Macroeconomics of the Labor Market By Christian Merkl CES-Lecture 1: Stylized Facts of the Labor Market Munich, August 2013
Outline 1 Labor markets and the business cycle 1. Stylized facts Descriptive view on the labor market Comparison: United States and Germany Reference point for model simulations 2
Outline 2 2. Establish canonical search and matching model in partial equilibrium Derivation of key equations Model vis-à-vis data? Simulation results Simple analytics Problems and potential solutions 3
Outline 3 3. Electives (all closely related to my own research): Alternative labor market framework ( labor selection ) Monetary and/or fiscal policy and the labor market Evaluating labor market policy measures in a search and matching framework (example: German short-time work) Heterogeneity based matching function 4
Target Group of this Lecture This lecture is targeted towards PhD students / researchers who would like to get an introduction/overview about macro labor topics. may have started reading macro labor articles on their own and who would like to discuss open questions. would like to discuss potential research topics. would like to discuss about data availability / comparability / German data (although my comparative advantage is more on the theory side). 5
RULES OF THE LECTURE I assume that the background of the audience is quite heterogeneous. Thus, my slides explain many concepts in detail (e.g. the Hodrick- Prescott-filter, derivation of search and matching model). BUT if we figure out that some of these concepts are very familiar to a big majority, I can speed up or skip some parts and make more room for stuff that is more interesting to you. Please interrupt me and ask whenever necessary! 6
Trend and Cycles Many macroeconomic time series (e.g. GDP) are not stationary. Thus, establishing stylized facts based on these raw series would generate spurious correlations. In addition, business cycle researcher are not interested in the trend component ( growth theory). There are various ways of decomposing trend and cycle. The most common one is the Hodrick Prescott filter. 7
Trend and Cycles: Example 14000 US GDP 13000 12000 11000 10000 GDP 9000 8000 7000 6000 5000 4000 70 75 80 85 90 95 00 05 10 12 Year U.S. real GDP 8
Trend and Cycles: Hodrick-Prescott Filter y is the time series, g is the growth component and c is the cyclical component See Hodrick and Prescott (1997) for details 9
Trend and Cycle: The Smoothing Parameter Two different smoothing parameters, lambda=1600 (red line) and lambda=100,000 (black line) 9.8 GDP 9.6 9.4 GDP and HP-trend, in logs 9.2 9 8.8 8.6 8.4 8.2 70 75 80 85 90 95 00 05 10 12 Year Lambda infinity linear trend 10
Cycle=Ln(Y)-HP(ln(Y)) Disentangling Growth and Cycles: The Cycle 0.04 GDP 0.03 0.02 Cyclical component of GDP 0.01 0-0.01-0.02-0.03-0.04-0.05 70 75 80 85 90 95 00 05 10 12 Year Can our simple HP-filter identify booms and recessions properly? Please try to identify well-known events! 11
Trend and Cycles: Intermediate Results The HP filter is a convenient (although ad hoc) tool to decompose trends and cycles. Thus, from now onwards, I will use HP-filtered data in this stylized facts section (unless otherwise mentioned). 12
Stylized Facts 1. Stylized Facts for the United States The matching function The Beveridge curve Amplification effects The role of job findings versus separations 2. How about Germany? 13
The Matching Function ln M t 0 1 lnvt 2 lnu t where M is the number of matches, V is the number of vacancies and U is unemployment. The literature probably contains hundreds of these estimations (starting with Blanchard and Diamond 1990). Bottom line: Evidence for a matching function (i.e. beta 1 and beta2 are statistically significant). Usually evidence for Cobb-Douglas form. Often evidence for constant returns (i.e. 1 2 1 ). 14
The Beveridge Curve US-Beveridge curve from 1951 to 2003 (quarterly data). Shimer (2005, p. 30). 15
The Beveridge Curve and the Great Recession Source: https://sites.google.com/site/robertshimer/ 16
The Job Finding Rate Definition: 0,7 JFR t M U t t 1 0,6 0,5 0,4 0,3 JFR 0,2 JFR Shimer 0,1 0 1948 01 01 1950 08 01 1953 03 01 1955 10 01 1958 05 01 1960 12 01 1963 07 01 1966 02 01 1968 09 01 1971 04 01 1973 11 01 1976 06 01 1979 01 01 1981 08 01 1984 03 01 1986 10 01 1989 05 01 1991 12 01 1994 07 01 1997 02 01 1999 09 01 2002 04 01 2004 11 01 2007 06 01 2010 01 01 2012 08 01 Remark: monthly averages! For details see Shimer (2012). 17
The Separation Rate Definition: 0,06 t S N t t 1 0,05 0,04 0,03 SepQ Shimer 0,02 Sep Rate 0,01 0 1948 01 01 1951 07 01 1955 01 01 1958 07 01 1962 01 01 1965 07 01 1969 01 01 1972 07 01 1976 01 01 1979 07 01 1983 01 01 1986 07 01 1990 01 01 1993 07 01 1997 01 01 2000 07 01 2004 01 01 2007 07 01 2011 01 01 Remark: monthly averages! For details see Shimer (2012). 18
Amplification Effects Shimer (2005, p. 28). 19
What Drives Unemployment Dynamics? Labor market with two states n and u: u t sr t 1 ut 1 (1 jfrt ) ut 1 In steady state: u sr sr jfr Apply variance decomposition 20
Do Separations Matter? For many years, students of the labor market believed that recessions periods of sharply rising unemployment were the result of higher separation rates from jobs as well as lower job-finding rates. In this view, a recession begins with a wave of layoffs, mainly in cyclical durable-goods industries. As the labor market becomes clogged with job-seekers, job-finding rates go down and the duration of unemployment rises. The second part of this account is not in dispute. ( ) But new research and new data have challenged the first part. The new view is that separations are not an important part of the story of rising unemployment in recessions. Unemployment is high in a recession because jobs are hard to find, not because more job-seekers have been dumped into the labor market by elevated separation rates. Robert Hall (2006, p. 101) 21
The Hall Proposition Hall (2006). Alternatively: NBER, WP. No. 11678, p. 6. 22
Much Debate about Little Difference? Fujita and Ramey (2009) argue that separations explain between 40 and 50% of unemployment fluctuations. Shimer (2012) that the separation rate accounts for about 25%. In some robustness checks, it is somewhat more. In the end, the differences seem to be driven by different methods (e.g. related to filtering: HP versus first differences). Thus, let s conclude that the job-finding rate drives roughly one half to three quarters of unemployment fluctuations. 23
Stylized facts: Intermediate Results: Stylized Facts for the United States 1. Matching function 2. Beveridge curve (although shifted or twisted in Great Recession) 3. Amplification effects 4. Job finding rate seems to be (somewhat) more important than separation rate for unemployment fluctuations. 24
How about Germany? Interesting case: completely different labor market institutions, e.g. larger firing costs, strong union coverage. Much smaller labor market flows. Excellent administrative labor market data Problem: comparability issue (e.g. survey unemployment versus registered unemployment) 25
The Beveridge Curve Gartner, Merkl and Rothe (2009) 26
Amplification: Germany versus US Gartner, Merkl and Rothe (2012) 27
Visual Inspection Gartner, Merkl and Rothe (2012), VoxEU 28
Job Finding Rate versus Separation Rate What drives unemployment fluctuations? There seems to be even less consensus. Some attribute almost no role to the separation rate (Bachmann 2005). Others attribute a very strong role to separations (Jung and Kuhn 2009). Reason: Treatment of out of the labor force!!! 29
Stylized facts: Comparison of stylized facts: Germany versus US 1. Matching function can be found in both countries 2. Beveridge curve strong negative correlation in both countries 3. Amplification effects seem to be even stronger in Germany 4. Job-finding rate seems to be more important than separation rate for unemployment fluctuations probably most debated fact, even more so in Germany! 30
Next steps Let s derive the canonical search and matching model. How well is it able to replicate the stylized facts? 31
References Bachmann, Ronald, 2005. Labour market dynamics in Germany: Hirings, separations, and job-to-job transitions over the business cycle. SFB 649 Discussion Papers 2005-045, Humboldt University. Blanchard, Olivier, and Diamond, Peter (1990). The Cyclical Behavior of the Gross Flows of U.S. Workers. Brookings Papers on Economic Activity, Vol. 2, pp. 85 155. Gartner, Hermann, Merkl, Christian, and Rothe Thomas (2012a). Sclerosis and Large Volatilities: Two Sides of the Same Coin, Economics Letters, Vol. 117, 106-109. Gartner, Hermann, Merkl, Christian, and Rothe Thomas (2012b). "The German labour market: Low worker flows and large volatilities", VoxEU.org, 08.08.2012. Gartner, Hermann, Merkl, Christian, and Rothe Thomas (2009). They Are Even Larger! More (on) Puzzling Labor Market Volatilities, IZA Discussion Papers, No. 4403. Hall, Robert (2006): Job Loss, Job Finding, and Unemployment in the US Economy over the Past Fifty Years. NBER Macroeconomics Annual, Vol. 20, pp. 101-137. Hodrick, Robert J. and Edward Prescott (1997): Postwar U.S. Business Cycles: An Empirical Investigation. Journal of Money, Credit, and Banking, 29 (1), 1-16. Jung, Philip, Kuhn, Moritz, 2009. Explaining cross-country labor market cyclicality: U.S. vs. Germany. Mimeo. Shimer, Robert (2005). The Cyclical Behavior of Equilibrium Unemployment and Vacancies. American Economic Review, Vol. 95, No. 1, pp. 25-49. Shimer, Robert (2012). Reassessing the Ins and Outs of Unemployment, Review of Economic Dynamics, Vol. 15, 127 148. 32