Aim of the paper Model Model Properties Data Rich Enviroment Other Minor Suggestions Discussion: Jan Br˚uha - A Small Labor Market Model for the Czech Economy Martin Slanicay 23.10.2010 Aim of the paper Model Model Properties Data Rich Enviroment Other Minor Suggestions Contents 1 Aim of the paper 2 Model General Assumptions Trend Component Cyclical Component 3 Model Properties Multivariate Filtering Second Moments Forecasting 4 Data Rich Enviroment 5 Other Minor Suggestions Aim of the paper Model Model Properties Data Rich Enviroment Other Minor Suggestions Aim of the paper a small model for joint modelling of the labor force, employment, wages, hours worked, output and GDP deflator model properties examination multivariate filtering second moments (data observations vs model implications) forecasting extension to the data rich enviroment Aim of the paper Model Model Properties Data Rich Enviroment Other Minor Suggestions Model - General Assumptions every variable is decomposed into the trend component and the cyclical component the trend component and the cyclical component are modelled jointly in the state-space framework - big advantage no data prefiltering before estimation (an arbitrary choice of detrending method can significantly change the behaviour of the model and obtained results can substantially differ) all obtained trends are filtred jointly and thus are consistent with each other, while those obtained by univariate filters can be inconsistent. the model is estimated by the prediction-error minimization method Aim of the paper Model Model Properties Data Rich Enviroment Other Minor Suggestions Model - Trend Component [1/2] based on log-linear specifications of production function and a labour demand equation every state variable can be described in the long run by some combination of trends in labor force, labor productivity, unemployment, hours per employee and GDP deflator. trends are modelled as ARIMA(1,1,0) (in contrast to Harvey and Jaeger (1993) framework where ARIMA(0,2,0) is used). Aim of the paper Model Model Properties Data Rich Enviroment Other Minor Suggestions Model - Trend Component [2/2] it allows for fluctuations of drifts in trends around some certain values, while Harvey and Jaeger model of trends does not and implies random walks for these drifts trend in labor productivity growth Harvey and Jaeger model implies that if growth in labor productivity becomes negative, all its future expected values will remain negative. This model implies that gradual recovery of this growth will happen. inflation trend in Harvey and Jaeger model follows random walk in this model fluctuate around inflation target In my opinion, this modification of modelling trends in time series has convincing economic reasoning and is one of the main contributions of this paper. Aim of the paper Model Model Properties Data Rich Enviroment Other Minor Suggestions Trend Component - Suggestions modification of the long run dynamics of unemployment in order to incorporate some long run hysteresis in the development of the unemployment including the capital into the production function and modelling the development of capital stock Aim of the paper Model Model Properties Data Rich Enviroment Other Minor Suggestions Model - Cyclical Component The cyclical dynamics of the model is determined by a VAR(3) model. Aim of the paper Model Model Properties Data Rich Enviroment Other Minor Suggestions Multivariate Filtering comparison of model-based filter with standard HP filter and HP filter with end-point bias reduction consistency of filtered trends from the model-based filter (in contrast to univariate HP filters) subject to smaller revisions than those obtained by HP filters more negative cyclical position in the current recession than HP filters the model-based filter also provides a better story for development of the inflation in the period 1998-2000 comparison is interesting maybe more interesting would be comparison of suggested filter with some multivariate filters, as univariate HP filters are not formidable rivals Aim of the paper Model Model Properties Data Rich Enviroment Other Minor Suggestions Second Moments [1/2] data observations vs model implications comparison of various cross-correlations, cospectra, quadrature spectra, sample coherences and population cross-corelations It seems that this model can replicate most of the business cycle features in the data. most problematic area - the relation between real wage, employment and hours worked The model can not explain some of the long run movements in these variables. Aim of the paper Model Model Properties Data Rich Enviroment Other Minor Suggestions Second Moments [2/2] problem of all the models with neoclassical view on labor market in the long run According to this view, unemployment is given in the long run by its natural rate, which is determined purely by institiutional factors while cyclical factors play no role. model predict no long run hysteresis of unemployment at odds with empirical evidence (as pointed out by author) my previous suggestion - modify long run dynamics of unemployment Aim of the paper Model Model Properties Data Rich Enviroment Other Minor Suggestions Forecasting experiment with conditional forecast How would model’s forecast of labor market variables development look like, if we had known the real decline in real output during recent recession. forecasting power is not bad (despite the model simplicity) model can do a good job in forecasting of some variables (real wage, hours worked and inflation) I consider this experiment very interesting. Aim of the paper Model Model Properties Data Rich Enviroment Other Minor Suggestions Data Rich Enviroment extension of the model to data-rich enviroment incorporate a large set of additional data in order to improve forecasting power of the model very difficult estimation technique with many steps only marginal improvement and only for a subset of variables possible reason - short length of available data for CZ interesting and very sophisticated tool which can prove to be useful in the future Aim of the paper Model Model Properties Data Rich Enviroment Other Minor Suggestions Other Minor Suggestions several extensions of the paper, mainly for reader’s convenience to show some non-technical interpretation of cospectrum, quadrature spectrum, sample coherence and population cross-corelations (more friendly for readers who are not so familiar with spectral analysis of time series) to show some technical details about the estimation technique (prediction-error minimization method) to show and interpret some of the impulse response functions Aim of the paper Model Model Properties Data Rich Enviroment Other Minor Suggestions Conclusion elegant model of labor market big advantage of modelling the trend component and the cyclical component jointly in one model reasonable specification of trends extension to the data rich enviroment - very interesting tool all suggestions should be regarded as topics for further research Aim of the paper Model Model Properties Data Rich Enviroment Other Minor Suggestions Thank you for your attention.