the variance at all horizons for all variables.
The paper presented by Helmut Lütkepohl (Hümboldt Universität zu Berlin), ‘Comparison of Bootstrap Confidence Intervals for Impulse Responses of German Monetary Systems’, which was written with Alexander Benkwitz and Jürgen Wolters, asked methodological questions about bootstrap confidence intervals for impulse-response functions in VAR systems. The methodology was illustrated on two VAR models aimed at investigating whether money supply shocks have an impact on inflation, as argued by the Bundesbank in support for its monetary targeting strategy. The first model was a cointegrated M1 system. An impulse-response analysis showed that the impact of monetary shocks on prices was very weak. According to standard bootstrap confidence intervals, this effect appears non-significant. But naive bootstrap strategies prove inadequate for estimating non-linear functions of estimated parameters, such as impulse-response functions, as the resulting confidence intervals are too wide, and the use of more sophisticated bootstrap methods did not alleviate this problem. As a solution, Lütkepohl proposed to estimate restricted systems using guidance from economic theory. This resulted in substantially narrower confidence intervals, since imposing a-priori restrictions removes a large part of estimation uncertainty. Monetary impulses then appeared to have a significant impact on prices. The same analysis was performed on a cointegrated German M3 system with a few more variables. Estimation of a restricted model yielded some support for the arguments of the Bundesbank: the confidence intervals showed that impulses in money or in the interest rate differential may have significant effects on the inflation rate. Lütkepohl concluded that the use of restricted VAR systems for impulse-response analysis was the best available method. However, he noted that the asymptotic and small sample properties of the resulting bootstrap confidence intervals are not fully understood yet, and some additional research should be performed.
‘An Empirical Study of the Cyclical Effects of Monetary Policy in Spain (1977-1997)’, written with Ramón Maria-Dolores, was presented by Juan-José Dolado (Universidad Carlos III and CEPR), in which he investigated the possible state asymmetries in the effects of monetary policy in Spain. He noted that there exists various types of monetary policy asymmetries which have been widely studied: the sign asymmetry, i.e. between positive and negative shocks due to one-way rigidities, and the size asymmetry, i.e. between the effects of large versus small shocks. However, there are very few instances on the state asymmetry, which relates to the issue of whether monetary policy affects real activity differently in different phases of the business cycle. The paper’s aims were to provide empirical evidence of such state asymmetries in Spain over the period 1977-1997, using both aggregate and sectoral data. Since there exists no dating of business cycle turning points in Europe, Dolado turned to a Markov Switching Model, in order to endogeneously determine from the data the dates and transition probabilities from one cyclical phase to another. Monetary policy shocks were extracted from a 6-variable VAR. His estimations showed that monetary policy shocks prove significantly more effective during recessions than during expansions: an 130 basis points increase in the marginal intervention rate can decrease output by 0.5% at the trough of a recession and by 0.1% at the peak of an expansion. Moreover, monetary policy shocks may have a direct impact on the probability of switching from one phase to another. For example, during an expansion three consecutive increases in the intervention rate by 15 basis points increase the probability of moving to a recession from 0.32 to 0.44. Dolado then turned to a disaggregate analysis using sectoral data, where he found that state asymmetries prove much larger in Construction and Services than in Agriculture or Manufacturing.
The high level of European unemployment is often attributed to labour markets rigidities but they may not be the sole reason. The paper presented by Etienne Wasmer (ECARES, Université Libre de Bruxelles, and CEPR), co-authored by Philippe Weil and called ‘The Macroeconomics of Labour and Credit Market Imperfections’, suggested that the interaction between labour market rigidities and capital market rigidities might be responsible for a large part of European unemployment. Wasmer builds a model of job creation/destruction with rigidities on both capital and labour markets. These rigidities are modelled in a symmetric way across both markets, and are summarised by two search processes. According to simulations, capital market imperfections appear as an interaction term: the higher the unemployment rate in absence of credit frictions, the larger the increase in unemployment stemming from credit frictions. A comparative statics analysis yielded three main results. First, in this model, a looser monetary policy increases output and lowers unemployment. Also, monetary policy appears more effective when credit market frictions are high. Second, the impacts of lower search costs for entrepreneurs are a fall in unemployment but an ambiguous response of credit market tightness. Third, an exogenous increase in firms net profits generates a mitigated impact on capital market tightness, but a large fall in unemployment through the interaction between capital market and labour market frictions, i.e. there is a profit multiplier.
‘Complementarity of Labour Market Institutions, Equilibrium Unemployment, and the Persistence of the Business Cycle’ was the title of the paper presented by Michael C. Burda (Humboldt Universität Zu Berlin and CEPR) and Mark Weder (Humboldt Universität Zu Berlin and CEPR). Conventional wisdom often blames institutions, such as social security or wage bargaining mechanisms, for the persistence of high unemployment rates in Europe. However, these institutions already existed in the 1960s, when unemployment was low. To solve this puzzle, the authors propose a model in which exogenous events and an economy with multiple equilibria interact. They constructed a dynamic general equilibrium model in which labour market institutions are complementary. Central features of the model are matching between workers and vacancies with endogenous search intensity, Nash-bargained wages, payroll taxation, and differential support for search and leisure unemployment. Parameter values were calibrated so that the model’s long-run properties correspond to the growth path of post-war Germany. Four parameters, however, are not fixed a-priori: the degree of returns to scale, the replacement rate, the degree of welfare generosity towards leisure unemployment, and the labour supply elasticity. For reasonable parameter values the model could exhibit indeterminacy without resorting to increasing returns, provided tax distortions and social security payments are introduced. Simulation results showed that the model was capable of replicating most business cycles facts without the need to assume technology shocks as the main driving force. ‘Pure sunspot shocks’ were able to explain the facts. Along several dimensions, the model fares better than the standard RBC approach. In particular, the introduction of institutions induces additional output persistence, which results in a closer fit to the actual data. The model yields a strong policy implication: only global labour market reforms can have significant impacts.
Second
full Network Meeting
The second network workshop took place in Hydra in May 2000 and was organised by Juan José Dolado (U. Carlos III and CEPR) and Harris Dellas (IMOP and CEPR).
The programme of this workhop is available as a pdf file (33k)
Return to Introduction
Workshops and Conferences