Comments on Modelling the business cycle in a multivariate - PowerPoint PPT Presentation

Comments on “ Modelling the business cycle in a multivariate structural time series model for the G7 and the Netherlands Gabriel Perez Quiros The usual disclaimer applies

Brief summary. Specification • Structural time series model that decompose the level of four time series in trend, medium term cycle, short term cycle and noise. • Eight countries. Housing prices, credit, IPI and GDP.

Brief summary. Results • Cyclical movements in financial variables driven by the medium term cycle • Macro-economic variables driven by medium and short term cycle • Comovement between macro and financial variables only in medium term cycle • Different medium term cycle of the financial variables across countries.

Detailed Results. Short term cycle • Italy 3.1 years. The Netherlands, 8.2 years • No common cycle for GDP and housing price for six countries. UK and Canada • No common cycle for credit and GDP. Only in the Netherlands. • No common cycle for credit and IP Only in Italy.

Detailed Results. Medium term cycle • Germany 8.5 years. The Netherlands, 20.6 years • For IP and GDP short term and medium term equally important. • Financial variables. Medium term • Cocyclicality house price and GDP and credit and GDP for most countries except Canada and Germany • For Italy and the Netherlands, indirect commonality through IP or housing prices

Main comments • The paper is still to be written. • The paper promises good results. • Technically, the paper is extremely well done. • Understand the relation between real and financial variables is interesting.

Main concerns • What do we learn from the paper? The results are spread across countries and it is difficult to obtain lessons of interest about the characteristics of each country. – Italy 3.1 years. The Netherlands, 8.2 years – No common cycle for GDP and housing price for six countries. UK and Canada – No common cycle for credit and GDP. Only in the Netherlands. – No common cycle for credit and IP Only in Italy.

Main concerns • What do we learn from the paper? The results are spread across countries and it is difficult to obtain lessons of interest about the characteristics of each country. – Germany 8.5 years. The Netherlands, 20.6 years – For IP and GDP short term and medium term equally important. – Financial variables. Medium term – Cocyclicality house price and GDP and credit and GDP for most countries except Canada and Germany – For Italy and the Netherlands, indirect commonality through IP or housing prices

Main concerns • Usefulness of the results? • Forecasting? • Policy analysis? Policy recommendations? • Theoretical models to validate?

Technical concerns. Robustness • Selection of the variables. – From the real side: • Sales, Income, Employment (Stock and Watson, 1991) – From the financial side: • Debt Insuance (Lopez Salido, Stein and Zakrajsek, 2014) • Credit spreads (Gilchist and Zakrajsek, 2012) • Term spreads (Estrella and Mishkin 1992) • Equity sentiment (Baker and Wurgler 2006) • Credit/Asset (Drehman and Juselius 2012) • Stock market

Technical concerns. Robustness • Transformation of the variables - Credit to GDP Gourinchas and Obsfeld (2011) - Credit to GDP variation (IMF and Jorda et al 2011) - Credit intensity (Jorda et al 2011) accumulated difference between credit and GDP growth/ N - Threshold values (Mendoza and Terrones, 2008) - Order of integration of credit. Variation in credit should be related to GDP.

Technical concerns. Robustness • More on transformation of the variables – Credit to GDP accumulates over time endogenously in different theoretical models, as in Gertler & Karadi (2011), Gertler & Kiyotaki (2010), Christiano, Motto & Rostagno (2010) or Nuño and Thomas (2012), and, therefore, it is endogenously high when the expansions are long.

US Credit/GDP ratio

Credit to GDP Trending expansions

Credit to GDP Trending expansions

Technical concerns. Robustness • Correct specification. – Independent trends. Same type of trends? Dependent Medium and short run cycles. Diagonal weights.

Technical concerns. Robustness • Sample period. – Is it the last recession what is driving the common behavior of the credit and GDP? – Some facts:

Technical concerns. Robustness a) For the sample of 39 OECD countries, between 1950.q1 and 2011.q2, we identify 149 recession periods. Out of these, only 45 coincide financial crises documented by Gourinchas and Obstfeld (2011), and 31 of them correspond to the recent crisis b) For this sample Gourinchas and Obstfeld (2011) identify 143 financial crises, of which only 45 correspond to a real crisis c) Eliminating the last 31 recent crises out of the 230 financial or real crises (143-31 financial , 149-31 real), we find that only 14 cases (6%) are both financial and real Definitely, too few coincidences to make a link

Technical concerns. Robustness • Ex-ante Ex-Post (Gadea and Perez Quiros, 2014) – Much has been written about why economists failed to predict the latest crisis. Reading the literature, it seems that this crisis was so obvious that economists must have been blind not to see it coming. We approach this failure by looking at one of the key variables in this analysis, the evolution of credit. We compare the conclusions reached in the recent literature with those that could have been drawn from an ex ante analysis. We show that the effect of credit on the business cycle can not be exploited from a policymaker’s point of view.

One hypothetical question • If changing a little the variables, the model, the transformation or the variables… etc … the results change and some cycles are more related across the different variables…. what do we learn? Do our policy analysis is different?

Something to remember • Because ….as of today … the academic knowledge seems to be: • IMF Global Financial Stability Report. September 2011 • "Credit to GDP growth is a particularly reliable indicator of recession when the experiences of both advanced and emerging economies are considered together

Something to remember • And the question of interest should be: - Does the level of credit to GDP (or its variation) observed in period "t" increase the probability of being in a recession in "t+1“? - Does that level affect the characteristics of future cyclical phases?

Something to remember If the answer to this questions is YES…credit is a variable to control…otherwise, the policymakers could cut good and healthy expansions in order to avoid a not- forthcoming recession

To Conclude • I would love to put their proposed specification to work to answer these applied questions that are key for the design of future stability policies.