What Lies Beneath: A Sub- National Look at Okun s Law for the United States. Nathalie Gonzalez Prieto International Monetary Fund Global Labor Markets Workshop Paris, September 1-2, 2016
What the paper does and why Provides estimates of Okun s Law for 51 U.S. states (we confer temporary statehood on the District of Columbia) Explores industrial structure as an explanatory variable for the cross state variation in Okun coefficients Why? Useful for state and national level design of labor market policies There are states that are bigger than some countries but they share some common features.
Okun s Law: What we estimate Gaps version Changes version u t u t = β y t y t + ε t Δu t = α + γδy t + ω t e t e t = β e y t y t + ε et Δe t = α e + γ e Δy t + ω et l l t = β l y t y t + ε lt Δl t = α l + γ l Δy t + ω lt
Summary statistics Distribution of Okun coefficients across states
Frequency Frequency Distribution of Okun Coefficient and R 2 : Unemployment, changes equation 15 Changes Specification 15 Changes Specification Adj-R2 10 10 5 5 0 -.5 -.4 -.3 -.2 -.1 0 0 0.2.4.6.8 Mean = -.25, Sd =.12, N = 51 Mean =.3, Sd =.16, N = 51
Frequency Frequency Distribution of Okun coefficient and R 2 : Employment, changes equation Changes Specification e Changes Specification Adj-R2 e 15 10 10 5 5 0 0.2.4.6.8 0 0.2.4.6.8 Mean =.46, Sd =.18, N = 51 Mean =.33, Sd =.17, N = 51
Frequency Frequency Distribution of Okun coefficient : Labor force, changes equation Changes Specification l Changes Specification Adj-R2 l 10 20 8 15 6 4 10 2 5 0 -.1 0.1.2.3.4 0 0.1.2.3.4 Mean =.18, Sd =.12, N = 51 Mean =.1, Sd =.11, N = 51
0 5 Frequency Frequency 10 15 Distribution of Okun Coefficient and R 2 : Unemployment, gaps equation Gaps Specification 15 Gaps Specification Adj-R2 10 5 -.6 -.4 -.2 0 Mean = -.29, Sd =.15, N = 51 0 0.2.4.6.8 Mean =.37, Sd =.21, N = 51
0 5 Frequency Frequency 10 Distribution of Okun coefficient and R 2 : Employment, gap equation Gaps Specification e 15 Gaps Specification Adj-R2 e 10 5 Mean =.52, Sd =.18, N = 51.2.4.6.8 0 0.2.4.6.8 Mean =.44, Sd =.19, N = 51
0 5 Frequency Frequency 10 15 Distribution of Okun coefficient and R 2 : Labor force, gap equation Gaps Specification l 20 Gaps Specification Adj-R2 l 15 10 5 -.2 0.2.4 Mean =.21, Sd =.15, N = 51 0 0.1.2.3.4.5 Mean =.17, Sd =.17, N = 51
Correlation Matrix β γ β e γ e β l γ 0.9574* 1 β e -0.6367* -0.6060* 1 γ e -0.7324* -0.7446* 0.8852* 1 β l 0.3285* 0.3195* 0.5186* 0.2737 1 γ l 0.0165 0.0458 0.6172* 0.6322* 0.7782*
Distribution matrix: Gaps High β (in absolute value) Low β (in absolute value) Low R 2 High R 2 Mississippi Alabama, California, Florida, Idaho, Illinois, Indiana, Kentucky, Michigan, Missouri, Nevada, North Carolina, Ohio, Oregon, Pennsylvania, Rhode Island, South Carolina, Tennessee, Utah, Washington, Wisconsin, New Jersey West Virginia, Alaska, Arizona, Massachusetts, Colorado, Delaware, District Arkansas, Maine, Maryland, of Columbia, Georgia, Connecticut, Minnesota, New Hawaii, Iowa, Kansas, Hampshire, Vermont, Virginia Louisiana, Montana, Nebraska, New Mexico, New York, North Dakota, Oklahoma, South Dakota, Texas, Wyoming
Industrial Structure Methodology Explaining the heterogeneity
Employment elasticity at the industry level How responsive is employment to changes in value added at the industry level? National data on value added at the industry level- VA I National data on employment at the industry level - Empl I ΔEmpl I,t = ω 0,I + ω I ΔVA I,t
Elasticities by sector - ω I Information Government Education and Health Services Other Services Leisure and Hospitality Financial Activities Trade, Transportation, and Utilities Professional and Business Services Manufacturing Construction 0.2.4.6.8
Industrial structure: State level IndStruc S = I Empl S,I Total Empl S ω I Once we have the elasticities (ω I ), we build a weighted average at the state level, using the share of employment of that industry in the states as the weighting factor. Given that the estimated elasticities are between zero and one, the industrial structure is also bound between zero and one.
Determinants Testing whether the industrial structure explains the heterogeneity of the Okun coefficient
The entrepreneurship index is the percent of individuals (ages 20-64) who do not own a business in the first survey month that start a business in the following month with 15 or more hours worked. Kauffman foundation. The data corresponds to 1996, the first year with available data Descriptive Statistics Obs Mean Std. Dev. Min Max Industrial Structure 51 0.38 0.02 0.29 0.42 Average Unemployment 51 6.10 1.14 3.56 8.30 Log-Labor Force 51 14.29 1.02 12.47 16.55 Entrepreneurial Index 51 0.00 0.00 0.00 0.01 Skill Mismatch Index 51 9.84 3.04 4.32 20.34 Oil 51 0.08 0.27 0 1
Okun coefficient and industrial structure 0.20 Industrial Strucure 0.00-0.20-0.40 DC ND AK MT HI WY SD NM NELA IA OK AR KS DE CO TX NY VT MD MN GA NH ME CT IDNV VA MA WV FL NJ AZ MS UT WA RI OR TNCA PAWI MO KY IL IN AL MI OH SC NC -0.60.3.35.4.45
Okun coefficient and average unemployment rate 0.00 ND Average Unemployment Rate AK -0.10 SD NE IA WY HI MT LA DC -0.20-0.30-0.40 NH KS VT MN VA UT OK DE CO MD CT MA WI AR TX GA NY ME NM NV ID FL NJ AZ NC WA RI OR TN CA PA MO IN KYIL MS WV -0.50 AL SC OH 4 5 6 7 8 MI
Okun coefficient and labor force 0.00 AKND Labor Force -0.10 SD MT WY DC HI NE IA LA -0.20-0.30-0.40 DE VT OK AR KS CO NH GA MN MD ME NM CT NV VA ID MA NJ WV MS AZ UT WA NC RI OR TN WI KY MO IN PA IL FL TX NY CA -0.50 AL SC MI OH 12 13 14 15 16 17
Okun coefficient and entrepreneurial index 0.00 Entrepreneurial Index ND AK -0.10 HI NE DC LASD WY MT IA -0.20-0.30-0.40 NJ AR DE TX KS NY MD NH GA MN VT CT ME VA NV MA ID WV FL MS AZ UT NC WA RI CA WI PA INIL KY MO OK NM OR TN CO -0.50 AL SC MI OH 0.002.004.006
Okun coefficient and skill mismatch index 0.00 Skill Mismatch ND AK -0.10 HI IA MT NE SD LA WY DC -0.20-0.30-0.40 CO NH VT NV KS GA MN MD CT VA MA NJ WA NC UT OR CA WI IN IL TX ME ID FL AZ RI TN PA MO AR DE MS KY OK NY NM WV -0.50 AL SC MI OH 5 10 15 20 SMI data comes from Estevão and Tsounta (2011). A Higher number indicates a higher mismatch
Employment coefficient and industrial structure e 0.80 0.60 0.40 0.20 DC Industrial Strucure AL MI FL WA IN NM MO NVOR CA OH UT AZDE ID IL GARI TN NH PAWI MS LA COVT WV VA WY MENJ KY OK NY MT MA AR TX MD MN CT HI AK KS SC NC ND SD NE IA 0.00.3.35.4.45
Employment coefficient and average unemployment rate e 0.80 0.60 0.40 0.20 NH Average Unemployment Rate VT VA UT WY MN KS HI OK DE WI CO AL FL IN WA NM MO SC NV OH ORCA AZ ID IL GA RI PA TN NC LA ME MA MD CT MT TX NJ NY AR KY MS DC AK MI WV NE SD ND IA 0.00 4 5 6 7 8
Employment coefficient and labor force 0.00 AKND Labor Force -0.10 SD MT WY DC HI NE IA LA -0.20-0.30-0.40 DE VT OK AR KS CO NH GA MN MD ME NM CT NV VA ID MA NJ WV MS AZ UT WA NC RI OR TN WI KY MO IN PA IL FL TX NY CA -0.50 AL SC MI OH 12 13 14 15 16 17
Employment coefficient and entrepreneurial index e 0.80 0.60 0.40 0.20 AL NJ Entrepreneurial Index MI RI WI PA WV HI MS OH IN DC VA NY MA AR MNMD WA SCMO CA IL DEUT NV AZ GA ID NH LANC TX CT FL KYWY ME VT KS NM OR TN OK MT CO AK NE SD IA ND 0.00 0.002.004.006
Employment coefficient and skill mismatch index e 0.80 0.60 0.40 0.20 NV NH VT CO Skill Mismatch MI AL WA FL IN NM CA SC OR OH MO IL UT AZ DE GA RI ID TN WI PA NC MS LA VA WV NJ ME KYWY OK NY MA AR CTMN MT MD TX KS HI AK DC IA NE SD ND 0.00 5 10 15 20 SMI data comes from Estevão and Tsounta (2011). A Higher number indicates a higher mismatch
Correlation Matrix Industrial Average Log-Labor Entrepreneurial Skill Mismatch Structure Unemployment Force Index Index Oil Average Employment Unemployment 0.13 1.00 Log-Labor Force 0.59* 0.28 1.00 Entrepreneurial Index -0.32-0.17-0.24 1.00 Skill Mismatch Index -0.56* 0.37* -0.26 0.02 1.00 Oil -0.18-0.05 0.01 0.09 0.24 1.00
Multivariate regressions γ γ e Industrial Structure -3.70*** -2.51*** 3.28*** 2.23* (0.53) (0.68) (0.95) (1.27) 0.09** Average Unemployment -0.05*** -0.04*** 0.09*** * (0.01) (0.01) (0.02) (0.02) Log Average Labor Force -0.07*** -0.02 0.05** -0.02 (0.01) (0.01) (0.02) (0.02) Entrepreneurial Index 41.70*** 14.00-33.27-6.49 (17.54 (13.78) (9.40) (20.67) ) Skill Mismatch 0.01* 0.00-0.01-0.01 (0.01) (0.01) (0.01) -0.01 Oil 0.14** 0.08** -0.09-0.00 (0.06) (0.04) (0.09) (0.08) Constant Standard errors in parentheses 1.16*** 0.06 0.73*** -0.39*** -0.36*** -0.26*** 1.12*** -0.79** -0.07-0.30 0.56*** 0.52*** 0.46*** -0.59 *** p<0.01, ** p<0.05, * (0.20) p<0.1 (0.09) (0.21) (0.05) (0.06) (0.02) (0.27) (0.36) (0.11) (0.34) (0.07) (0.08) (0.03) (0.50) Observations 51 51 51 51 51 51 51 51 51 51 51 51 51 51 R-squared 0.50 0.22 0.32 0.16 0.07 0.09 0.70 0.20 0.31 0.09 0.05 0.01 0.02 0.48 Adjusted R-squared 0.49 0.20 0.30 0.14 0.05 0.07 0.66 0.18 0.30 0.08 0.03-0.01 0.00 0.40 Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1
Conclusions The Okun coefficient is heterogeneous within the country Industrial structure explains a big part of this heterogeneity The degree of employment responsiveness to changes in value added accounts for 50% of the heterogeneity across states This relationship holds even when we include other explanatory variables