Structural Growth in Labor Costs: Evidence from the Employment Cost Index

Transcription

1 Journal of Forensc Economcs 25(1), 2014, pp by the Natonal Assocaton of Forensc Economcs Structural Growth n Labor Costs: Evdence from the Employment Cost Index Gary R. Albrecht and Kurt V. Krueger * Abstract Wage growth forecastng s a necessary part of forensc economcs. In ths paper, we present a tme seres methodology to test whether wage and compensaton growth n the Uned States vares by ndustry and occupaton. If growth vares, then the common use of "all-worker" net dscount or wage growth rates would not be accurate for every forensc economc case. Usng the Employment Cost Index, we fnd that total compensaton, wage, and benef growth n some, but not all, ndustres and occupatons has been sgnfcantly dfferent from that of the wage growth of all workers. That fndng may concern the forensc economst who needs to construct a varety of net dscount or wage growth rates. As an alternatve to constructng multple forecasts, ths paper provdes estmated ndustry and occupatonal specfc dfferentals from the growth n all workers' wages. I. Introducton Wage growth forecastng s a necessary part of forensc economcs. In ths paper, we present a tme seres methodology to test whether wage and compensaton growth n the Uned States vares by ndustry and occupaton. If growth vares, then the common use of all-worker net dscount or wage growth rates would not be accurate for every forensc economc case. Usng the Employment Cost Index, we fnd that total compensaton, wage, and benef growth n some, but not all, ndustres and occupatons has been sgnfcantly dfferent from that of the wage growth of all workers. That fndng may concern the forensc economst who needs to construct a varety of net dscount or wage growth rates. As an alternatve to constructng multple forecasts, ths paper provdes estmated ndustry and occupatonal specfc dfferentals from the growth n all-workers wages. Armed wh the dfferentals and one net dscount or growth rate forecast based on all-workers wage growth, the forensc economst can adjust the all-worker forecast to a specfc ndustry or occupaton. The paper dscusses the ssue of wage growth varaton and provdes a tme seres method to measure. The data and emprcal results are presented along wh example applcatons. *Gary R. Albrecht, Albrecht Economcs, Inc., Wnston-Salem, NC; Kurt V. Krueger, John Ward Economcs, Prare Vllage, KS. 17

2 18 JOURNAL OF FORENSIC ECONOMICS II. The Issue of Wage Growth Varaton Industres and occupatons utlze technology and laborer sklls dfferently whch creates varaton n the growth of s laborers wages (Bronars and Famular, 1997). Reasons for an ndustry or occupaton exhbng growth n labor costs unque to nclude: dfferng utlzaton of technology n the partcular ndustry or occupaton that changes s margnal productvy of labor relatve to the entre economy; changes n the demand for the product produced by the specfc ndustry or occupaton relatve to all other products whch change relatve prce (and thus margnal revenue) to the entre economy; or, relatve prce changes of nputs used n the producton of the product may change the growth n labor costs n the ndustry or occupaton relatve to the entre economy. Hstorcal wage growth dfferences s a frequent topc n the general economc lerature wh many studes focusng on the wdenng gap of wages by worker skll level (Juhn, et al., 1993; Autor, et al., 2008). Whle specfc ndustry and occupaton wage growth measures are tracked, the macroeconomc varable of all-workers wage growth receves the most attenton as a forecasted varable. Whn forensc economcs, the wage growth of all workers s an mportant varable 1 and many forensc economsts use forecasts of all-worker wage growth produced by prvate frms or government agences. 2 Despe that forensc economc casework spans the gamut of workers by ndustry and occupaton, only a few forensc economc studes such as those by Pelaez (1991) and Payne, et al., (1999) have gven attenton to wage growth dspary (those two studes gve net dscount rates for a handful of ndustres but they dd not test for varaton n net dscount rates by ndustry). Trends n wage growth dspary should be accounted for because, absent cases nvolvng chldren, most forensc evaluatons call for wage growth forecasts over near perods of tme n whch the observed varatons of wage growth across ndustry and occupaton are lkely to persst. 3 III. Measurng Wage Growth Varaton Assume that the wages n any ndustry or occupaton (subscrpt ) at tme t s represented by W and that the wages of all workers at tme t s repre- 1The phrase net dscount rate appears n 102 past Journal of Forensc Economcs artcles. The concept of net dscount rates based on all-worker wage growth s a common enough statstc that leads the questonng n the Survey of Forensc Economsts paper seres (Slesnck et al., 2013). 2For example, s common to see forensc economsts usng the all-worker wage growth estmates of the Socal Secury Admnstraton. 3In 2012, the medan age of all employed persons was 42.3 years Bureau of Labor Statstcs data publshed from the Current Populaton Survey ( The worklfe expectancy of 42-year-olds ranges from 16 to 26 years based upon gender and educaton (Skoog, et al., 2011)..

3 Albrecht & Krueger 19 sented asw At. Defne the perod-to-perod rato n the wages n a chosen ndustry or occupaton to that of all workers as W (1) R W The perod-to-perod growth rate n At R can be wrten as (2) G R R 1 1. Conversely, changes n W and G can also be thought of as a lnk between the perod-to-perod W At as (3) W W W W (1 G ) At At G At 1 At1 At1 At1 W W W W If the growth dfferental seres G s statonary and non-zero, there s a constant mechansm separatng the wage growth n the partcular ndustry or occupaton from that of all workers. Econometrcally, we have to (1) determne the statonary ofg, and (2) test whether G s statstcally dfferent from zero. Studyng the tme seres propertes of the growth n medcal care costs, Bowles and Lews (2000) commented on the mportance of statonary n a forecasted tme seres. A covarance statonary seres has a tme nvarant mean and varance such that the seres fluctuates around a constant long-term mean. The unbased forecast of a statonary seres s s long-term mean. The forensc economc lerature has devoted much attenton to the statonary of the net dscount rate f the net dscount rate s not statonary, then s use produces based forecasts. If G s statonary and the all-worker net dscount rate s statonary, then G can be subtracted from the all-worker net dscount rate forecast to produce a vable net dscount rate forecast for the partcular ndustry or occupaton. An accepted test for determnng the statonary of a growth seres s the Phllps-Perron test. The Phllps-Perron test s an augmented Dckey-Fuller test that recognzes autocorrelaton and heteroscedastcy n the dsturbance. The Phllps-Perron test for a sngle mean n G s determned from the autoregressve model (4) G G ( 1) u where α s the constant, s the coeffcent on the lagged growth dfferental value, and u s the dsturbance. Snce frst order autocorrelaton mght be ex-

4 20 JOURNAL OF FORENSIC ECONOMICS pected n G, to get an unbased estmate of the sngle mean growth, the dsturbance adds a frst order autocorrelaton parameter so ut s specfed as 2 (5) u u t nd : N 1 ( 1) a 0,. The autoregressve model s estmated usng the Yule-Walker method whch frst estmates the parameters wh generalzed least squares, then φ1 wh ordnary least squares, and then usng 1 the parameters are re-estmated wh generalzed least squares. The generalzed least squares approach corrects for the heteroscedastcy n the varance of the tme seres observatons. Usng the standard error of the estmated mean, we perform a t-test to dscover whether the mean growth dfferental statstc for each ndustry and occupaton selected s sgnfcantly dfferent from zero. IV. Data In order to fnd the hstorcal dfferental growth n labor costs, we utlze the Employment Cost Index ( ECI ). 4 The Employment Cost Index s a prncpal federal economc ndcator 5 publshed by the Bureau of Labor Statstcs and s based on the change n the average straght-tme hourly cost of labor, free from the nfluence of employment shfts among occupatons and ndustres. Detaled nformaton on ECI concepts, coverage, and methods can be found n the BLS Handbook of Methods, Chapter 8, Natonal Compensaton Measures. 6 In 2006, the ECI underwent revson whch left many ndustres and occupatons whout data before For our analyss, we utlzed all of the so-called contnuy ECI ndexes that span back as early as September along wh the remanng top-level ndustry ECI seres that begn n The data are current through December By ndustry and occupaton, we use the three ECI measures of labor costs: total compensaton, wages and salares, and benefs. 8 The ECI seres utlzed are those that are seasonally adjusted. 9 The denomnator n each ndustry or occupaton from equaton (1) s the ECI for the wages and salares of all prvate ndustry workers. In Fgure 1, we show an example ncremental growth data seres as the total compensaton ECI for prvate ndustry offce and admnstratve support occupatons. The sold lne represents R, the smple rato of the offce and admnstratve support occupaton total compensaton ECI dvded by the allworker wage and salary ECI. If the offce and admnstratve support occupaton total compensaton ECI grew over tme at approxmately the same annual rate as the all-worker wage and salary ECI, the sold lne would be relatvely 4The methodology could be used n the same manner wh other wage tme seres data. 5For nformaton on prncpal federal economc ndcators, see OMB Statstcal Polcy Drectve No. 3, found December 27, 2013 on the Internet at 85.pdf. 6As of June 4, 2013 on the Internet at 7For a lstng of these ECI s, see Employment Cost Index Hstorcal Lstng Volume V accessed on the Internet December 27, 2013 at 8The sum of wages and salares and benefs equals total compensaton. 9A separate analyss usng not seasonally-adjusted data acheved nearly dentcal results.

5 Albrecht & Krueger 21 flat around the value of 100 across all tme perods. 10 In ths example, the sold lne R s not flat but ncreasng showng that offce and admnstratve support occupaton total compensaton growth has hstorcally outpaced all-worker wage and salary growth. The black dashed lne shows G, the quarter-to-quarter growth n the sold lne ncremental growth ndex. Our econometrc task wll be to dscover whether G s statonary and sgnfcantly dfferent from zero. Because the total compensaton G for offce and admnstratve support occupaton s relatvely flat (or non-trendng), a vsual apprasal suggests statonary. Whle the mean value of G looks posve, s hard to vsually predct whether the varance of G keeps s mean value statstcally dfferent from zero a t-test wll produce that sgnfcance value. Fgure 1. Growth n the total compensaton pad to offce admnstratve support workers relatve to the wages and salares pad to all workers V. Results In Table 1, we show the estmaton results of mean growth dfferentals by ndustry and occupaton for total compensaton, wages and salares, and benefs. The name of the ndustry or occupaton s shown n the frst column and the second column contans the seres analyss start date. All of the Phllps- Perron test statstcs were sgnfcant at the 99% confdence level supportng statonary n all of the G wage growth seres. 11 The estmates of G, the 10Whn the context of examnng the flatness of the tme seres rato of two wage ndexes, f the rato s flat then one ndex seres could be descrbed as havng mean revertng tendences to the other ndex seres. 11Ths artcle s electronc supplemental materals at the Journal of Forensc Economcs Internet se contans a table showng the Phllps-Perron and Durbn-Watson test statstcs.

6 22 JOURNAL OF FORENSIC ECONOMICS mean annual growth n the partcular ECI whch s n excess of the all-worker wage and salary ECI growth s presented along wh the level of statstcal sgnfcance of the growth beng dfferent from 0% (** represents 99% confdence that the annual growth s sgnfcantly dfferent from 0% and * represents 95% confdence). If an annual growth dfferental s not statstcally sgnfcant from 0%, then the growth n the wages and salares of all workers s a suable ndcator of the partcular labor cost growth for that ndustry or occupaton. If the growth dfferental s sgnfcantly dfferent from zero, the growth dfferental can be used to adjust an all-worker net dscount rate or wage growth level forecast to f the emprcal characterstcs of the partcular ndustry or occupaton. Table 1 Growth dfferentals n total compensaton, wages and salares, and benefs relatve to the wages and salares pad to all workers Startng Date Total Compensaton Wages & Salares Benefs Industry & Occupaton Prvate Industry All workers 1984Q1 0.24% ** 0.85% ** Management, busness, and fnancal 1985Q4 0.41% ** 0.30% * Professonal and related 1985Q4 0.40% ** 0.16% Sales and related 1986Q1 0.15% 0.04% Offce and admnstratve support 1985Q4 0.48% ** 0.17% ** Servce occupatons 1984Q1 0.06% 0.24% * 0.93% ** Goods-producng ndustres 1984Q1 0.14% 0.17% ** 0.78% Constructon 1985Q1 0.05% 0.36% ** Manufacturng 1984Q1 0.17% 0.13% * 0.78% Servce-provdng ndustres 1984Q1 0.30% ** 0.09% 0.89% ** Transportaton and warehousng 1985Q1 0.06% 0.51% ** Utles 1987Q4 0.76% ** 0.17% Fnancal actves 1985Q1 0.55% * 0.35% Educaton servces 1988Q4 0.65% * 0.32% Junor colleges, colleges, and unverses 1988Q4 0.63% * 0.26% Hospals 1986Q2 0.73% ** 0.48% * Nursng care facles 1992Q2 0.23% 0.21% State & Local Government All workers 1984Q1 0.55% * 0.17% 1.14% ** Professonal and related 1989Q2 0.21% 0.07% Offce and admnstratve support 1989Q2 0.41% * 0.15% Servce occupatons 1984Q1 0.74% ** 0.25% Educaton servces 1989Q1 0.20% 0.10% Schools 1984Q1 0.55% 0.24% Elementary and secondary schools 1984Q1 0.55% 0.24% Hospals 1988Q1 0.49% * 0.20% Publc admnstraton 1984Q1 0.57% ** 0.07% The Employment Cost Index does not contan a benefs ndex for every ndustry or occupaton.

7 Albrecht & Krueger 23 For total compensaton growth dfferentals, we utlzed 17 prvate ndustry and occupatons ECI seres and nne state and local government seres. All but one of the growth dfferentals were posve and 15 of the 26 were statstcally sgnfcantly dfferent from zero. At the top level for prvate ndustry, the annual growth dfferental between the total compensaton pad to all workers and the wages and salares pad to all workers s a statonary and statstcally sgnfcant at 0.24%. Ths result means that when forecastng the sum of earnngs and benefs, 0.24% should be subtracted 12 from an all-worker wage and salary net dscount rate or added to a wage and salary growth level forecast. Assume that a chosen all-worker wage and salary based net dscount rate s 2%. If the forensc economst s projectng earnngs and benefs n a case and he or she beleved that the all-worker classfcaton s relevant to the case, then the approprate total compensaton net dscount rate wll subtract 0.24% from the 2% all-worker wage-based net dscount rate. If a nomnal all-worker wage and salary growth forecast s 3%, then the approprate nomnal allworker compensaton growth forecast wll be calculated by addng 0.24% to the 3% all-worker wage and salary growth forecast. If a real, nflaton-free allworker wage and salary growth forecast s 0.75%, then the approprate real, nflaton-free all-worker compensaton growth forecast wll be calculated by addng 0.24% to the 0.75% all-worker wage and salary growth forecast. Lookng down the lst of prvate ndustry and occupaton total compensaton growth dfferentals, those that are statstcally sgnfcantly dfferent from 0% are n predomnantly whe-collar occupatons or hgh-sklled ndustres. 13 Blue-collar occupatons or low-sklled ndustres have not had growth n total compensaton whch s statstcally sgnfcantly dfferent from the growth n wages and salares of all workers. Ths result suggests that when the forensc economst s evaluatng the earnngs and benefs of blue-collar type workers, the net dscount rate or wage growth level forecast of all workers s suable n computng present value. The growth dfferental for all state and local government workers total compensaton s over double that of the growth dfferental for all prvate ndustry workers (0.55% versus 0.24%). When measurng the growth dfferentals usng wages and salares by ndustry and occupaton, we found only three of the 26 ndustry and occupatons had statstcally sgnfcant posve growth dfferentals whe-collar orented management, busness, and fnancal; offce and admnstratve support; and, hospals). Sgnfcant and negatve growth dfferentals were n blue-collar orented servce occupatons, goods-producng ndustres, constructon, manufacturng, and transportaton and warehousng. As an example of usng a negatve growth dfferental, when determnng the present value of the wage earnngs of a constructon worker, the forensc economst would add 0.36% to hs or her all-worker wage and salary net dscount rate (or subtract 0.36% from hs or her all-worker wage and salary growth level forecast). 12Geometrc subtracton would be requred. For example, to geometrcally subtract a growth rate of 0.25% from a net dscount rate of 2%, the calculaton would be % 1 2% %. To geometrcally add a growth rate of 0.25% to a net dscount rate of 2%, the calculaton would be % 1 2% % 13Ths s the same result found by Juhn, et al. (1993).

8 24 JOURNAL OF FORENSIC ECONOMICS In all ndustres and occupatons except goods-producng and manufacturng ndustres, the benef growth dfferentals were posve and statstcally sgnfcant. Ths result suggests that nearly always net dscount rates or wage level growth forecasts based on the wages and salares of all workers are not approprate for computng the present value of total employment benefs. 14 VI. Concluson Although s an mportant part of forensc economcs, wage growth by ndustry and occupaton has receved ltle attenton n the lerature. In ths paper, usng a tme seres methodology we found that the growth n labor costs n some ndustres and occupatons vares from that of all-worker wage growth. Our fndng suggests that the blanket use of all-worker net dscount rates or growth level forecasts s not consstent wh the wage growth found n many ndustres and occupatons. In provdng statonary, constant mean growth dfferentals, we show the addons or subtractons necessary to adjust all-worker net dscount rates or growth level forecasts for use n specfc ndustry and occupaton forensc economc cases. References Autor, Davd H., Lawrence F. Katz, and Melssa S. Kearney Trends n U.S. Wage Inequaly: Revsng the Revsonsts. Revew of Economcs and Statstcs, 90(2): Bowles, Tyler J. and W. Crs Lews A tme-seres Analyss of the Medcal Care Prce Index: Implcatons for Apprasng Economc Losses. Journal of Forensc Economcs, 13(3): Bronars, Stephen G. and Melssa Famular Wage, Tenure, and Wage Growth Varaton whn and across Establshments. Journal of Labor Economcs, 15(2): Juhn, Chnhu, Kevn M. Murphy, and Brooks Perce Wage Inequaly and the Rse n Returns to Skll. Journal of Polcal Economy, 101(3): Pelaez, Rolando F Valuaton of Earnngs Usng Hstorcal Growth-Dscount Rates. Journal of Forensc Economcs, 5(1): Payne, James E., Bradley T. Ewng, and Mchael J. Pette An Inqury nto the Tme Seres Propertes of Net Dscount Rates. Journal of Forensc Economcs, 12(3): Phllps, P.C.B and P. Perron Testng for a Un Root n Tme Seres Regresson. Bometrka, 75: Skoog, Gary R., James E. Cecka, and Kurt V. Krueger The Markov Process Model of Labor Force Actvy: Extended Tables of Central Tendency, Shape, Percentle Ponts, and Bootstrap Standard Errors. Journal of Forensc Economcs, 22(2): Slesnck, Frank L., Mchael R. Luthy, and Mchael L. Brookshre Survey of Forensc Economsts: Ther Methods, Estmates, and Perspectves. Journal of Forensc Economcs, 24(1): In a forensc economc case, some of the employer benef costs mght be mathematcally lnked to the employee wage rate as a constant percentage (e.g., a 3% of wages 401(k) contrbuton). When calculated alone, forecastng constant percentage benefs would be more sued to wage and salary growth as opposed to total benef cost growth.