19668 ISeptember 1999 Labor Market Flexibility in Thnirteen Latin American Countries and the United States Revisiting and Expanding Okun Coefficients gr35 [3rSt[0(3 (tO6e3g3i3tleDG3 1-~~~~~~~~~~~~~zdlW~,po WORLD BANK LATIN AMERICAN AND CARIBBEAN STUDIES Vziew0points Labor Market Flexibility in Thirteen Latin American Countries and the United States Revisiting and Expanding Okun Coefficients Jose Antonio Gonzdilez Anaya The [It orld Bank WaVshington, D.C. ©) 1999 The International Bank for Reconstruction and Development / the world bank 1818 H Street, N.W Washington, D.C. 20433, U.S.A. All rights reserved Manufacttired in the United States of America First printing September 1999 The findings, interpretations, and conclusions expressed in this paper are entirely those of the author(s) and should not be attributed in any manner to the World Bank, to its affiliated organizations, or to members of its Board of Executive Directors or the countries they represent. The World Bank does not guarantee the accuracy of the data included in this publication and accepts no responsibility whatsoever for any consequence of their use. The boundaries, colors, denominations, and other information shown on any map in this volume do not imply on the part of the World Bank Group any judgment on the legal sta- tus of any territory or the endorsement or acceptance of such boundaries. The material in this publication is copyrighted. Requests for permission to reproduce portions of it should be sent to the Office of the Publisher at the address shown in the copyright notice above. The W\Aorld Bank encourages dissemination of its work and will normally give permission promptly and, when the reproduction is for noncommercial purposes, without asking a fee. Permission to copy portions for classroom use is granted through the Copyright Clearance Center, Inc., Suite 910, 222 Rosewood Drive, Danvers, Massachusetts 01923, U.S.A. The complete backlist of publications from the World Bank is shown in the annual Index of Publications, which contains an alphabetical title list (with full ordering information) and indexes of subjects, authors, and countries and regions. The latest edition is available free of charge from the Distribution Unit, Office of the Publisher, The World Bank, 1818 H Street, N.W., Washington, D.C. 20433, U.S.A., or from Publica- tions, The World Bank, 66, avenue d'lena, 75116 Paris, France. Library of Congress Cataloging-in-Publication Data Gonzalez Anaya, Jose Antonio, 1967- Labor market flexibility in thirteen Latin American countries and the United States / Jose Antonio Gonzalez Anaya. p. cm. - (World Bank Latin American and Caribbean studies) Includes bibliographical references. ISBN 0-8213-4489-7 1. Labor market-Latin America. 2. Labor market-United States. 1. Title. II. Title: Labor market flexibility in thirteen Latin American countries and the United States. III. Series. HD5730.5.A6G66 1999 331.12'098-dc2l 99-15436 CIP 09 ......... AHdVAdDoIohIl .I .. 1 XICN9ddV ....S....... . ...................... SNOISfYIDNO) A L ...........IdIIlO (INV IN? O'Ida NIA SdIHSNOILVqT C1VflNlIDfYUd IS WN[II-9lNEY Al L ....I...lO .Nf 'SZIDVA\ 'IN1IV'LAO'Id1JNflI CNV ILNHW1AOIdINa DNOINV SNOIIVIVVA IVDIIDAD 111 ................... DNVWdOddJ VI o0V1 0 SThfISVIJV SV INJNAO~IdVMHNf1 CINV 'IN'INXO'1dV\ 'SADVA 'IVTh 'II SI NVI NO9 It'L'FWa1AN9 lv TABLES Table 3.1. Unemployment Okun Coefficients for Some OECD Countries .................I.......... 21 Table 3.2 Summary of Cyclical Results Based on Okun Ratios .............................................. 25 Table A.2. 1. Regression Results for the Last Ten-Year Window .................................................... 48 FIGURES Figure 3.1 Ten-Year Rolling Wage, Employment and Unemployment Okun Ratios: The Exception ................................................. 13 a. Ten-Year Rolling Wage Okun Coefficients: Argentina, Bolivia, Chile, Costa Rica, Peru, and the United States (HP50) .............................................. 13 b. Ten-Year Rolling Employment Okun Coefficients: Argentina, Bolivia, Chile, Costa Rica, Peru, and the United States (HP50) ...................................... 14 c. Ten-Year Rolling Unemployment Okun Coefficients: Argentina, Bolivia, Chile, Costa Rica, Peru, and the United States (HP50) ............... ........ 14 d. Ten-Year Rolling Wage Okun Coefficients: Brazil, Venezuela, and the United States (HP50) ................................................ 15 e. Ten-Year Rolling Employment Okun Coefficients: Brazil, Venezuela, and the United States (HP50) ............................................... 15 f Ten-Year Rolling Unemployment Okun Coefficients: Brazil, Venezuela, and the United States (HP50) ............................................... 16 g. Ten-Year Rolling Wage Okun Coefficients: Mexico and the United States (HP50) ................................................ 16 h. Ten-Year Rolling Employment Okun Coefficients: Mexico and the United States (HP50) ................................................ 17 i. Ten-Year Rolling Unemployment Okun Coefficients: Mexico and the United States (HP50) ................................................. 17 j. Ten-Year Rolling Wage Okun Coefficients: Colombia, Panama, Uruguay, and the United States (HP50) .............................................. 18 k. Ten-Year Rolling Unemployment Okun Coefficients: Colombia, Uruguay, Panama, and the United States (HP50) .............................................. 18 v 1. Ten-Year Rolling Employment Okun Coefficients: Colombia, Panama, Uruguay, and the United States (HP50) ............................................................ 19 m. Ten-Year Rolling Wage Okun Coefficients: Guatemala and the United States (HP50) ............................................................ 19 n. Ten-Year Rolling Employment Okun Coefficients: Guatemala and the United States (HP50) ............................................................ 20 o. Ten-Year Rolling Unemployment Okun Coefficients: Guatemala and the United States (HP50) ............................................................ 20 Figure 3.2 Wage and Unemployment Okun Coefficients for the United States and Latin America, 1990-1996 ............................................................ 21 Figure 4.1 Output Elasticity of Employment and Labor Shares of Output .............................. 29 a. Long-Term Output Elasticity of Employment for Argentina, Bolivia, Chile, Costa Rica, Mexico, Peru, and the United States (HP5) ..................................... 29 b. Long-Term Labor Share of Output for Argentina, Bolivia, Chile, Costa Rica, Mexico, Peru, and the United States (HP5) ......................................................... 29 c. Long-Term Output Elasticity of Employment for Brazil, Venezuela, and the United States (HP5) ............................................................ 30 d. Long-Term Labor Share of Output for Brazil, Venezuela, and the United States (HP5) ............................................................ 30 e. Long-Term Output Elasticity of Employment for Colombia, Guatemala, Panama, and the United States (HP5) ............................................................ 31 f. Long-Term Labor Share of Output for Colombia, Guatemala, Panama, Uruguay, and the United States (HP5) ............................................................ 31 Appendix 1 United States ............................................................ 36 Unemployment Rate Index of Real Wages in Manufacturing Employment and Real Gross Domestic Product Argentina ............................................................ 37 Unemployment Rate Index of Real Wages in Manufacturing Employment and Real Gross Domestic Product Vi Bolivia ...................................... 37 Unemployment Rate Index of Real Wages in Manufacturing Employment and Real Gross Domestic Product Brazil ..................................... 38 Unemployment Rate Index of Real Wages in Manufacturing Employment and Real Gross Domestic Product Chile ..................................... 38 Unemployment Rate Index of Real Wages in Manufacturing Employment and Real Gross Domestic Product Colombia ..................................... 39 Unemployment Rate Index of Real Wages in Manufacturing Employment and Real Gross Domestic Product Costa Rica ....................................................... ...... 39 Unemployment Rate Index of Real Wages in Manufacturing Employment and Real Gross Domestic Product Guatemala ..................................... 40 Unemployment Rate Index of Real Wages in Manufacturing Employment and Real Gross Domestic Product Mexico ..................................... 41 Unemployment Rate Index of Real Wages in Manufacturing Employment and Real Gross Domestic Product Panama ..................................... 41 Unemployment Rate Index of Real Wages Employment and Real Gross Domestic Product Paraguay ..................................... 42 Unemployment Rate Index of Real Wages Employment and Real Gross Domestic Product vii Peru ............................................................ 42 Unemployment Rate Index of Real Wages in Manufacturing Employment and Real Gross Domestic Product Uruguay ............................................................ 43 Unemployment Rate Index of Real Wages Employment and Real Gross Domestic Product Venezuela ............................................................ 43 Unemployment Rate Index of Real Wages Employment and Real Gross Domestic Product Appendix 2 Ten-Year Rolling Wage, Employment, and Unemployment Okun Coefficients a. Ten-Year Rolling Wage Okun Coefficients: Argentina, Bolivia, Chile, Costa Rica, Peru, and the United States (Differences) ........................................ 49 b. Ten-Year Rolling Employment Okun Coefficients: Argentina, Bolivia, Chile, Costa Rica, Peru, and the United States (Differences) ............................ 49 c. Ten-Year Rolling Unemployment Okun Coefficients: Argentina, Bolivia, Chile, Costa Rica, Peru, and the United States (Differences) ........................................ 50 d. Ten-Year Rolling Wage Okun Coefficients: Brazil, Venezuela, and the United States (Differences) ........................................................... 50 e. Ten-Year Rolling Employment Okun Coefficients: Brazil, Venezuela, and the United States (Differences) ........................................................... 51 f Ten-Year Rolling Unemployment Okun Coefficients: Brazil, Venezuela, and the United States (Differences) ............................................................ 51 g. Ten-Year Rolling Wage Okun Coefficients: Mexico and the United States (Differences) ........................................................... 52 h. Ten-Year Rolling Employment Okun Coefficients: Mexico and the United States (Differences) ........................................................... 52 i. Ten-Year Rolling Unemilployment Okun Coefficients: Mexico and the United States (Differences) ............................................................ 53 Vill j. Ten-Year Rolling Wage Okun Coefficients: Colombia, Panama, Uruguay, and the United States (Differences) ............................................................ 53 k. Ten-Year Rolling Employment Okun Coefficients: Colombia, Panama, Uruguay, and the United States (Differences) ..................................................... 54 1. Ten-Year Rolling Unemployment Okun Coefficients: Colombia, Panama, Uruguay, and the United States (Differences) ..................................................... 54 m. Ten-Year Rolling Wage Okun Coefficients: Guatemala and the United States (Differences) ............................................................ 55 n. Ten-Year Rolling Employment Okun Coefficients: Guatemala and the United States (Differences) ............................................................ 55 o. Ten-Year Rolling Unemployment Okun Coefficients: Guatemala and the United States (Differences) ............................................................ 56 ix ACKNOWLEDGMENTS 7" y * J I WOULD LIKE TO THANK Guillermo Perry for his encouragement and support of the pro- ject. I would also like to thank Eliana Cardoso, Norman Loayza, Bill Maloney, William Pizer, Mario Cuevas, Humberto Lopez, Daniel Lederman, and participants at the Latin American and Caribbean Cross-Fertilization Seminar at the World Bank and the Econometric Society Meetings in Peru for many valuable comments. Finally, I am grateful to Conrado Garcia for research assis- tance well beyond the call of duty. The opinions expressed in this paper belong to the author, and do not necessarily represent the opinions of the World Bank, its Board of Directors, or the coun- tries they represent. x ACRONYMS AND ~~~~~~~~~~~~~~~~~~~~~~~~~~~~MM.gSe ABBREVIATIONS CEPAL Centro Estudios para America Latina INEGI Instituto Nacional de Estadistica y Geografia CPI Consumer price index ITAM Instituto Tecnol6gico Aut6nomo de DANE Departamento Administrativo Nacional Mexico de EstadIstica NAIRU Nonacceleration inflationary rate of ECLAC Economic Commission for unemployment Latin America and the Caribbean OCEI Oficina Central de Estadistica e ENEU Encuesta Nacional de Empleo Urbano Informatica GDP Gross domestic product OECD Organisation for Economic Co-operation and Development HP Hodrick-Prescott OIT Organizaci6n Internacional de Trabajo IBGE Instituto Brasileiro de Geografia e Estatistica OLS Ordinary least squares IFS International Financial Statistics UN United Nations IICE Instituto de Investigaciones de Ciencias WDI World Development Indicators Econ6micas ILO International Labour Office 1 ^ UMMARY ,0>?~~~~~~~~~~~~~~~~~''- -7_ ' 1 111 THIS PAPER STUDIES LABOR MARKET FLEXIBILITY in 13 Latin American countries since the 1960s and 1970s by looking at the sensitivity of employment and unemployment, and real wages with respect to output. It finds that price stabilization has brought real-wage stability, but that it has tended to increase uncertainty of job security. It argues that declining inflation makes labor market rigidities binding because labor markets cannot absorb output shocks through prices. Cyclical relationships are studied by constructing Okun coefficients for unemployment, employ- ment, and wages using first differences and the cyclical component of a Hodrick-Prescott (HP) decomposition of the series. This paper finds that compared with the indicators are labor market reforms. Long-term United States, output fluctuations in Latin Amer- relationships are studied using a standard produc- ica have a small effect on the quantity variables tion function and the permanent component of employment and unemployment, but a large the HP decomposition of the series. In all seven effect on real wages. For five of the six countries countries that implemented a price stabilization that implemented a price stabilization program, program, the output elasticity of employment the wage sensitivity of output decreased. Con- increased, implying higher productivity and lower versely, with respect to output, unemployment- employment generation. Long-term labor shares employment sensitivity tended to increase. Coun- were also computed, and it was found that in five tries in the sample with stable inflation levels did of these countries labor shares increased, imply- not exhibit such an inverse relationship. The ing that labor is reaping the benefits of increased most important determinants of the flexibility productivity. 2 INTRODUCTION ECONOMIC GROWTH IS GOOD FOR WORKERS. World Bank (1995b) ONCE AGAIN, THE QUICK CAPACITY TO OVERCOME ECONOMIC DIFFICULTIES [IN 1995] WAS INSUFFICIENT TO MARK IMPROVEMENTS ON THE LABOR FIELD. ILO Latin America (1996) The two quotes above are at the heart of the debate policymakers' renewed attention to labor market on the impact, or lack thereof, of "Washington issues. Thus, in high-inflation countries, a reduc- Consensus Reforms"1 on labor markets in Latin tion in real wages, obtained through less-than-ftill America (see Williamson 1995). This paper studies nominal wage adjustment, counterbalances the labor market flexibility in 13 Latin American coun- costs imposed by labor market rigidities. This paper tries since the 1960s and compares them with the argues that in the long term, in stabilization coun- United States by looking at the sensitivity of tries, the output elasticity of employment increased, employment and unemployment, and real wages reflecting improvements in productivity, providing with respect to output in the short and long term. support for the hypothesis that postreform growth The two main findings in the short term are (a) is not as labor-intensive. This paper also finds, Latin American labor markets adjust to output however, that labor shares of income have shocks predominantly through wages rather than increased, indicating that workers are reaping pro- through employment or unemployment when ductivity increases in the form of wages. cormpared with the United States, and (b) price sta- To test the short-term hypotlhesis mentioned bilization makes labor market rigidities binding above, this paper estimates the response of unem- because labor markets cannot absorb output shocks ployment-employment and real wages to changes through adjustment in real wages, which explains in output throughout the period. It argues that out- REVISITING AND EXPANDING OKUN COEFFICIENTS * 3 put shocks in each country will be reflected entirely tries where the quantity adjustment channel has not in fluctuations in real wages (prices), employment increased, as in Argentina, labor reform has become and unemployment (quantities), or turnover (not a high priority. The investigation of rigidities as a investigated in this paper)-regardless of such fac- function of policy variables is left for future tors as labor market regulations, institutions, or research. idiosyncrasies. I constructed Okun coefficients, Long-term relationships between these vari- which for now can loosely be defined as the effect ables allowed me to address whether reforms are of a 1 percent deviation of output from its trend on benefiting wage earners as much as expected. To unemployment, employment, and wages. The determine whether postreform growth is generat- paper finds that output fluctuations in Latin Amern- ing as much employment as expected, I used a sim- can economies have a smaller effect on quantity ple production function framework and con- variables (unemployment and employment) and a structed long-term output elasticities of bigger effcct on real wages when compared with employmcnt. Long-term elasticities were con- the United States. In particular, the paper finds structed using the permanent component of the large wage Okun coefficients in Latin America on Hodrick-Prescott (HP) decomposition of each of the order of 1.0, while in the United States the the series. In many countries, a U-shaped pattern comparable number is close to 0.5, and conversely, emerged, indicating that despite employment pro- unemployment and employment Okun coefficients tection, import-substituting regimes had high out- of 0.1 to 0.3 compared with 0.3 to 0.4 in the United put elasticities of employment. The United States States. The magnitude of the unemployment coef- also has a slight U-shaped pattern, indicating that ficients in Latin America is closer to those in productivity gains were smaller in the 1970s. Europe and Japan. Postreform growth has been associated with larger In the sample, there are six countries- productivity gains than in the 1980s. The result was Argentina, Bolivia, Chile, Costa Rica, Mexico, and expected and is desired. Structural reforms have Peru-that carried out a price stabilization program translated into larger gains in productivity. In the during the sample period. Brazil's stabilization plan long term, it is the only way to increase welfare. was launched in 1994 and thus is included as a The question of the well-being of workers is counterexample, along with Venezuela, as a country better addressed by looking at labor participation in with rising inflation. Okun coefficients for real output. I constructed long-term labor shares and wages fell in five of the six countries and increased found that labor shares rose after reforms in four in Venezuela and Brazil. Conversely, in four of the (Bolivia, Chile, Costa Rica, and Mexico) out of the six stabilizing countries the unemployment coeffi- six reforming countries, and the precipitous fall in cients rose, and in all six countries the employment Peru was contained. Moreover, Venezuela follows coefficients rose. Therefore, the newly acquired the pattern of a counterstabilization country. The concern on the part of Latin American policymak- long-term effects of structural reforms and stabi- ers with labor market rigidities is justified. With lization are that output elasticity of employment lower inflation, real-wage adjustment is not as increases because of productivity gains, while wages effective a mechanism of adjustment. Real wages rise to more than compensate lower employment are easier to adjust downward in a higher inflation- per unit of output. ary regime by awarding nominal increases below The study of the transmission mechanism the rate of inflation. With lower inflation, adjust- between output and labor market performance vari- ment to shocks-which were previously easily ables offers insights on the effects of labor market accommodated through real-wage adjustments- reforms in countries that have not experienced now have to be translated into layoffs and closings. extreme price instability. This paper finds that the Thus, employment rigidities, which have been in gradual liberalization of labor regulation in Colom- place for a long time, have become more binding bia has resulted in a steady increase in the unem- today because the wage-adjustment channel is no ployment-employment sensitivity with respect to longer available. It is not surprising that in coun- output, and has madc Colombia one of the few 4 * LABOR MARKET FLEXIBILITY countries with countercyclical wages. Moreover, at answering whether the labor market in country along with Chile, Colombia is the only country X is more or less flexible than the labor n arket in with a sustained increased labor share of output. country Y The study is meant to lay the ground- Panama presented an increasingly more work for further refinements by country. Finally, inflexible labor market until the trend was at least this paper provides a measure of "broadness of partially reversed in the last few years. The 1972 growth." How much of growth is translated into labor code had devastating consequences for quan- employment? Is growth more labor-biased in one tity and price adjustment channels and became a country than in another? major obstacle to growth. However, legislation From an academic point of view, the paper approved in 1995 has reversed the trend. Okun investigates whether basic macroeconomic relation- coefficients for employment and unemployment ships take different shapes across countries depend- rose significantly because modifications of the law ing on history, customs, regulations, and institu- eliminated automatic conversion of temporal to tions, despite the fact that the basic economic logic permanent contracts after three months. behind these factors should be quite universal. This In Guatemala, the benefits of peace since 1986 paper stretches Okun's contribution to bridge the dwarf the benefits of stabilization for any other often-discussed gap between macro and labor eco- country. There were robust and immediate nomics by claiming that his logic for the relation- improvements in short-term labor market flexibil- ship between output and unemployment applies to ity, and long-term worker well-being. The story in real wages. Moreover, I argue that it is the most Uruguay is less clear; in the short term there appropriate measure of labor market flexibility in seemed to be labor market liberalization both in the short term for Latin American countries, where quantities and prices until 1993, when a sharp output growth is volatile. reduction in both was observed. There has been a This paper is structured as follows: Section II deterioration of the labor share since the early describes the interpretation of short- and long-term 1970s. fluctuations in our relevant variables and the justifi- This paper makes three important contribu- cation for using real wages, unemployment, and tions to the debate over labor market reform in employment as performance variables. Section III Latin America. First, this is the first time that Okun describes the transmission mechanisms aniong the coefficients or any other uniform methodology has variables, and presents estimation procedures and been applied over such a long time period and wide results. Section IV presents the long-term structural range of countries. In fact, there is surprisingly little relationship between employment and output. comparative econometric work on Latin American Using a conventional production function frame- labor markets. In addition, this paper compares work and the permanent components of the series, results with the United States. Second, this paper long-term output elasticities of employment and does not capture the richness provided by survey labor shares are estimated. Section V offers a con- data but "looks at the forest" and takes a first crack clusion and presents future avenues of researclh. 5 REAL WAGES, EMPLOYMENT, AND UNEMPLOYMENT AS MEASURES OF t LABOR MARKET PERFORMANCE IN STUDYING LABOR MARKET FLEXIBILITY, as in other areas of economics, there are two types of variables: policy and performance variables. There is an extensive literature concentrating on the links between labor market policy variables and the resulting performance variables using microeconomic survey data. USING PERFORMANCE the United States for the last 20 to 30 years as a way VAR IA B L E S of measuring the ability of the labor market to absorb output shocks. This is not only a simple and elegant way, but also the appropriate way, to mea- Since this paper concentrates on the link between sure labor market flexibility in an environment output and the labor market, it was natural to con- with volatile output. centrate on performance variables. All the idiosyn- crasies of such factors as labor market regulations, institutions, and taxes will be reflected in a perfor- DECOMPOSING THE SERIES mance variable. Surprisingly, there is consensus that INTO LONG- AND SHORT-TERM there are only three broad labor market perfor- C O M P O NE N T S mance variables:2 (a) a price variable best measured by real wages, (b) a quantity variable measured The related literature on output, employment, alternatively by either employment or unemploy- unemployment, and wages has divided the study of ment, and (c) turnover (labor markets are peculiar the relationships of these variables into long- and in that turnover is another measure of perfor- short-term categories. This work will follow that mance). Because comparable turnover data are convention because of the nature and length of the scarce, measuring a labor force's ability to adjust to time series, and for clarity. Each log series is made up sector shocks is left for future research. of two components: a long-term trend (or permanent This paper looks into the structural relation- component), and a short-term or business cycle com- ship between output and performance variables (a) ponent. The econometric decomposition is described and (b) across 13 countries in Latin America and in section III, as shown below by equation 2.1: 6 * LABOR MARKET FLEXIBILITY y = yp + Y, The interpretation of the decomposition of employment and unemployment is more involved. e = ep + ec Long-term employment is determined exoge- (2.1) nously by the growth of the economically active u up + uc population. This may differ from population growth because of differing age distribution para- wT wp + Iv meters. Cyclical employment varies due to the changing participation rate. Traditionally, unem- TIhe interpretation of the coefficients is ployment was thought of as a stationary variable straightforward: long-term growth, yp, can be that returned to its long-term value, the rionaccel- thought of as the term explained by Solow's eration inflationary rate of unemployment growth literature (Solow 1957). Cyclical output (NAIRU). However, recent studies have ,uggested variations have been traditionally explained by that the NAIRU may change over time. The mat- short-term macroeconomic models. Long-term ter is far from settled and is beyond the scope of real wages are determined by productivity, while this paper. The nature of long-term unemploy- short-term deviations depend on labor demand. ment is widely covered in the literature.3 7 CYCLICAL VARIATIONS AMONG X EMPLOYMENT AND UNEMPLOYMENT, WAGES, AND OUTPUT THE LITERATURE GOING BACK TO OKUN (1962) has concentrated on the cyclical rela- tionship between output and unemployment. The innovative approach of computing Okun coef- ficients for all three variables can provide insight into labor market flexibility across countries and time periods. S H O R T - T E R M T RAN S M I S S I O N lated the following transmission mechanisms, MECHANISMS BETWEEN which were all reflected in the unemployment OUTPUT AND THE LABOR rate:5 MARKETS A procyclical behavior of the labor force-that is, increased participation in the labor force The reasoning derived from the unemployment during an upturn, and decreased participation link can be readily applied to employment and during a downturn. There are two opposing wages. Okun postulated that each percentage point forces on the size of the labor force during decrease in the unemployment rate is associated business cycles: with an increase of 3 percent in output. This 3-to-1 (a) Substitution effects cause a procyclical labor ratio has come to be known as Okun's Law, or force. During the upswing, wages rise, more appropriately, Okun's coefficient.4 The moti- increasing the opportunity cost of leisure vation of Okun's paper was to establish the tremen- and drawing more participants, particularly dous gains in output that would be obtained from teenagers and females, into the labor force. reduced unemployment. The result was a big sur- Conversely, in a downturn one observes a prise. The finding that a 1 percent reduction of "discouraged worker" effect as people stop unemployment translated into a 3 percent increase looking for work.6 in output violated the basic postulates of diminish- (b) Income effects work toward a countercycli- ing marginal returns to labor, and constant returns cal labor force; that is, when the head of the to scale. Surely if employment increased by 1 per- family becomes unemployed, the rest of the cent the effect in output would be smaller than, but household has to enter the labor force.7 This close to, 1 percent. Recognizing this, Okun postu- paper finds that contrary to Organisation for 8 * LABOR MARKET FLEXIBILITY Economic Co-operation and Development hours, labor force participation, and man-hour (OECD) countries, several countries in productivity, the magnitudes of all these effects Latin America exhibit countercyclical labor are related to the unemployment rate. With this forces during the business cycles, indicating assumption, the unemployment rate can be that income effects dominate, pointing to an viewed as a proxy variable for all the ways in important qualitative difference. which output is affected by idle resources. (Okun 1962, pp. 98-104) • Labor hoarding-that is, the average number of hours worked by each worker-moves pro- The above quote has had a tremendous effect cyclically, amplifying the fluctuations in unem- on the literature. Much effort in the literature for ployment. As output expands, firms prefer to the following 30 years has been devoted to making increase the number of hours per worker and this "huge leap from unemployment to output" a the number of shifts, rather than hiring work- "step by step" process. Throughout the period, ers, and vice versa.8 there has been some debate on the size of the coef- ficient. In general, estimates have tended to be * Labor productivity moves procyclically for rea- higher than the 0.3 originally proposed by Okun, sons not entirely understood. Various mecha- but they have been at most 0.45. With increasing nisms have been documented: (a) Labor is fixed quality of labor market statistics, including number to some degree because of overhead costs, con- of people employed and hours worked, there has tractual commitments, technological constraints been a strong consensus that in the United States, that prevent frictionless hiring and firing, large for each percentage point reduction in unemploy- transaction costs of layoffs, learning curves, and ment, the first two effects translate into increases in morale effects;9 (b) shifts occur in industrial output of around 1.8 percent, implying that a little composition toward more productive activities less than half is taken up by fluctuations in produc- during periods of growth;10 (c) ladder climbing tivity.14 Efforts made to incorporate data on produc- or the upgrading of labor takes place;"' (d) wage tivity explicitly are not conclusive owing to difficul- differentials narrow in the upswing;12 and (e) ties in the construction of potential output series."5 employment shows a lagged response to varia- Efforts to estimate Okun coefficients outside tions in output."3 the United States have been limited to OECD countries. Results suggest that in the United States, Okun's coefficient assumes that all three output variations translate into larger fluctuations effects change parn passu with the unemployment in unemployment when compared with Europe rate, and the combined effects are captured in the and Japan-in other words, the United States labor coefficient. Furthermore, Okun argued that market is more flexible in quantities. The explana- because productivity and hours-per-worker data tion for this result is not clear. A first line of argu- were notoriously unreliable, it was preferable to use ment centers on differences in the definition of a reliable statistic that embodied all of the changes unemployment across countries. However, the above: U.S. Bureatu of Labor Statistics has constructed a compatible unemployment series for manv OECD Ideally the measurement of potential output countries directly from survey data, but the results would appraise the various possible influences remain robust (that is, unchanged). Others argue of high employment on labor input and pro- that employment stability plays a larger role in ductivity and evaluate the influences step by Europe and Japan.16 Section III will show that as in step.. . The basic technique I am reporting con- Europe and Japan, output in Latin America trans- sists of a leap from unemployment rate to lates into small fluctuations in unemployment. potential output ... Strictly speaking, the leap The intuition for the structural relationship requires the assumption that, whatever the between output and employment requires a brief influence of slack economic activity on average recapitulation of the definitions of employment and REVISITING AND EXPANDING OKUN COEFFICIENTS * 9 unemployment. In a nutshell, variations in unem- shocks as exogenous and measures the degree of ployment are usually accompanied by fluctuations adjustment in the labor market. in the size of the labor force. Thus, a rise of 1 per- For the United States output and unemploy- cent in the employment rate does not always trans- ment series, there is a long, involved, unsettled late into a fall of 1 percent in the unemployment econometric debate on the appropriate econometric rate. Employment growth can be quite vigorous, decomposition of the original series.18 Elements of but if a segment of the population decides to the debate are briefly mentioned here to justify the engage in job searching-women or teenagers, for techniques chosen. The disagreement centers on example-unemployment may rise. In relation to whether the series contains a unit root or is trend- the unemployment transmission mechanism, only stationary.19 If the series contains a unit root, a posi- the first channel does not apply because we are tive innovation leads to a revision of the forecast for directly controlling for the number of workers. all horizons, whereas if it is trend-stationary, the Procyclical hour variations and productivity still future path remains the same-that is, in the pres- apply. Thus, we expect Okun employment coeffi- ence of a unit root in the series, a positive error in cients to be higher than traditional unemployment period t causes the entire series in levels to be Okun coefficients if the labor force fluctuates pro- shifted upward without the tendency to return to cyclically, as in OECD countries, or lower if it fluc- its trend.20 Traditionally, it was assumed that the tuates countercyclically. logarithm of output deviated around a deterministic Macroeconomists since Adam Smith have trend. Nelson and Plosser (1982) argued that out- been preoccupied with the cyclical relationship put series contained a unit root, and therefore between real wages and output. Keynes wrote, in econometric models that assumed deviations The General Theory, that "an increase in employ- around a deterministic trend were mis-specified ment can only occur to the accompaniment of a (that is, they were yielding biased estimators) .21 A decline in the rate of real wages." However, the similar debate exists in the decomposition of evidence suggests that, if anything, real wages are unemployment. Solutions include using vector procyclical. The debate is far from settled, but the autoregressors, cointegration techniques, and general consensus is that a real wage index for a Kalman filters.22 This paper provides two alterna- country is procyclical in industrial economies.17 tives: (a) Assume output is trend-stationary and This paper finds a strong procyclical relationship "detrend the series" in order to avoid spurious cor- in most of the Latin American countries in our relation between the variables (the choice of sample. detrending technique is discussed below); and (b) accept the unit root hypothesis, reject a cointegra- CYCLICAL ESTIMATION tion relationship between the variables, and run the PROCEDURES regressions in first differences. The issue of the choice of dependent variable Data between output and employment-unemployment was first raised in a comment by Solow in 1973 (see The figures in appendix 1 present plots of unem- Okun 1973) and has not been resolved. Conceptu- ployment, inflation, real wages, output, and ally and econometrically, the matter is difficult to employment for each of the countries. In addition, settle because variables Granger cause each other the detailed sources of each of the series are pre- (see Evans 1989). In a bivariate regression such as sented. A constant price output scrics and con- the one presented here, estimates differ only sumer price indexes (CPIs) were obtained from because the sum of horizontal rather than vertical various issues of the publication International Finan- squared residuals is minimized. Okun (1973) and cial Statistics (IFS). There were numerous sources Perry (1971) both point out that one gets slightly for information on employment, unemployment, different results by interchanging dependent and and real wages. Real wages in industry were either independent variables. This paper takes output taken directly or constructed from nominal wages 10 * LABOR MARKET FLEXIBILITY using the IFS CPI series. The series captured the First Differences broadest definition of industry obtained from sur- veys of formal establishments. Thus, they represent Okun's (1962) initial method relied on first differ- formal industrial wages. ences. Although the approach has given way to With the exception of Guatemala, all of the more sophisticated methods, this paper wrill show employment and unemployment numbers in Latin that improvements make a difference only when a America are products of urban employment sur- series is "unstable." The estimation equations are veys. 23 This has two implications. First, rural statis- given by (3.1) below: tics are not included, which does not allow the identification of rural-to-urban migration.24 (More- AUt aud + /3udAy, + gud over, in countries with large rural sectors, urban employment may have a small impact on the econ- AE a ed + Pied AY + £ed (3.1) omy as a whole). Second, informal employment is included. Surveys typically ask whether a person AW, = aed + Pivdi + wd has engaged in any productive activity, formal or informal, in the last few weeks. This fact prevents where AUy., AE, AW.,, and AY. are, respec- inferences of switches between forrnal and informial tively, changes in percentage terrns of unemploy- employment, leading to the issue of comparability ment, employment, real wages, and real output in of the definition of unemployment. A casual country i at time t. Most papers for the United inspection reveals that statistical institutes in Latin States use quarterly data and introduce a lag struc- America have similar surveys and follow the Inter- ture. However, lags longer than four quarters are national Labour Office (ILO) guidelines. The bet- not significant. Not surprisingly, with annual data a ter alternative is the Herculean task of constructing single lag of output was discarded in most cases. an cmployment-unemployment series directly from For uniformity, the estimation did not incorporate a survey results. lag structure. The figures in appendix 1 show the great het- The advantages of using first differences to erogeneity of the statistical properties of unem- estimate cyclical variations are simplicity--errors ployment, real wages, inflation, and output and suffer from little autocorrelation, and estimators employment across countries in the sample. The remain consistent when the series contains a unit log axes in the output and employment panels are root with no cointegrating relationships. The such that the percentage increase is the same in Dickey-Fuller tests suggest the presence of unit both series. This allows inferences of productivity roots for some of the output series in some increases.25 The U.S. output series is well behaved period. However, the tests rejected cointegrating when compared with output series in the rest of relationships among any of the labor market vari- Latin America. Allowing for a break in 1973, it can ables with respect to output, eliminating the need be argued that output deviates around a trend (see for an error correction term or a cointegration endnote 21). It is obvious that a single determinis- technique.27 The reason for the lack of a cointe- tic trend is not the correct assumption for any of grating relationship is that each of the variables the Latin countries, and unit roots tests are period- possesses a different trend and the employment sensitive.26 In most cases, there are at least three series has rejected the unit root hypothesis. As dis- distinct growth regimes: (a) a relatively high and cussed in the section on long-term components, stable growth period in the 1960s and early 1970s, employment is expected to deviate around an (b) a period of stagnation and instability since the exogenous trend equal to the growth of the eco- debt crisis in the early 1980s, and (c) a period of nomically active population. The trend in output renewed growth in the late 1980s until today. Only is determined by the sum of employment growth, the United States and Panama show stable and capital accumulation, and productivity gains. positive productivity increases throughout the Thus, using a cointegrating technique required period. the assumption that variables followed a random REVISITING AND EXPANDING OKUN COEFFICIENTS * 11 walk, but that the difference between their drifts changes in the natural rate of unemployment. was deterministic. Okun, as well as other authors, postulated an arbi- The disadvantages of using first differences trary natural rate of unemployment-namely, 4 are that (a) the technique eliminates "memory" in percent. However, choosing an arbitrary natural levels by construction in order to properly estimate unemployment rate across countries seems ill- series with unit roots, (b) insight into the long- advised and calculating a "natural" rate of unem- term trends of each of the series is lost since one ployment using a concept such as a NAIRU is well observes only the ratio of the trends in the series, beyond the scope of this exercise. and (c) the procedure implicitly assumes that the Cyclical components are obtained by subtract- series is first difference-stationary in the relevant ing the fitted trend values from the original series, period. Simple inspection of the output series in as demonstrated by equation group 3.4: appendix 1 rejects this assumption. Allowing for different intercepts in time periods corrects for this Yit = Y- t restriction, but the choice is arbitrary, and switches in long-term growth rates are discrete. u= u t- ua (3.4) Deterministic Detrending e= e-t _ t The basic procedure of using output trends and wit = - wa cyclical gaps was Okun's (1962) second method, but this paper follows Weber's (1995) exposition for where y,t, ut, e', and w' are the fitted values clarity. The authors regress the natural logarithms from equations 3.2 and 3.3, respectively. of each series on a deterministic trend and use the The static ordinary least squares (OLS) Okun residuals as cyclical compoinenits. The relationship coefficients are given by regressing the cyclical is given by equation group 3.2 below: components, as shown by (3.5): Yi= a. + Di + Oyt + 0.D.t + t U' = putyc eit= a. + Di + O't + 0,D t + v. (3.2) ef = petcy (3.5) i t I z 2i i itI i wit =a(X + Di + oxi't + 02iD.t + vi W, = owl c where a lower case denotes the natural loga- where fu'i, ft3, 13w7 are our second set of Okun y e i rithm of the series, t is a time trend, and Oi, 0 tl coefficients. and 0w are long-term growth rates for output, A technical disadvantage of this approach is employment, and real wages, respectively. The that if the series contains unit roots, deterministic inclusion of dummies for changes in slopes and detrending leads to inconsistent estimators. In prac- intercepts is crucial for some of the countries in tical terms, deterministic detrending poses two Latin America to avoid misinterpretation of the major setbacks-estimates are extremely period- residuals.28 sensitive, and it seems unreasonable to assume that "Trend" unemployment is constructed by cal- long-term growth rates change abruptly and dis- culating the unconditional mean, as shown by cretely.29 Given growth stability in the United equation 3.3: States and the other OECD countries, determinis- tic detrending was the most popular technique used ui = HXi + 0 1D. + (3.3) to estimate Okun's coefficient. For unemployment, it is worth noting that even in the case of postwar where 0u. is the average rate of unemployment United States, authors have found it necessary to and Di is a dummy variable, which allows for allow for at least a one-time break in the natural 12 * LABOR MARKET FLEXIBILITY rate (Evans 1989; Perron 1988; Weber 1995). For adjust to output shocks dramatically more through Latin American countries each of the growth adjustment in real wages than U.S. labor markets, regimes was allowed to have a different natural as reflected in larger Okun ratios;31 (b) conversely, unemployment rate. The procedure is admittedly Latin American economies tend to have lower- arbitrary, and results are available on request. quantity responses to output both in terms of Since the series converges to a deterministic unemployment and employment than the United trend, the estimation process has the advantage that it States; and (c) as inflation falls, output shocks are has "memory" in levels; that is, a large deviation from absorbed less through real wages, and there is a ten- its trend today results in a return to trend in the dency for larger unemployment-employment future. Therefore, the interpretation is more in accord responses. with business cycle theories. Moreover, insight into Given the period-sensitivity of the computation the long-term trend of the series is gained. of Okun coefficients, 10-year rolling regressions were constructed. Figures 3.1.a through 3.1.o present Using the Hodrick-Prescott Filter results for unemployment, employment, and wages, respectively, using the HP filter.32 Each coefficient in Hodrick and Prescott (1997) proposed an econo- the figures represents the effect of a 1 percentage metric procedure for "representing a time series as point output deviation from trend on employment, the sum of a smoothly varying trend component real wages, or the unemployment rate in the previ- and a cyclical component." The trend component is ous 10 years. That is, the Okun coefficient for 1971 the sum of the squares of its second difference and is the result of an OLS regression using the 1961-71 the cyclical components are deviations from the sample, the coefficient for 1972 represents data for trend component. The regression ran uses the cycli- 1962-72, and so forth. Therefore, changes in the cal component from the HP decomposition in coefficients from one year to the next can be inter- equations 3.5, where Okun coefficients are given by preted as innovations of the year incorporated.33 For betas. There are certain advantages to using the HP example, the 1993 Chilean unemployment Okun filter, three of which are as follows. First, it is not coefficient dropped with respect to 1992 because the period-sensitive and thus there is little arbitrariness. Chilean economy grew close to 10 percent that year Second, long-term trends vary smoothly over time, and the country had reached full employment. If the which is more intuitively appealing. And third, like 1993 observation were omitted, the coefficients deterministic detrending, regressions have level would continue to be around the -0.5 previously "memory," and deviations from trend can readily be reached, instead of the -0.3 reported. Coefficients interpreted as business cycles. The procedure is capture the cyclical response for the previous 10 widely used, but not without criticism. Harvey years; that is, they have a 10-year memory. Though (1992) calls it "making the same mistake [as linear each regression contained only 11 observations, the detrending] but in a more sophisticated way." He coefficients are precise, the explanatory power of the proposes Kalman filters as an alternative. The regressions is high, and each Okun coefficient com- biggest disadvantage of using the HP filter, as puted covers a 10-year period, which is a long time opposed to the Kalman filter, for example, is that it for these studies. Standard errors for the United does not "solve" the nonstationary nature of the data States are one-tenth the size of the employmnent and and may lead to inconsistent estimators.30 unemployment coefficients. Table A.2.1 shows regression results for the last window in each of the CYCLICAL RESULTS: regressions. IN TERP RE TAT ION OF Labor markets in Latin America adjust to out- COEFFICIENTS AND GENERAL put shocks predominantly through adjustmnents in DISCUSSION real wages, as evidenced by large Okun ratios (see figures 3.1.a, d, g, j, m). With the exceptions of Results for the sample independent of the estima- Brazil, Colombia, and Panama, the price-adjusting tion process are (a) Latin American labor markets ability of Latin American labor markets is greater REVISITING AND EXPANDING OKUN COEFFICIENTS * 13 and more volatile than that of the U.S. labor mar- both estimation techniques.34 In Latin America, ket. Okun ratios greater than 1 indicate that output with the possible exception of Chile and Bolivia, fluctuations magnify wage responses in Argentina, employment and unemployment Okun coefficients Bolivia, Chile, Costa Rica, Mexico, Peru, Uruguay, are lower and more volatile. Table 3.1 shows that and Venezuela. The transmission mechanism is the quantity adjustment ability of labor markets in that inflation allows for rapid real-wage adjust- Latin America is closer to estimates obtained for ments throtigh less-than-full nominal wage index- Europe and Japan. It is ironic that "Okun's 1/3 ation. The observation corroborates the fact that Law" is only a law in the United States, and that in output shocks in Latin America place a large bur- the rest of the world the unemployment and den on salaried employees rather than creating employment response to output shocks is lower unemployment. and more volatile. The fact that in the United Unemployment and employment Okun coef- States unemployment and employment are more ficients in figures 3.1.b, c, e, f, h, i, k, 1, n, and o responsive to output does not necessarily mean that indicate large and constant quantity responses of the U.S. labor market is more flexible or more effi- the U.S. labor market to cyclical output shocks cient than others. If real wages were perfectly flexi- when compared with any country in Latin America. ble, then unemployment would never fluctuate. A 1 percent output shock in the United States Hamada and Kurosaka (1984) have argued that cul- translates into a reduction in unemployment of tural reasons in Japan allow for greater wage flexi- between 0.35 and 0.45 throughout the period with bility in exchange for employment stability. Figure 3.1 Ten-Year Rolling Wage, Employment, and Unemployment Okun Ratios: The Exception a. Ten-Year Rolling Wage Okun Coefficients: Argentina, Bolivia, Chile, Costa Rica, Peru, and the United States (HP50) 3,5- 3 _ 2.5 --USWCI55 ARWCC55 BOWC155 2 - CHWCC55 'k-PEWCI55 CRWCI55 1.5 0.5 0 1971 1974 1977 1980 1983 1986 1989 1992 1995 14 * LABOR MARKET FLEXIBILITY b. Ten-Year Rolling Employment Okun Coefficients: Argentina, Bolivia, Chile, Costa Rica, Peru, and the United States (HP50) 0.9 - USEI111 0.8 ARED6 l 0.7 CHES11 0.2 0.1 0 1971 1974 1977 1980 1983 1986 1989 1992 1995 c. Ten-Year Rolling Unemployment Okun Coefficients: Argentina, Bolivia, Chile, Costa Rica, Peru, and the United States (HP50) 0~~B5 CHDOM55 X. -0.1 - \CR55 1 MI Ln (Output) LI*SLN31DI)1JJRO9 Nfl)O DNICINVdXa C[NV DNI I SIA'2H Percent Percent 1960 19601960 1962 -~1962 1964 1964 1966- 1966 1968 1968 1970~~~~~~~~~~~~~- ~~~~1970 1972 ~-21972 1974 t----- 1974 1976 1976 1978 1982 1980 I~ 980 _ 1982 ~~~~~~~~~~~~~~~~~1982 1982~~~~~~~~~~~~~94 1984 w18 1986 1986-' 1988 1988 1990 ~~~~~~~~~~~~~~~~~1990- 1992 1992 1994 ~~~~~~~~~~~~~~~~~1994 1996 -1996 Index (1990=100) Index (1990=100) 1950 -v1960 1962L 1962 1964 -1964 - 1966 1966 A 1968- 19682 1974 1974 ~ 976 - ------196.- 10 ~ ~ ~ ~~18 1980 O-. 90 1982 E 19821- 1984 *2 0 19847 1986 1986 1988 -199884 1990 1990 1992 ~~~~~~~~~~~~~~~1992 1994 219947 -- 1996 _____1996- lInfation Inflation Ln (Employment~ Ln (Employment) 1960 \21970 F7777-777 1962- 217 1964 1966 217 CL 1968 r -~~~~~~~~~~~~~~~~~17 - ~~~~~ ~ 1976 0' 1970 m 98 1972 - a -~~~~~~~~~~~~~~~~17 3' 19742< 18 1976 ~ ~ ~ ~ ~ ~ ~ ~ ~~~~~~~~ ~~ 1982. 1974 94 z 1986 1(82 (0r 19"84 2 9880 1992~~~~~~~~ 1 9901 .19 1992 --1994 I1994.- " 1996.' 0 1996 -0 ' t~ Ln (Output) A II-I IOI X21a IMI q VIN d OV1 -I REVISITING AND EXPANDING OKUN COEFFICIENTS 39 661 9661 661 E661 1661 1661 U 6861 6861 L861 r L861 "61 "61 E861 la 1861 1861 El 4 6L61 6L6 T LL61 LL61 9L61 9L61 4- EL61 00 r- 'D -Z `6 145 1413 (juam(ojdwg) U'l uoiluuul UO"RUUI 9661 9661 t,661 i,661 Z661 Z661 0661 0661 8861 8861 9861 9861 t,861 t,86T Z861 P" Z861 0861 086T SL61 8L61 9L61 9L6T tL61 tIL61 ZL61 ZL61 OL6 OL61 4. If l 9961 8961 9961 9961 i7961 i,961 Z961 Z961 096 T 0961 (001=0661)x4PUl (001=0661)XaPul 966T T 9661 t,661 t,661 Z66T Z661 066T 0661 8861 S86T 9861 9861 t,86T t,861 Z861 Z861 0961 t 0861 8L6f 8L6I 9L61 T 9L6I tL61 tIL61 ZL61 ZL6T OL61 + OL61 8961 896T 9961 9961 V961 i,96T Z961 Z961 0961 0961 luaajad jua3jad 1960 -v - _________> ;4_______________ 1962 1964 1966 : 1968 1970 1972 1974j 1976 I 1978 T 1980 1982 1984k 19861 1988 1990i 1992 , 19941 ' 19961 Index (1990=10O) 1960 - 1962w 1964 1966 19684 1970k 1972--a 1974 - 1976 7 c 1978~ 19802 1982 1984~~ 1986 _ ' k Sfi ;Ak(:L2 ;Ni1¢0 SC 4:7 (-d?(t; Rf: e 1990 - --' _: 1992 -t 1994 u ; 1996 ---i--~-________Index_(990=100 Inflation Ln (Employment) 19620 1962 t;t ;:$:;0t:; t :L:;; i X:;0 ;-N 1964 1 1966± 8 196811 1972 19744 1976 19784 1980 -t 1982 'A 1984-i ° 1986 C 19884 - 1990 -. 1992± 1994 1996 .- --.-- Ln (output) A,jjqj91 IH HV U!RV, po Seot P Percent Percent 19000 -0 ----- 001960 ~ 0 0 1960 1960 1964 1964_ 1966 1966 1968 -1968-- ., 1970 - 1970 1972 - 1972- 1974- 94 - 19761 1976- 1978 ~~~~~~~~~~~S~978- 1980- -980 1982- - 1- 1982- 1984 t -1984- 1986 ~ ~~~~~~~~~~1986- 1988 - ---- 1988 /.- 1990 ~~~~~~~~~~~~~~~1990 i 1990 99 1992 4 1994 --~-~ 1994-- - ~~~~~~~~~~~~~1996+ 1996- - Index (1990=100) Index (1990=100) 1960 1960 1962-0 196 0- 1964i 1964 1966 -- 1966k ___ 1968 -1968 1970 -1970j- 1972 - 972 - - - 1974-4 1974 . 19762 - 1976 1978 + ~~~~~~1978 1980 ˇ- I-1980 -_. 1982 18 -- --- 1984k K 1984 '- - 1986 n 1986- 1988 1988 1990 ---- ~~~~~~~~~~~~1990 1992I 1992 C 1994 - - 1994 j _ _ _ _ _ _ _ _ _ 1996 ~- -~1996 --+ Inflation ~~~~~~~~~~~~~~~~~Inflation Ln (Employment)Lu(mlyet 1963 51970 1965 - 01972 1967 S 1969 et97 n 1971 17 197311971978 19751018 19771~~~~~~~~~~~~~~~18 0 ~~~~~~~~~~~~~1982 ~ 1979 >- 1981- 19840 19831 1986 a 1985118 o 198710C S 1~~~~~990 e '~ 1989 CD- 19911 1992 1993+ 1994 4~~~~ 19951*0 _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ ~~~1996 1 - -~~~ ~Ln (Output) Ln (oupu) -. 0 - 14 *SN aI DII A A oD Nfl XO DN ICUN V(I X G NV 0 N I TIIS I A 3 Paraguay Unemployment Rate Index of Real Wages Employment and Real Gross Domestic Product> 18 63~~~~~~~~~~~~~~~~~~~~~~8 6.6 2956 ,161 (o 29.4 14) 64~~~~~~12 M4 12 - 1--2272 4~~6.O 2~~~7506. 9 ~~~~~~~~~~~~~~~~~~~~~~~~~5.8 288 40 O 6fIIIII5626 ++4IF-f 0 5.4 I 4 r , F' 284 N It ~~~O w 0 10 0 N 10 0 N a 10 Z1, -~~~0 C -~ -0 -0 - - -~ -O 00 - -r - 10 10 ~ = = w a, r SC a, Yew0 00 0 0 NO N NO NO N O NONO NO YCM ~ ~ ~ ~ ~ ~ ~ ~ --Real Wage -- Inflation -Ln EmploynientI Ln Real GDP Peru' Unemployment Rate Index of Real Wages in Manufacturing Employment and Real Gross Domestic Product .2).....-.. .. .. M149 JC 23.3 14.8 23.2 340 ~ ~ ~ ~ ~ ~ ~ -14.7 -23.1 140 CC -% C -- 123 ~~~~~~~~~~14.6 23.0 E 14.5 T 29~ 1l2 -, - -m 6. 14.4 ' 28~ Pk 0 - 3 14.3- 2 7 r 4 1'X) ~314.2 226 8 -. V ~~~ ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~40 141 225 c 143 - ~~~~ ~~~~ ~ ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~14.0 ~V<~22.4 4 - - 4++-+±+'~~±±-+~ 3.9 c+ P V~A*4+H fm--j--H- 223 2 1r. 7~ N r N N N 0000 0 0ONO O N Real Wage Inflatioii - Ln Employment --Ln Real GDP Pereent 5 38 1960 1960 1962 1962 1964 1964 1966 1966 1968 1968 1970 1970 1972 < f" 1972 + 1974 7 1974 10 1976 - 1976 1978 1978 1980 980 1982 + 1982 1984 :L 1984 1986 1986 1988 1990 1988 1992 1990 1994 1992 1994 1996 1996 Index(1990=100) Index(1990=100' F3 Ls, 1960 1962 962 1964 1964 966 1966 1968 1968 197 1970 1972 1972 - 974 1974 1976 m :5 1976 1978- 1978 1980T TC 1982 1982 1984 w 1984 1986E m 1986 1988 1988 1990 1992 1990 1994 1992 1996 1994 1996 Inflation Inflatior Ln (E ployrnent Ln (Ernployment) m P' p- PO C, j 00 968 1967 1969 1970 1971 1972 1973 1974 1975 1976 6- 1977 1978 1979 1980 1981 1982 t + 1983 + 01 1984 0 rA rA 1985 z 1986 1987 1988 1989t 1990 1991 993 1992 1995 1994 + CL Ln (Output) Ln(Output) SINAIDILILIdOD N--IMO 9NICINVdXa CINV ONIIISIAgN 44 * LABOR MARKET FLEXIBILITY UNEMPLOYMENT RATES Argentina 1964-1979: United Nations Economic Commission for Latin America and the Caribbean (UN ECLAC). Precios, Salarios y Empleo en la Argentina: Estadfsticas Econ6micas de Corto Plazo; 1980-84: International Labour Office (ILO); and 1984-1996: International Financial Statistics. Unemployment rate of Greater Buenos Aires. Bolivia 1970-76: ILO, official estimates. 1978-96: UN ECLAC. National urban unemploy- ment rates; official estimates. After 1987 the unemployment figures are based on a survey conducted by the National Statistical Institute. Brazil 1978-96: UN ECLAC. Twelve-month average of the unemployment rate of six met- ropolitan areas: Rio de Janeiro, Sao Paulo, Belo Horizonte, Porto Alegre, Salvador, and Recife. 1997: Instituto Brasileiro de Geografia e Estatfstica (IBGE), Banco Central do Brazil World Wide Web home page. Chile Banco Central de Chile, Direcci6n de Estudios, Indicadores Econ6micos y Sociales, 1960-1985 and Boletin Mensual del Banco Central de Chile, several issues. Unemploy- ment rate in Greater Santiago obtained from a household survey. Colombia UN ECLAC. From 1990-96 the unemployment rate corresponds to seven metropoli- tan areas; from 1974-89 it corresponds to four metropolitan areas: Bogota, Barran- quilla, Medellin, and Cali. 1974-84: average for March, June, September, and Decem- ber; 1985: average for March, July, September, and December; 1986: average for April, June, September, and December; 1987: average for March, July, and September. Costa Rica UN ECLAC. National urban unemployment rate, April-October average; 1986: October average; 1976-84, average for March, July, and November. From 1987 onward the figures are not strictly comparable with the data for preceding years owing to methodological changes. Ecuador UN ECLAC. Country total, official estimates. From 1986 onward, the unemployment rate is based on a household survey of Quito, Guayaquil, and Cuenca. Guatemala Country total unemployment rate, official estimates. Honduras 1968-78: ILO; the numbers were calculated using the number of unemployed from ILO, which are official estimates; these in turn were divided by the labor force from World Development Indicators (WVI). 1980-96: UN ECLAC. Unemployment rates are estimates for five cities; 1986: urban labor survey; 1987: March, Central District. Mexico Instituto Nacional de Estadfstica y Geografia (INEGI). Encuesta Nacional de Empleo Urbano for the main cities. The unemployment rate is a four-quarter average of the metropolitan areas of Mexico City, Guadalajara, and Monterrey. Centro Estudios para America Latina (CEPAL). 1990-95: main urban areas; 1970, 1973-89: metropolitan areas of Mexico City, Guadalajara, and Monterrey, four-quarter REVISITING AND EXPANDING OKUN COEFFICIENTS * 45 average; 1987: July average. 1997 is an average of first and second quarter of 1997 as published by INEGI in Estad&sticas Econ6micas, Indicadores de Fmpleo y Desempleo. Panama Direcci6n de Estadistica y Censo de Panama, from a household survey. UN ECLAC: National nonagricultural unemployment rate up to 1977. After 1977 the unemployment rate is from the metropolitan region. Unemployment rate for 1980 is from a population census of that year. Paraguay UN ECLAC: Unemployment rate for Asunci6n, Fernando de La Mora, Lambare, and the urban areas of Luque and San Lorenzo, annual averages; 1981: first semester; 1985: average for November and December. After 1993 the data are national urban unemployment rates. Peru UN ECLAC: Unemployment rate of Metropolitan Lima; 1978: July-August average; 1979: August-September average; 1985: Official estimates. After 1995 the figures cor- respond to total urban unemployment. ILO. Uruguay UN ECLAC: Montevideo unemployment rate. Two-semester average until 1979; four-quarter average, thereafter except for 1987, in which the unemployment rate is the average of the first three quarters. Venezuela Oficina Central de Estadistica e Informatica (OCEI): Household survey. UN ECLAC: National urban unemployment rate. Two six-month averages. 1984 and 1985 are national unemployment rates; 1986, second six months; 1987: first six months. United States Economic Report of the President 1997. WAGES IN MANUFACTURING Argentina ILO. Wages in the manufacturing industry; skilled and unskilled wage earners. UN ECLAC. 1970-80: wages of laborers in the manufacturing industry in the metro- politan area. 1980-96: average total monthly wages in the manufacturing industry. Bolivia ILO. Wages of employees; 1995-96: private sector in La Paz. Brazil (Rio deJaneiro): Coase (1993). The average nominal wage for industry used to con- struct the real wage was the annual average of IBGE's monthly average salary for workers in industry. It refers to the average wages in industry in general, deflated by the CPI for Rio de Janeiro. UN ECLAC. Average wages in basic industry, deflated by CPI for Rio de Janeiro. Chile ILO. Wages of employees in manufacturing. Colombia ILO. Index of monthly earnings in manufacturing. It includes salaried employees. 46 * LABOR MARKET FLEXIBILITY Costa Rica Employees. Labor force sample survey. July of each year. Includes salaried employees and insured persons. Ecuador ILO. El Salvador ILO. 1961-85: wages in the department of El Salvador. 1986-96: wages in the whole country. Guatemala ILO. Wages of employees. Prior to 1974: Guatemala City. Establishments of five or more persons employed. Mexico ILO. October of each year. Methodology revised in 1986. Bank of Mexico, from the Encuesta Industrial. Panama 1961-84: ILO. Wages of employees. 1985-96: Estadfstica Panamefia. Median of weekly salaries (in Balboas). The salaries are those of males in the metropolitan area. 1995 and 1996 were monthly; they were converted into weekly. Paraguay UN ECLAC. Wages of manual workers in Asunci6n; average for June and December. Peru ILO. Wages of employees in Lima. UN ECLAC. Wages of laborers in the private sec- tor in the metropolitan area of Lima. Average for 12 months. Uruguay ILO. Wages of employees. Index of average monthly wages in the private sector. Venezuela 1964-73: ILO. Manufacturing industry; 1974-89: Estadfsticas Socio-laborales de l'enezuela. Series Hist6ricas, 1936-1990, Tomo I; 1990-96: Industria fabril; 1990-96: ECLAC. Aver- age income per urban worker, deflated by the variation in consumer prices in the Caracas metropolitan area. United States ILO. Private sector: production and construction workers and nonsupervisory employees. EMPLOYMENT Argentina Informe Econ6mico No. 21, 1997, and Subsecretaria de Programaci6n Macro- econ6mica. Greater Buenos Aires. 1997 is for the first semester, other years are average of two surveys. Total urban employment, Ministerio de Trabajo. Bolivia ILO. Persons aged 10 years and older; civilian labor force employed in urban areas. Brazil 1976-88: ILO. Persons aged 10 and older; civilian labor force employed. Excludes rural population of Rodonia, Acre, Amazonas, Roraima, Para, and Amapa. September of each year. 1989-97: Monthly Bulletin of the Central Bank of Brazil (February, 1997, table 1.8). REVISITING AND EXPANDING OKUN COEFFICIENTS * 47 Chile ILO. Civilian Labor force employed. Persons aged 15 and older. Fourth quarter of each year. Indicadores Econ6omicos y Sociales 1960-1985; Bulletin Mensual del Banco Central de Chile; Statistical Synthesis of Chile 1984-1988; and Compendio Estadfstico 1996. Employ- ment in the Greater Santiago Region. Annual average of four surveys. Colombia ILO. Persons aged 12 and older. Seven main cities of the country. September of each year. 1994-96: data are from Departamento Administrativo Nacional de Estadistica (DANE). Costa Rica ILO. Persons aged 12 years and older. Civilian labor force employed. July of each year. IICE (Instituto de Investigaciones de Ciencias Econ6micas). Direccion General de Estadisticas y Censos 1980-96. Data are from household surveys. Guatemala Bank of Guatemala. Number of workers enrolled in the Guatemalan Social Security System. It includes banking, insurance, and public administration sectors. Mexico INEGI. Results from the Encuesta Nacional de Empleo Urbano (ENEU) for the 16 largest cities. Panama ILO. Persons aged 15 and older. August of each year. Peru Metropolitan Lima. 1994-96 were calculated with data obtained also from Ministerio de Trabajo y Promoci6n Social, Direcci6n General de Empleo y Formaci6n Profe- sional (Compendio Estadfstico 1990-91). Based on a survey of establishments with 100 or more employees; and Boletin del Banco Central de Reserva del Peru. Uruguay ILO. Urban employment. Includes professional army; excludes compulsory military service. Venezuela Central Bank of Venezuela. OCEI household sample survey. United States ILO. Persons aged 16 and older. Civilian labor force employed. APPENDIX 2: REGRESSION RESULTS FOR THE LAST TEN-YEAR WINDOW-WAGE, EMPLOYMENT, AND UNEMPLOYMENT OKUN COEFFICIENTS Table A.2.1 Regression Results for the Last Ten-Year Window Rolling Wage, Unemployment, and Employment Okun Coefficients Unemployment Employment Wages First differences HP filter (y-50, u=50) First differences HP filter (y 50, Liu50) First differences HP filter (y=50, u- 50) Country Coefficlent (SE) (R ) Coefficient (SE) (R2) Coefficient (SF) (W2) Coofficient (SE) (R2) Coefficient (SE) (RP2) Coefficient (SE) (R2) Argentina -0.21 0.12 0.26 -0.11 0.11 0.11 0.14 0,11 0.14 0.07 0.11 -0.02 0.64 0.16 0.64 0.84 0.26 0.52 Bolivia -0.38 0.37 0.11 -0.08 0.21 0.01 0.27 0.89 0.01 0.81 0.43 0.23 0.37 2.34 0.22 2.49 1.20 0.30 Brazi -0.19 0,04 0.75 -0.19 0.03 0.78 0.01 0.17 0.00 -0.01 0.10 -0.08 2.84 2.02 0.18 1.15 1.26 0.06 Ch le -0.20 0.25 0.07 -0.16 0.19 -0.14 0.78 0.47 0.23 0.32 0.32 0.00 -0.21 0.28 0.06 0.68 0.32 0.20 Colombia - 0.39 0.26 0.20 -0,37 0.29 0.13 -0.17 0.21 0.07 -1.17 1.38 0.04 0.09 0.11 0.06 0.90 0.64 0.16 Costa Rica -0.35 0.15 0.39 -0.32 0.10 0.47 0.19 0.71 0.01 O,55 0.25 0.33 0.19 1.75 0.00 0.70 0.78 0.07 Guaternala -0.75 0.35 0.36 -0.72 0.21 0.54 -0.85 0.15 0.78 0.14 0.45 -0.04 0.65 0.69 0.09 0.32 1.56 0.00 Mexico -0.22 0.06 0.56 -0 lb 0.06 0.38 0.30 0.14 0.19 0.05 0.14 - 0.12 0.12 1.44 0.00 2.29 0.33 0.82 Panama -0.29 0.03 0.93 -3.21 0.02 0.90 0.32 0.33 0.12 0.25 0.07 0.54 0.38 O.b5 0.05 0.17 0.23 0.04 Peru -0.18 0.08 0.38 -0.10 0.05 0.27 0.56 O.b6 0.10 0.19 0.22 0.06 4.74 2.39 0.33 2.51 0.29 0.88 Uruguay -0.5 0.10 0.20 -D.18 0.11 0.04 0.53 11.20 0.00 -4.99 2.27 0.35 0.04 1.21 0.00 0.91 0.33 0.41 Venezuela -0.23 0.10 0.41 -0.24 0.09 0.39 0.06 0.13 0.03 0.08 0.09 0.06 0.26 0.63 0.02 1.21 0.46 0.41 Un ted States -0.37 0,11 0.59 -0.50 0.06 0.85 0.62 0 13 0.73 0.67 0.08 0.87 0.18 0.24 0.05 0.37 0.22 0.22 REVISITING; AND EXPANDING OKUN COEFFICIENTS * 49 Figure A.2.a Ten-Year Rolling Wage Okun Coefficients: Argentina, Bolivia, Chile, Costa Rica, Peru, and the United States (Differences) 2.5 2 USWILO ARWCEP 2 - BOWILO -CHWILOT- Y \/. 1.5 CRWILO PEWCEP \ 0, 0n/' y -0 5 -1- -1,5 -Il lll ll ll lX 1971 1976 1981 1986 1991 1996 Figure A.2.b Ten-Year Rolling Employment Okun Coefficients: Argentina, Bolivia, Chile, Costa Rica, Peru, and the United States (Differences) USEILO -AREDO1 1 - BOEILO - CHEDSA 0~8 --/ CREILO | | -~~~PEED02/\ / 0~6 0 4 0,2 1971 1976 1981 1986 1991 1996 50 * LABOR MARKET FLEXIBILITY Figure A.2.c Ten-Year Rolling Unemployment Okun Coefficients: Argentina, Bolivia, Chile, Costa Rica, Peru, and the United States (Differences) 0,1 0 -0.2 -0,3 -0.4 -0.5 -~CH DOM ~CR PE 1969 1974 1979 1984 1989 1994 Figure A.2.d Ten-Year Rolling Wage Okun Coefficients: Brazil, Venezuela, and the United States (Differences) USWILO 0.8 - -'- VEWICD --BRWCEP | 0.6 0.4 0.2 -0. -0.4 -0.6 1 1 l 1996 1971 1976 1981 1986 199' 1996 REVISITING AND EXPANDING OKUN COEFFICIENTS * 51 Figure A.2.e Ten-Year Rolling Employment Okun Coefficients: Brazil, Venezuela, and the United States (Differences) 0.7 0.6 0.5 0.4 0.3 ~ - USEILO - BREILO 0.2 -VEEILO 0.1 0 -0. 1 -0.2- 1971 1976 1981 1986 1991 1996 Figure A.2.f Ten-Year Rolling Unemployment Okun Coefficients: Brazil, Venezuela, and the United States (Differences) -0.2 -0.3 -0.4 -0.5 -BR -VE -0.6 I i i l l i l l i l 1969 1974 1979 1984 1989 1994 52 * LABOR MARKET FLEXIBILITY Figure A.2.g Ten-Year Rolling Wage Okun Coefficients: Mexico and the United States (Differences) 2 USWILO MXWCEP 1.5 0. 1971 1976 1981 1986 1991 1996 Figure A.2.h Ten-Year Rolling Employment Okun Coefficients: Mexico and the United States (Differences) 1.2 - USEILO 1 - MXEDO 0.8 0.6 0.4 0.2 1971 1976 1981 1986 1991 1996 REVISITING AND EXPANDING OKUN COEFFICIENTS * 53 Figure A.2.i Ten-Year Rolling Unemployment Okun Coefficients: Mexico and the United States (Differences) 0 2 -0.1 -0.2 -0.3 -0.4 -0.5 us UZ MX -0.6 I I I I I 1969 1974 1979 1984 1989 1994 Figure A.2.j ,Ten-Year Rolling Wage Okun Coefficients: Colombia, Panama, Uruguay, and the United States (Differences) 2 - -USWILO - COWILO 1.5 - -URWILO PAWI LD 0.5 0 1971 1976 1981 1986 1991 1996 54 * LABOR MARKET FLEXIBILITY Figure A.2.k Ten-Year Rolling Employment Okun Coefficients: Colombia, Panama, Uruguay, and the United States (Differences) 1 -~USEILO -COEILO 0.8 U UREILO1 0.6 0.4 0.2 0 -0.2 -0.4 -0.6- 1971 1976 1981 1986 1991 1996 Figure A.2.1 Ten-Year Rolling Unemployment Okun Coefficients: Colombia, Panama, Uruguay, and the United States (Differences) 0.2 0.1 0 -0.1 -0 2 -0.3 -04 5s ,]~c U7R 1969 1974 -1979 1984 1989 1994 REVISITING AND EXPANDING OKUN COEFFICIENTS * 55 Figure A.2.m Ten-Year Rolling Wage Okun Coefficients: Guatemala and the United States (Differences) 1.2 USWI LO 1 ,GU\/\ILO 0.8 0.6 0.4 0.2 0~~~~~~~~~~~ -0.2 -0.4 - I I 1971 1976 1981 1986 1991 1996 Figure A.2.n Ten-Year Rolling Employment Okun Coefficients: Guatemala and the United States (Differences) 1.2 0.9 0.6 0.3 0 -0,3 -0.6 - -USEDO -0.9 -1,2 -1.5- I I I I I 1996 1971 1976 1931 1986 1 991 1996 56 * LABOR MARKET FLEXIBILITY Figure A.2.o Ten-Year Rolling Unemployment Okun Coefficients: Guatemala and the United States (Differences) 0 * -0.1 -0.2 03; -0.4 -0.5 -0.6 -0.7 -0.2 - U -0.9 1969 1974 1979 1984 1989 1994 57 NOTES I See Williamson (1995) for the accepted definition of tunately, owing to the lack of reliable data on the hours 'Washington Consensus." worked in Latin America, evidence for these countries is scarce. 2 See Nickell (1997) for a recent review on this subject. 9 This was identified by Okun himself and documented 3 See theJournal of Economic Perspectives (fall 1997) for a in Okun (1973). review of this subject. 10 Documented by Okun (1973), Perry (1971), and Gor- 4 The literature has been split in following Okun's origi- don (1984). nal convention of reporting the coefficient as the change in output per percentage point of unemployment (that l Okun (1973), Perry (1971), Gordon (1984), and Kuh is, 3) versus its reciprocal, which indicates changes in (1966). unemployment caused by a percentage change in output (that is, 0.33). For clarity, and because we always have 12 Haddy and Tolles (1957) wrote the seminal paper on labor market variables on the left-hand side of the wage differential narrowing with a small output gap. regression, we will use the second option. Thus, a larger Okun (1973) provides the evidence. coefficient indicates greater sensitivity of labor market indicators with respect to output. 13 The lack of a proper lag structure was first postulated by Kuh (1966). Improvements on the estimation tech- S All the transmission mechanisms were proposed by niques of the coefficient, especially separating a short- Okun in his 1962 and 1973 papers. To my knowledge, term from a long-term coefficient, have corroborated his no one has proposed any new transmission mechanisms assertion, but the effect is small. between these two variables. 14 See Perry (1971) and Gordon (1984) for reviews of the 6 This has been consistently documented not only in the estimations up to that time. United States but also in Japan and in the other OECD countries. See Kuh (1966), Perry (1971 and 1977), Okun 15 See Gordon (1984) and Kaufman (1988) for a recent (1973), Gordon (1984) for reviews, and Hamada and decomposition of the literature. Kurosaka (1984), Knoester (1986), and Kaufman (1988) for documentation of a very strong effect in Japan and in 16 Hamada and Kurosaka (1984) estimate that the coeffi- Europe. For papers on labor force participation from a cient for Japan is 28. Kaufman (1988) presents evidence supply perspective, see Friedman (1962) and Knoester for Canada, Germany, Japan, Sweden, the United King- (1986). The U.S. Bureau of Labor Statistics has esti- dom, and the United States using unemployment data mated that in a recession, for every 10 workers listed as adjusted by the U.S. Bureau of Labor Statistics. Moosa unemployed above 4 percent, there exist 3 additional (1997) uses Kalman filters and seemingly unrelated potential workers not actively seeking a job. regressions for Canada, Germany, Italy, Japan, the United Kingdom, and the United States 7 Maloney (1997) and Hernandez Licona (1998) present evidence for Mexico. 17 See Sumner and Silver (1989) for more on this topic. 8 See endnote 5: All of the papers present procyclical 18 There is little debate that employment deviates around hours worked. Moreover, Black and Russell (1969) and a slowly evolving trend exogenously determined by nat- Thurow and Taylor (1966) review the evidence. Unfor- ural population growth. 58 * LABOR MARKET FLEXIBILITY 19 Evans (1989) and Clark (1987) raised the point in the 29 Even for the United States, for which various authors Okun literature, but the seminal paper on the subject is suggest the introduction of a break in the slope coeffi- by Nelson and Plosser (1982). cient, Banerjee and others (1993) have suggested that the break varies if it is allowed to be determined erndoge- 20 See Greene (1990) for an introduction to the subject nously by the data. and Hamilton (1994) for a more rigorous exposition. 30 Using a Kalman filter to decompose the series is left for 21 For output, there is an extensive literature on the sub- future research. Another notable omission from the esti- ject just for the United States, and the matter is not set- mation techniques is a vector autoregression approach to tled. Perron (1988) argued that deviations of log postwar create the cyclical components of the series. The interpre- U.S. real GDP with a break in the trend coefficient in tation of the cyclical components is not business cycles, 1973 are stationary. Banerjee and others (1993) argue but how innovations in output are translated into innova- that the stationarity depends on whether the break is tions in any of the labor market performance variables. imposed or endogenously determined by the data. For the unemployment case seeJournal of Economic Perspectives 31 To my knowledge no one has computed Okun wage (1997), and Evans (1989), Perron (1988), and Weber ratios for either the United States or other OE>CD coun- (1995). tries. 22 Weber (1995) presents a review of the effects of esti- 32 Estimates for first differences are presented in appen- mation techniques for the United States Although the dix 1. The discussion concentrates on results from the effects are significant, the magnitude in the difference of HP estimation because there are breaks in output and point estimates is small. wage trends even for 10-year windows. The HP proce- dure is more appropriate because first differences 23 The statistics for Guatemala come from the country's assumes a stationary growth trend. National Security Agency, implying that only formal employment is included. The series was included 33 Actually it is the sum of the innovation of the new because of the lack of another source, and because of the year plus the fluctuation from the year dropped. time period covered. 34 This result is in line with the literature. See Gordon 24 The crude, but accurate, reasoning from some of the (1984) for a review He argues that the coefficient is 0.4 agencies is that in the "campo" one cannot be unem- instead of 0.3, as Okun originally postulated. ployed. 35 In general, employment is procyclical while unemploy- 25 Labor productivity is increasing if the slope of the out- ment is countercyclical; thus the coefficients have oppo- put series is steeper than the employment series, and site signs. For clarity, I did not change the signs of either. decreasing if the slope is flatter. 36 Maloney (1997) and Hernandez Licona (1998) show 26 Dickey-Fuller tests are available from the author on strong evidence of increased labor participation in Mex- request. ico as real wages fall. 27 Engle and Granger (1987) argued the point. 37 It is not clear why the wage Okun coefficient rose in Johansen's maximum likelihood cointegration estima- 1987. Moreover, it is not clear why there was a rise in tion was actually the author's original choice of estima- 1994. tion technique, but the variables rejected the tests. 3 The instability of the first difference coefficient is that 28 In some countries more than one break was allowed in the break in output growth trend lasts until about 1983; the trends. thus the estimation is biased. REVISITING AND EXPANDING OKUN COEFFICIENTS * 59 39 Moreover, throughout the 1980s the country's econ- the period. TheJournal of Econiomic Perspectives (1997) omy did not grow at all. dedicated an entire issue to this subject. 40 I owe this suggestion to Jaime Saavedra who suggested 47 Okun coefficients are inherently cyclical measures; that the fall in real wages in Peru was overestimated dur- they are long-term estimates of the cyclical variations. ing that period. The fact that in first differences the sen- They are not long-term elasticities between the variables. sibility has been increasing since 1990 is attributable to a The point may seem obvious, but there has been some break in the growth trend in 1990. degree of confusion in the literature about the interpre- tation of Okun coefficients. 41 Recall that fluctuations from year to year can be inter- preted as innovations. 48 Note that in fluctuations in hours worked per worker, labor hoarding is no longer an issue. The reason is that 42 The point cannot be overemphasized that had the 1993 unless there are permanent increases or decreases in the observation been dropped, the coefficient would have work week, the effect will simply be a temporary damp- maintained 1992 levels. In other words, quantity adjusting ening of the effect of output on employment. Permanent is not that much lower today than it was in the early 1990s. changes in the work week will have one-time effects on output. In the long term, fluctuations in hours worked is 43 Appendix 1 presents the coefficients under first differ- simply measurement error. ences, and the increase in the coefficients is more pro- nounced. 49 A good exercise for further research is to determine how much of the increments in output per unit of 44 Cyclical fluctuations are relatively unimportant com- employment is attributable to productivity, and how pared with the fall in real wages. Unemployment much to increases in capital per worker. dropped until 1990, aided by the fall in real wages. 5'Although there are changes in the long-term slope of 45 The corresponding jump in using the HP estimation employment, most of the variation comes from changes process is from 0.15 to 0.20. Moreover, an extension in output growth. Inspection of the employment series with preliminary data from 1997 indicates a further rise in appendix 2 confirms this assertion. In addition, "true" of the coefficients. The genesis of this paper was a dis- long-term employment growth is inherently determined cussion about the effectiveness of the labor reform. by population growth. Some argued that since unemployment had not fallen, reforms were not effective; I argued that one had to look S Martinez (1997) has argued that wage increases sur- at growth performance. passed productivity gains in the period. 46 Unemployment is not discussed because in the long 52 Brazil has grown vigorously since 1992. However, the term, unemployment should be stationary or exhibit fact that total employment decreased since 1990 makes small variations; that is, it cannot increase or decrease one suspect the data. Various sources showed employ- continuously. There has been a renewed debate on the ment reduction in the 1990s. level and stability of the natural rate, the NAIRU, in the United States. Estimation of the NAIRU for each of the 53 There was some degree of social unrest since 1985 countries is beyond the scope of this paper, particularly until the peace accords were signed in 1989, but militant given the instability of prices in Latin America during activities were reduced in 1985. 60 BIBLIOGRAPHY Adams, Charles, and David T. Coe. 1990. 'A Systems . 1993. "Okun's Law and Potential GNP." Federal Approach to Estimating the Natural Rate of Reserve Board of Governors, Federal Reserve Sys- Unemployment and Potential Output for the tem, June. Federal Reserve, Washington, D.C. United States." LIF Staff Papers 37(2). Interna- tional Monetary Fund, Washington, D.C. De Gregorio, Jose. 1992. "Economic Growth in Latin America."Journal of Development Economics 39(July): Arkelof, George, William T. Dickens, and George L. 59-84. Perry. 1996. "The Macroeconomics of Low Infla- tion." Brookings Papers on Economic Activity 1: 1-59. Dolado, Juan J., David L6pez-Salido, and Juan Luis Brookings Institution, Washington, D.C. Vega. 1996. "Short- and Long-Run Phillips Trade- Offs and the Cost of Disinflationary Policies." Dis- Attfield, Clifford L. F., and Brian Silverston. 1994. cussion Paper No. 1483, October. Human "Okun's Coefficient: A Comment." Discussion Resources and International Macroeconomics, Paper No. 94/377, March. University of Bristol, Centrc for Economic Policy Research, London. Bristol, England. Dolado, Juan J., and Juan F. Jimeno. 1997. "The Causes Banerjee, Anindya, Juan J. Dolado, John W Galbraith, of Spanish Unemployment: A Structural VAR and David F. Hendry. 1993. "Co-Integration, Error Approach." European Economic Review 41: Correction, and the Econometric Analysis of Non- 1281-1307. Stationary Data." Advanced Texts in Econometrics. New York: Oxford University Press. Dolado, Juan J., Tim Jenkinson, and Simon Sosvilla- Rivero. 1990. "Cointegration and Unit Roots: A Bentolila, Samuel, and Juan J. Dolado. 1994 "Labour Survey." Banco de Espafia, Servicio de Estudios, Flexibility and Wages: Lessons from Spain." Eco- No.9005. Madrid, Spain. nomic Policy: A European Forum 9(18): 53-99. Engle, Robert F., and Clive W J. Granger. 1987. "Co- Black, Stanley W., and Robert R. Russell. 1969. 'An Integration and Error Correction: Representation, Alternative Estimate of Potential GNP" Review of Estimation and Testing." Econometrica 55(2): Economics and Statistics 51(1): 70-76,. 251-76. Blanchard, OlivierJ., and Danny Quah. 1989. "The Evans, George W 1989. "Output and Unemployment Dynamic Effects of Aggregate Demand and Supply Dynamics in the United States: 1950-1985."Jour- Disturbances." The American Economic Review 79(4): nal of Applied Econometrics 4: 213-37. 655-73. Fischer, Stanley. 1977. "Long-Term Contracts, Rational . 1989. "A Traditional Interpretation of Macroeco- Expectations and the Optimal Money Supply nomic Fluctuations." The American Economic Review Rule."Journal of Political Economy 85(1): 191-206. 79(5): 1146-64. Friedman, Benjamin M., and Michael L. Wachter. 1974. Clark, Peter K. 1987. "The Cyclical Component of U.S. "Unemployment: Okun's LaNv, Labor Force and Economic Activity." QuarterlyJournal of Economics Productivity." The Review of Economics and Statistics 102(4): 797-814. 56(2): 167-76. REVISITING AND EXPANDING OKUN COEFFICIENTS * 61 Friedman, M. 1962. Price Theory: A Provisional Text. . 1997. "Trends, Cycles and Autoregressions." The Chicago: Aldine. EconomicJournal 107(440): 192-201. Royal Eco- nomic Society, London. Froot, Kenneth A., and Emil Dabora. 1998. "How are Stock Prices Affected by the Location of Trade?" NBER Hernandez Licona, Gonzalo. 1998. "The Effects of Working Paper No. 6572: 1-22. National Bureau of Household Poverty on Participation: Total Labor Economic Research, Cambridge, Mass. May. Supply and Unemployment in Mexico." Instituto Tecnol6gico Aut6nomo de Mexico (ITAM). Gordon, RobertJ. 1973. "The Welfare Cost of Higher Unemployment." Brookings Papers on Economic Hodrick, R., and Edward Prescott. 1997.Journal of Activity 1: 133-205. Brookings Institution, Wash- Money, Credit and Banking 29(1): 1-16. ington, D.C. IFS (International Financial Statistics). Various issues. Inter- . 1980. "Arthur M. Okun." Brookings Papers on national Monetary Fund, Washington, D.C. Economic Activity 1: 1-5. Brookings Institution, Washington, D.C. ILO (International Labour Office) Latin America. 1996. Editorial. Labor Ouilook 1996. . 1984. "Unemployment and Potential Output in the 1980s." Brookings Papers on Economic Activity 2: Journal of Economic Perspectives. 1997. Fall. 130-33. Brookings Institution, Washington, D.C. Kaufman, Roger T. 1988. "An International Comparison Greene, William H. 1990. Econometrics Analysis. New of Okun's Laws."Journal of Comparative Economics York: Macmillan Publishing Company. 12: 182-203. Gregory, Peter. 1986. "The Myth of Market Failure: Kennedy, Peter. 1992. A Guide to Econometrics. 3rd ed. Employment and the Labor Market in Mexico." Cambridge, Mass.: MIT Press. World Bank Research Publication. World Bank, Washington, D.C. Knoester, Anthonie. 1986. "Okun's Law Revisited." Weltwirtschaftliches Archiv 122(4): 657-66. Haddy, Pamela, and Arnold Tolles. 1957. "British and American Changes in Inter-Industry Wage Struc- Kuh, Edwin. 1966. Measurement of Potential Output. ture under Full Employment." Review of Economics The American Economic Review 56(4): 758-76. and Statistics 39(4): 408-14. Loayza, Norman, and J. Humenrto L6pez. 1997. "Mis- Hamada, Koichi, and Yoshio Kurosaka. 1984. "The Rela- alignment and Fundamentals. Equilibrium tionship between Production and Unemployment Exchange Rates in Seven Latin America Countries." in Japan." European Economic Review 25: 71-94. Mimeo. February. World Bank, Washington, D.C. Hamilton, James D. 1994. Time Series Analysis. Princeton, Loayza, Norman, and Luisa Palacios. 1997. "Economic NewJersey: Princeton University Press. Reform and Progress in Latin America and the Caribbean." Policy Research Working Paper, Sep- Harvey, Andrew. 1992. Time Series: Volume l: Introduction. tember. World Bank, Washington, D.C. Elgar Reference Collection, Intemational Library of Criti- cal Writings in Econtometrics, 5. Aldershot, U.K.: Lodewijks, John. 1988. "Arthur Okun and the Lucasian Elgar. Distributed in the United States by Ashgate, Critique." Australian Economic Papers 27(51): Brookfield, Vermont. 253-71. 62 * LABOR MARKET FLEXIBILITY Lustig, Nora, and Darryl McLeod. 1996. "Minimum Nelson, C. R., and C. I. Plosser. 1982. "Trends and Ran- Wages and Poverty in Developing Countries: Some dom Walks in Economic Time Series. 'Journal of Empirical Evidence." Brookings Discussion Papers Monetary Economics 10(September): 139-62. in International Economics No. 125 June). The Brookings Institution, Washington, D.C. Nickell, Stephen. 1997. "Unemployment ar.d Labor Market Rigidities: Europe versus Nor-h America." Lutkepohl, Helmut. 1993. Introduction to Mvlultiple Time Journal of Economic Perspectives 11 (3): 55-93. Series Analysis. 2nd ed. New York: Springer-Verlag. OIT (Organizaci6n Internacional de Trabajo). 1996. Maloney, William. 1997. "Labor Market Structure in Panorama Laboral '96: Informe deAm6rica Latina y el LDCs, Time Series on Competing Views." Uni- Caribe. Geneva, Switzerland. versity of Illinois at Champaign-Urbana, Ill. Okun. Arthur M. 1962. "Potential GNP: Its Measure- Marimon, Ram6n, and Fabrizio Zilibotti. 1996. "Actual' ment and Significance; Proceedings of the Busi- versus 'Virtual' Employment in Europe: Is Spain ness and Economic Statistics Section of the Ameri- Different?" ECO No. 96/21, May. European Uni- can Statistical Association," pp. 98-104. Reprinted versity Institute, Florence. in Arthur M. Okun, The Political Economy of Pros- perity (pp. 132-145). 1970. Washington, D.C.: Marquez, Gustavo, and Carmen Pages-Serra. 1997. Brookings Institution. "Structural Reform and Labor Market Perfor- mance in Latin America and the Caribbean during . 1973. "Upward Mobility in a High Pressure the 90's: Much Ado about Nothing?" Inter- Economy." Brookings Papers on Economic Activity American Development Bank, Washington, D.C. 4(1): 207-61. Brookings Institution, Washington, D.C. 1998. "Protecci6n al Empleo y Funcionamiento del Mercado de Trabajo: Una Aproximaci6n Com- Palley, Thomas I. 1993. "Okun's Law and the Asymmet- parativa." Inter-American Development Bank, ric and Changing Cyclical Behaviour of the USA Washington, D.C. Economy." International Review of Applied Economics 7(2): 144-62. Martinez, Lorenza. 1997. "Real Wages and Productivity in Mexico." Banco de Mexico Occasional Discus- Perron, Pierre. 1988. "Trends and Random Walks in sion Papers, Mexico City, Mexico. Macroeconomic Time Series: Further Evidence from a New Approach."Journal of Economic Dynam- Moosa, Imad A. 1997. "A Cross-Country Comparison of ics and Control 12(2/3): 297-332. Okun's Coefficient."Journal of Comparative Econom- iCs 24(3): 335-56. Perry, George L. 1971. "Labor Force Structure, Potential Output, and Productivity." Brookings Papers on Eco- Nehru, Vikram, and Ashok Dhareshwar. 1994. "New Esti- nomic Activity 3: 53-65. Brookings Institution, mates of Total Factor Productivity Growth for Devel- Washington, D.C. oping and Industrial Countries." Policy Research Working Paper No. 1313, June. World Bank Intema- . 1977. "Potential Output and Productivity." tional Economics Department, Washington, D.C. Brookings Papers on Economic Activity 1: 11-60. Brookings Institution, Washington, D.C. Nehru, Vikram, Eric Swanson, and Ashutosh Dubey. 1993. "Sources, Methodology and Results." Policy Prachowny, Martin F. J. 1993. "Okun's Law: Theoretical Research Working Papers (April). World Bank, Foundations and Revised Estimates." The Review of Washington, D.C. Economics and Statistics 75(2): 331-36. 63 Rama, Martin. 1995. "Do Labor Markets Policies and Thurow, Lester C., and L. D. Taylor. 1966. "The Interac- Institutions Matter? The Adjustment Experience tion between the Actual and the Potential Rates of in Latin America and the Caribbean." Working Growth." The Review of Economics and Statistics paper supported by World Bank's Research Com- 48(4): 351-61. mittee through grant RPO 678-46. World Bank, Washington, D.C. May. Watts, Martin, and William Mitchell. 1991. "Alleged Instability of the Okun's Law Relationship in Aus- Revenga, Ana L., and Samuel Bentolila. 1995. "What tralia: An Empirical Analysis." Applied Economics 23: Affects the Employment Rate Intensity of 1829-38. Growth?" Documento de Trabajo No. 9517, Banco de Espafia. Madrid, Spain. Weber, Christian E. 1995. "Cyclical Output, Cyclical Unemployment, and Okun's Coefficient: A New Reyes Trejo, Sa6l. 1973. "Industrializaci6n y empleo en Approach."Journal of Applied Econometrics 10(4): Mexico." Fondo de Cultura Econ6mica, Mexico. 433-45. Sheehan, Richard G., and Frank Zahn. 1980. "The Vari- Williamson, Oliver E. 1995. "What Washington Means ability of the Okun Coefficient." Southern Economic by Policy Reform." In Stephen Haggard, ed., The Journal, 47 (October): 488-97. International Political Economy and the Developing Countries. Elgar Reference Collection Library of Interna- Siebert, Horst. 1997. "Labor Market Rigidities: At Root tional Political Economy, 7: 514-28. Aldershot, U.K.: of Unemployment in Europe."Journal of Economic Elgar. Distributed in the United States by Ashgate, Perspectives 11(3): 37-53. Brookfield, Vermont. Solow, Robert. 1957. "Technical Change and the Aggre- World Bank. 1995a. "Reformas Laborles y Econ6micas gate Production Function." Review of Economics and en America Latina y el Caribe." Informe sobre el Statistics 39(3): 312-20. Desarrollo Mundial 1995: Perspectivas Regionales. Washington, D.C. . 1965. Unpublished Presidential Address to Econometric Society, delivered December, as cited . 1995b. "Workers in an Integrating World." World in Okun (1973). Development Report 1995, World Development Indica- tors. Washington, D.C. Sumner, Scott, and Stephen Silver. 1989. "Real Wages, Employment, and the Phillips Curve."Journal of . 1996. "Labor Reform and Job Creation: The Political Economy 97(3): 706-20. Unfinished Agenda in Latin America and Caribbean Countries." Second Annual World Taylor, John B. 1979. "Staggered Price Setting in a Bank Conference on Development in Latin Macro Model." American Economic Review 69(2): America and the Caribbean, Bogota, Colombia, 108-13. June 30-July 2. THE WORLD BANK 1818 H Street, N.W Washington, D.C. 20433 USA Telephone: 202-477-1234 Facsimile: 202-477-6391 Telex: MCI 64145 WTORLDBANK MCu 248423 NVORLDBANK Internet: wwvw.worldbank.org E-mail: books@Nworldbank.org ISBN 0-8213-4489-7