Full text of Review (Federal Reserve Bank of Dallas) : November 1991

The full text on this page is automatically extracted from the file linked above and may contain errors and inconsistencies.

FEDERAL RESERVE BANK OF DALLAS  \..  ,  , I  .  .  November 1991  COnOllllC eVleW  Income Growth in the Southwest: Implicatz'ons for Long-Term Development Robert W. Gilmer  The Effect of the Growing Service Sector on Wages in Texas Keith R. Phillips  This publication was digitized and made available by the Federal Reserve Bank of Dallas' Historical Library (FedHistory@dal.frb.org)  Economic Review Federal Reserve Bank of Dal/as Robert O. McTeer, Jr. President and Chief Executive Officer  Tony J. Salvaggio First Vice President and Chief OperBling Officer  Harvey Rosenblum Senior Vice President end Director of Research  W. Michael Cox Vice President and AssociBCe Director of Research  Gerald P. O'Oriscoll, Jr. Vice President and Economic Advisor  Stephen P. A. Brown Assistant Vice Prasidant and Senior Economist  Economists Robert T. Clair John V. Duca Kenneth M. Emery Robert W. Gilmer David M. Gould William C. Gruben Joseph H. Haslag John K. Hill Evan F. Koenig D'Ann M. Ozment Keith R. Phillips Fiona D. Sigalla Lori L. Taylor John H. Welch Mark A. Wynne Kevin J. Yeats Mine K. YOcel Research Associates Professor Nathan S. Balke Southern Methodist University Professor John Bryant Rice University Professor Thomas B. Fomby Southern Methodist University Professor Scott Freeman University of Texas Professor John H. Wood Wake Forest University Editors Rhonda Harris Diana W. Palmer Virginia M. Rogers The Economic Review is published by the Federal Reserve Bank of Dallas. The views expressed are those of the authors and do not necessarily reflect the positions of the Federal Reserve Bank of Dallas or the Federal Reserve System. Subscriptions are avai lable free of charge. Please send requests for single-copy and multiplecopy subscriptions. back issues. and address changes to the Public Affairs Department. Federal Reserve Bank of Dallas. Station K. Dallas. Texas 75222, (2141651-6289. Articles may be reprinted on the condition that the soutce is credited and the Research Department is provided with a copy of the publication containing the reprinted material.  Contents Page 1  Income Growth in the Southwest: Implz'catz'ons for Long-Term Development Robert W. Gilmer  Many obs Iv ers vi ew eco nomic d iv rsifi aLi o n ' IS the antidote for th Southwest's boo m-a nd-bu st cycl es and as th e k ey to long-t rm co nom ic growth. Ho bert WI. Gilmer addre s s th is issue by analyz ing w hether regio nal growth has been domin ated by industry-s pecifi c cycli 'II h cto rs o r wheth er int rn al facto rs-such as labor forc qu ality , in.fra tructure, and th edu ca ti o nal system- h'lve primaril y acco unted fo r lo ng-term grow th. Gilm r's analy is suggests that, for lo ng-term growth , in t m al factors ar mor critica l than indu stry mix. Gilm er oncludes th at po li cies design d to promote loca l inv stm ent in human and phys ical ca p ital sho uld b hvo r d over those designed fo r th specific pu rpo e of d iv ersificati o n. Intern al facto rs are largely what prepare the regio n fo r the fu ture.  Page 15  The Effects ofthe Growing Service Sector on Wages in Texas Keith R. Phillips  In Texas du ring (h 1980s, s Ivi ce-s ctor mployment rose, goods-secto r mpl oyment declin d , and th e average r al wage increased only slightly. B ca use se rvi c -se to r jobs pay lower av rage w'lges than goods-s cto r jobs, analysts have suggested that th e growing pro po rti o n o f jobs in the s rvic sector was an impo rtant factor suppress ing ov rail wage gains in th sta t . Keith H. Philli ps finds that th e incr asing share o f service-secto r jobs o nly slightl y cl amp ned wag growth in Texas during th 198005. low w ag growth primarily resulted fro m w ea k w ag expansio n in both the goocls and se rvice sectors. Ph illips also find s that the shift to service-sector jobs had li ttle ffect o n wage in equ ali ty amo ng w orkers in the state. Wl hLl ov rail wage in equali ty increased , the increase r suited almost ntir Iy from a larg increas in wag in quality in the goods secto r and a relatively small increas in wage ineq uality in the s Iv ic secto r.  Robert W. Gilmer Senior Economist and Policy Advisor Federal Reserve Bank of Dallas Houston Branch  Income Growth in the Southwest: Implications for Long-Term Development  W  h n explaining differences in regiona l growtil, economists often rely o n th e variations in industlY mix among regions. From such a perspective, ' I region dominated by o il and gas extra tion will g row when that industry grows natio n' tily and wil l shrink when LiYlt industry shrinks. However, evidence suggests that o ther, interl1'ti hctors may also conrribute to growth differenc s among regions. Furth r, th se in terna l hceors appear to grow in r lative importance as the time frame onside red for analysiS becom es lo nger. S parating extern al factors , such as the natio nal w II-being o f key loca l industri es, from th es interna l f~l ctors all ows us to ask questions about a regio n 's 'lbility to motiva te expansion . Tn th is articl e, I exa mine trends in per ca pita incom growth for thirty-six metro politan ar as in T xas and Louisiana between 1969 and ]988. The 1970s 'lIld ea rly ]980s wer a time of rapid expansion throughout tilis I' gion , and the late 1980s were larg Iy marred by recess ion 'lIld slow growth. This stud y does not focus o n the well-known regiona l cycl ica l events o f this period-oil bust, constru ction downturn, and banking crisis-hut rather on w hat these data tell us about the rundamental economic hea ltil o f th m tropolitan South west and its prospects for long-term econom ic development. I n rh eir classic book on the determin ants of regiona l growth , Borts and Stein 0964.) obselved manuf:lcturing employment by state and divided its growth rate into l:\VO pailS. h 'st, part o f the Clclual growt'h rate W'IS ' ltt ributed [() differences in industrial com position fro m staLe to st'lte, as some states sta rt w ith a prepond rance of rapidl y growing natio nal industries. Second, the autho rs defined the inlernCiI growth nte as the difference between the actu' ti growth rate and the part 'ILrributable to industry mix . This intern al growth rate consists of Economic Review -  November 1991  all the loca l factors important and specific to each stat. Bo tts and Stein 's surprising conclusion was that actual lo ng-term growth, as measured from l ea k Lo pea k in the busin ss ycl e, is dominated by loca l factors and the interna l growth rate. The primary 1'01 for industty mix is as a cyclica l, rather than lo ng-term , in11uenc on development. This conclusion rais s interesting qu estions fo r a region such as the Southwestern Unit d States, where twe nty y aI's of boo m-'lnd-bust cy les are ofren attributed to d pend nce on oil , lumber, paper, h nning, and other prim'lry commoditi s. The r g ion 's econ om ic future is ofte n describ d in renns of a post-petrol um industry mix; discussio ns of eco no mi c po li cy focus on diversification of the I' gio nal industrial base. Could I' g io nal growth in Texas and Lo uisiana ove r the past twenty years actua lly have been dominated by loca l infernCiI hctors and not by cye/icCiI facto rs, such as o il and farm pri ces? To answer this questio n, T exa mined per apita incolll growth in metropo litan areas of Texas and Louishna. Per cap ita incomc is the best and most comprehensive measure of regional co nom ic hea lth ava il able to r sea rchers. U nlike Borts and Stein , w ho focused strictly o n employment in manufacturing, I include all sectorsfarming, co nstru ction , mining, various private serv ices, and governm ent. Further, while Borts ' md St in focused their discuss io n of industry mix effects o n high- versus 10w-g rowl l.? industri es, we w ill be able to deal w ith changes in the mix of high- versus low-wage industri es as well . T bcg in this articl e by describing growth patterns o f the 1970s and 19805 and many or th e resulting eco nom ic dislocations in Texas and Louisiatl'l. Next] expl o re trends in per ca pita income growth in the regio n, plac ing emphasiS o n 1  long-term growth measured from pea k to pea k in the bu siness cycle. My analysis highlights metropolita n growth by state, population size, and dependence on o il exploration. Finall y, I dif~ rentiate the ro le of indu stJy mix versus intern al factors in the growth o f Texas and Lou isiana industlY. My concl usio ns broadly concur with those o f Borts and Stein . Intern al factors dominate the actual growth rate from pea k to p ak in the busines cycle. 'I he resul ts also suggest that, I as d onl y o n inte rn al facto rs ind pendent o f d1 ef~ cts of industry mix, thes Texas and Lo uisiana cities would have been hard pressed through out this period to k eep up w idl the rest of dl metJ'opolitan Unit d States. In thi s article, T do no t address the mo re di ffi cul t q uesti o n o f exactl y w hat determin es lhe in tern al g ro wth rate. Th methodol ogy I empl oy is ' 1 va ri anl o f th we ll -kn ow n sb·(ft- sbare tecbnique. ' As in shift-s hare analysis, the I ca l o r in te rn al ffecr is simpl y a res idu al repr senting th distinctiv influ nc s that differenti ate th e regio n from lh nation. Factors such as the area's du catio nal syste m , labor skills, infrastJ'uctur , transpo rtatio n sysl em , env ironm ent, and qualilY of life spring to mind . Certainly, the impo rtance o f loca l factors, as highlig hled by dlis slud y, suggest less emphas is o n po licies gea red towa rd ind ustrial recruitm ent fo r purposes o f di ve rsifica tio n o r on the s arch fo r specifi c industries th at bring high va lu -'Idded o r high te hno logy. Instea d , my analys is suggests th e " post-petro leum" Southw st sho uld evo lve fro m ca refLtI attentio n to how w II we ca rry o ut in vestment in human 'lI1d phys ica l ca pi ta l al lhe loca l I v l.  Structural Change and the Southwest Economy Figure 1, a pl o t o f non agricultural wage and s't! ary employment in Texas, shows th e path of mining, co nstru cti o n , and tot,t! empl o yment from 1969 to 1990. All valu es have b n ind xed to 1969 = 1 to give a better perspective on th rate o f growth o f the vari ous compo nents. Miningcomp sed primaril y o f the o il and gas industry in  r 2  See Isard (1960) for a description; see Jackson et al. ( 1981) for  a typical application.  Figure 1  Nonagricultural Employment in Texas, 1969- 90 Percent 3 :.· ····... Minlng  / \..  2.5  2  Total  1.5  •  ~  ~  ~  n  ~  ~  ~  U  ~  u  1969 = 1 for all variables  th se states-a nd a massi ve constru ction boom starled the o uthwe t's cyclica l roller coaster rid e, a rid e that continued fo r much of lhis peri od. 1 he count o f working o il and natural g'ls ri gs in lhe U nited ta tes soa red to 3,970 during 1981, o nl y to fall bac k to an ave rage o f 937 workin g ri gs during 1988. N w r sidenti 'll building permits in Texas numbered slightly mo re tban 250,000 in 1983 but o nly 40,500 by 1988. Texas-based banks, which had fin anced a sig nifi ca nt p art of bOlh boo m and I ust, sustain d a declin in equity ca pital fro m $13.6 billi on in 1985 to $7.8 billi on in 1988. In term s of total employment, the downturn in Texas was relati vel y bri ef, and recovery to th e prio r p ea k ca m quickl y . Further expan io n has b n quite slow, howev er. Th e Lo uisiana data indica le a different story, as the state has yet to I' turn to prio r levels of employment (Fig u re 2) . Lo uisi'lI1a's depend nce on o il and g'IS, mu ch great I' even than Texas, exaggerated the heights and depths of ev ents there (Gruben and Hay s 1991). Tn lh 'Iggrega te fi gur s fo r bo th states, however, Texas do mjnate the statistics. For th e m tro po litan data us d in this stud y, fo r exa mple, Texas in 1988 had 84 perce nt of total arnin gs and 82 p erce nt o f the popul ation. These v nts brought long r-tenn chang Texas and Louisian'l as w ell. To III asure this stru t't.lral Fede.·a) Resel've Bank of Dallas  Figure 2  Nonagricultural Employment in Louisiana, 1969- 90 Percent 2.5  .......... ~:.~ing  2  ..... ..... 1.5  '69  '71  '73  '75  '77  '79  '81  '83  '85  '87  '89  1969 = 1 for all variables  chang , T used the foll owing formul a, employed by Lawrence (1984) and Gilmer and Pulsipher (1989), to measure structur'ti change in manufacLu ring. H er , how vel', T applied the formul a to a comprehensive list of in lustly s ctors that tog th I' make up tota l mployment and income: 11/  L, ISIj,' SIj = J=1  slj.l+ll x100, i = l ,,,. , fJ ;j = l ,,,.m  2n  5 /, 1+ 1 /  n  tru ctural change fo r pl ace i and for the p riod beginning at time I, hal' of industry j, place i, and tim I, hare of industry j , p lace i, and tim 1+1 , and umber of yea rs in the pe ri od fro m I to 1+ l.  T measured th shares of ea h industlY as a part of the I' gional e ono my in an unusual way, bUL in a way that provided some consisLency widl later discussions of changes in indusLry mix. h'II'es were computed from bYPolbelical income, in w hich the hypothetica l income for each indusuy is its reg ional mployment multiplied by the matching /1ational wage o r arnings rate. Total hypothetica l Economic Review -  Novembel' 1991  income is the sum across all sectors. As 'I practi al matter, reg i0 11'1 I shifts are mostly dominated by changing loca l employm nt patterns. T computed d1is index for thr tim p riod 1969-79, 1979-85, and 1985-88. Tabl 1 shows the va lu e of dle index fo r dl dljrLy-six Texas and Louisiana m tropoJitan areas used in thjs articl . The to p portion of dle table shows the total ind x va lu fo r the thilty-six m tropoliran ar a and the tota l for each st'lte. T defined five differ nt population groups an I computed th e index fo r each: more than 1 millon ( fiv 'ities), 500,000-750,000 (thr cities), 250,000-500,000 (fiv ities), 150,000250,000 ( I ven cities), less tl1'ln 150,000 (twelve cities). T us d this index to define o il-d pendent metropolitan areas. Citi s in w hich 20 percent or more of the regi0 11'1 I tructural change from 1985 to 1988 o riginated in dl mining sector weI' assumed t:o have significa nt Jink'lges to o il exploration. This distinctio n divided the numb I' of m tropo litan ar as venly: eighte 11 weI' o il-dependent and eightee n w re not. On av rage, 32.4 percenL of dle structural cl1"lnge during dlis period riginated in dle mining sector for the d p nd nt group; for the lessd pendent group, the figure was onl y 10.1 percent. Th third section of Table 1 lists Lh ind x of structural change for evelY metro poli ta n area during the thr p ri ods. The index shows a clea r tend ncy to incr ase across tim ,but index valu s also dem nstl"l te considerabl diversity in each peri od. In 1985-88, for exa mple, the index fo r our xplorationd pendent cities averag d 186. Midland , Odess'I, and L-t fayette bro ke 300; mea nwhile, citi s such as EI Paso and Baton Roug barely broke 100. Tabl 2 shows th industry sectors dlat contr'ibuted to structural change for the combin d index values for the dlirty-six metropolitan areas. For ]969-79, the index was relatively small , and ne'lrl y half the m asure originated in "other" ca tegori s-prim arily gov rnment 'Ind th growth of proprietorships. The contributio n of mining W'IS ] 3.2 percent, hrge compared w ith the sector's 2Ap rcent sll'lre of ov rail empl oyment. As would be exp cted, in 1985-88 conslruction and mining both play d a big ro le, especially I' lativ to their size. Although private services, w hen taken together, we re a Significant contributor to structu ral change after ]979, these s Iv ice indusLries also do minated regional mployment. Strong servic 3  Table 1  Index of Structural Change Texas and Louisiana Metropolitan Areas, 1969-88 1969- 79  1979-85  1985-88  Total Metropolitan Metropolitan Texas Metropolitan Louisiana  63.3 67.5 61 .6  109.5 106.7 130.1  164.3 168.4 148.1  Population: Group 1 Group 2 Group 3 Group 4 Group 5  64.5 53.9 59 .7 81.1 82.1  107.6 89.7 131 .4 132.5 112.8  166.5 114.1 174.3 166.0 200.1  Oil Dependent Less Dependent  66.4 63.8  133.4 93.7  186.0 147.2  103.1 64.5 75.3 76.7 64.5 59.0 151.4 96.8 63.8 95.9 56.2 88.0 82.8 60.9 93.6 68.8 105.7 148.7 71 .1 156.4 85.9 82.3 110.8 83.8 74.3 51 .5 87.5 128.2 96.4 90.9 76.6 163.0 74.1 122.9 56.2 167.3  114.2 94.5 114.8 127.8 116.2 153.9 235.1 146.7 129.3 130.1 101 .8 73 .0 97.7 161 .5 116.8 144.9 97.0 131.1 200.2 145.3 130.4 159.5 147.4 113.4 128.4 136.9 167.1 106.6 98.6 96 .5 120.5 130.5 141.4 86.0 96.9 110.7  248.8 141.6 180.4 162.5 108.4 167.2 116.5 161.3 172.1 224.5 172.3 105.1 121 .2 175.0 265.0 193.3 158.0 333.4 197.4 173.8 156.6 140.4 116.4 314.5 149.0 149.2 336.7 213.2 180.7 178.8 190.4 152.6 235.5 293.0 163.7 268.3  Abilene Alexandria Amarillo Austin Baton Rouge Beaumont- Port Authur Brazoria Brownsville-Harlingen Bryan- College Station Corpus Christi Dallas EI Paso Fort Worth Galveston- Texas City Houma- Thibodaux Houston Killeen- Temple Lafayette Lake Charles Laredo Longview- Marshall Lubbock McAllen- Edinburgh Midland Monroe New Orleans Odessa San Ange lo San Antonio Sherman- Denison Shreveport Texarkana Tyler Victoria Waco Wichita Falls  4  Federal Reserve Bank of Dallas  Table 2 Structural Change in Thirty-Six Metropolitan Areas: Contribution of Various Sectors, 1969-88 1969- 79  1979- 85  1985- 88  Index Value  63.3  109.5  164.3  Percent Contribution of Construction Mining Manufacturing Private Services Farming Other  9.0 13.2 10.9 14.2 5.0 47.7  11 .9 7.4 17.3 43.4 3.0 17.0  13.3 22.4 5.5 40.3 1.4 17.2  sector growth seemed to ameliorate overall struct1Jral change. For example, I computed correlations across the thirty-six metropolitan area between the percentage of change emanating from services and the total index. I found signjficant and negative oefficients, indicating that when selv ic s were a more impoltant component of structural change, the overall index was typically smaller and dle local economy more table. On dle other hand , a large part of structural change stemming from manufact"llring or mining was associated strongly and positively with large overa ll index values. Personal Income Growth In this section, I look closely at the growth of real per capita persona l incom in metropolitan Texas and Louisiana. Table 3 summari zes longterm trends in per onal in come, population, per capita income, and the positio n of the region relative to the nation. In 1969, m tropolitan per ca pita income in Texas and Louisia na was 85 percent of that found in the typica l U.S. metropolitan area, and by 1979 this figure had risen to 94 percent. Metropolitan per capita income fell to 92 percent by 1985 and by 1988 had lost all the ground previously regained, falling back to only 84 percent of the U.S. metropolitan level. Growth rates for total income, population, and per capita income, shown in the lower half of Table 3, indicate dlat from 1985 to 1988 the average of per Economic Review - November 1991  capita income in Texas and Louisiana grew at an annual rate a remarkable -3.1 percent below dle nation's. This growdl deficit occurred not only because of regional decline but also because dle decl ine coincided with a period of rapid national ex:pansion. 2 To determine the ca uses of these flu ctuatio ns in personal income per ca pita, it is user"lll to co nside r changes in the components of this variable. Personal income consists of ea rnIngs, property income, and transfer payments. Ea1'l'lings are wages, salaries, proprietor's income, and othe r labor income, such as empl oyer contributions to private pen ion funds. Property income consists of dividends, rent, and interest. Transfers are payments to individuals for which no current service are rendered. Social Security, unemployment compensation, railroad and mil italY retirement, and Medicare payments are examples of transfers. Each component can be divid ed by total current population to put it on a per capita basis. We have eliminated general inflationary trends and  2  The general approach to the descriptive material regarding personal per capita income is from Garnick and Friedenberg (1982) and Garnick (1990). The mechanics of making these calculations is detailed in the same papers. I based my industry mix calculations partly on these sources and partly on more standard shift- share calculations, such as those used by Borts and Stein.  5  Table 3  Long-term Trends in Real Per Capita Income: Thirty-Six Metropolitan Areas in Texas and Louisiana  Personal Income (billions) Population (millions) Per Capita Income : Regional Metropolitan U.S. Metropolitan Per Capita Ratio : Regional to U.S. Metropolitan  1969  1979  1985  1988  78.6 11 .0  133.7 13.9  159.4 16.3  155.1 16.8  8417 9954  11492 12193  12222 13295  11845 14154  .845  .942  .919  .837  1969-79 Personal Income Population Per Capita Income: Regional Metropolitan U.S. Metropolitan Difference: Regional Minus U.S . Metropolitan  con vert d ac h compo nent to co nst'lJ1t 1982 do llars by using the per. o nal consumptio n expenditure d Aato r fro m the Na tio nal Income and Product Acco unts. Earnings per ca pita is further bro ken down by using this simpl identity: Ea rnin gs  Ea rnin gs  Po pul atio n  Employment  ---=-- =  Employ ment  x - --'-----'---  Po pu latio n '  The first term is th earnings rate o r, mo re simply, the ave rage annu al wage o r s'li ary per empl oyed work r. Th e s cond term is the employment to p opuJafiol1 ratio. This measure has in creas d steadily since 1969 because of lo ng-t rm trends, such as th increase in the numb r of w o rkin g wo men and th baby boom generatio n's ntry into the labo r fo rce. This ratio also ri ses and falls cyclic-lil y w ith the avaihbility of jobs. Tabl 4 summ ari zes this list of defini tio ns and shows som alcul atio ns fo r 1969-79. The first co lumn sho ws the growth rate fo r income per ca pita fo r vari.ous m tro po litan r g io ns and subregio ns. Th e oth er columns sho w th co ntribu6  1979- 85  1985- 88  5.46 2.35  3.75 2.72  -. 17 .87  3.11 2.05  1.03 1.42  - 1.04 2.09  1.07  -.39  - 3.13  tors to th is towl , stated as perce ntag -po int co ntributi o ns. Th figure g iven fo r ea rnin gs, pro perty income, and transf rs p r ca pita all 'Idd up to th fi gure fo r total in come per capita. Ea rnin gs ar bro k n in to the contributio n of th arnin gs rat and the emplo ym nt- population ratio. Tables 5 and 6 show simil ar resul ts fo r the 1979-85 and 1985-88 peri ods. Ea rnings is the ca tegory comribuLing th hrg st share of p rsonal in com and r ce iving the largest w ight in th e ca lcuhtio ns. Tn 1969, ea rnin gs w r abo ut 80 perce nt o r total p rsonal inco me, 'Ind property in com e and transf rs w ere ro ughly 10 p rce nt each . By 1988, regiona l 'Irnings had shrunl to 68 perce nt, w ith the oth er conl'ribu to rs at 'Ibout 16 perce nt ach. Fro m 1969 to 1979, ve ry thing seemed to be workin g w ell fo r metro politan areas in T xas and Loui siana . Ea rnin gs, in particul ar, showed strong g rowth o ri ginating from both the ea rnings rate and job growth relative LO po pulati on. From 1979 to 1985, rh e co ntributi on of ea rnin gs f II , but th other factor ros . The oth er fa cto rs we re not sufficient, how ever, to compensate fo r th e sIowFedel'al Reserve Bank of Dallas  Table 4  Contributors to the Growth of Real Per Capita Income, 1969-79 (Percent per year) Percentage-Point Contribution : Earnings per Capita  Total  Property Income per Capita  Transfers per Capita  Income per Capita  Earnings per Worker  Jobs per Capita  3.1  1.4  1.5  0  .3  0 0  .3 .3  Texas Louisiana  3.1 3.1  1.4 1.3  1.5 1.5  Population : Group 1 Group 2 Group 3 Group 4 Group 5  3.1 2.6 3.0 3.6 3.2  1.4 .9 1.2 1.6 1.6  1.5 1.2 1.3 1.8 1.1  -. 1 .1 .1 0 .2  .2 .4 .4 .2 .4  Oil Dependent Less Dependent  3.6 2.7  1.5 1.2  1.9 1.1  -. 1 .1  .2 .3  NOTE: Totals may differ from the sum of the parts because of rounding errors.  dow n in growth o f jobs per ca pita and th eros ion o f arnin gs p er worker G IU S d by inflation. Overall p r ca p ita income growth slowed sharply. Du ring 1985-88, the results for I' al ea rnin gs pushed p er ca pita in com e growth to the nega tive sid . Property incom and transfers remained pos itive but now were mo re than offs t by results for ea rnings .  versus low wages operates through th ea rnin gs rate. D fine hypotheti ca l ea rnings (E) as th b s ctoral employment in th regio n tim es the U.S. ea rnin gs rate in ach sector and summed across all sectors. G'lrnick and Fri edenberg then divide ea rnin gs per worker (EI]) in to the fo llow ill g identity:  E/1  =  (E/]) (EIE,).  Industry Mix Versus Internal Factors The influ nc of industry mix wo rks through ea rnin gs. Ga rni ck and Fri edenberg ass um e that the important factor is the ea rnings rat , and they ask w heth r changes in the mix of indu stry hvor high- versus low-wag industry. Borts and Stein, o n the other hand, loo k at: whether changes in industry mix favor high- o r low-growth industlY· Both perspectives are im po rtant, of course, and both ca n easily be incorpo rated in the fram work I have pI' sented. Garnick and Fri ed nberg's conc rn al out high Economic Review -  November 1991  The first part of th is expr ssio n represents hypoth tica l ea rnin gs per worker, and it measures the influence o f industry mix on incom as work rs move in to o r o ut of high- or low-wag employm nt. The second expression is sse ntialiy a I' sidua l, measuring other loca l o r intern al influences on ea rnings rates. A ltern atively, concern about mix focus s on high-growth ve rsus low-growth industry . In this case, defin e hypothetica l employm nt C./,) as the empl oyme nt that would have been achieved in th e I' gio n if ach industry grew at nati onal rate . 7  Table 5 Contributors to the Growth of Real Per Capita Income, 1979-85 (Percent per year) Perce ntage-Point Contribution : Earnings per Capita  Property Income per Capita  Transfers per Capita  Income per Capita  Earnings per Worker  Jobs per Capita  Total  1.0  - .3  .4  .7  .2  Texas Louisiana  1.1 .6  -.2 - 1.0  .5  -. 1  .7 1.1  .5  Population : Group 1 Group 2 Group 3 Group 4 Group 5  1.0 1.8 -. 1 .8 .9  - .2 - .1 - 1.1 - 2.0 - .7  1.2 -.8 -.2 .3  .6 .6 1.3 1.0 1.0  .1 0 .6 .4 .2  Oil Dependent Less Dependent  .1 1.9  -.7 .1  - .4 1.1  .9 .6  .3 0  .5  .1  NOTE: Totals may differ from the sum of the parts because of rounding errors.  Then di vide the mployme nt-po pu lation ratio as fo llows:  l i P = C.I,/ P) (PI,), Th first part i hypothetica l emp loyment as a m as ur of industry mix; the second te rm is aga in a I' s idual, measuring other loca l influences o n empl oyment growth. l a bles 7 and 8 show the result of these ca l ul atio ns. Ta ble 7 divid s the growth in the ea rnings rate into an industry mix and an inte rnal compo ne nt; Ta ble 8 does the sa m for gro wth in the employme nt- popu lation ratio. In Tabl 7, actual growth rates match the colu mn fo r ea rn ings per worke r in Tables 4, 5, and 6. Her , we have simply tak n o n mo re ste p and divided th e g rowth rate into industry m ix and in ternal ffects. The industry mix component is hea lt hy and positiv fro m 1969 to 1979, contributing abo ut 1 percent to rea l I e r ca pita income growth in all 8  metro politan area. During 1979-85, th is growth rate turned sJig htl y nega tive, the n reversed and beca m positive after 1985 . Thi positive turn from 1985-88 was surprising because on of the p rim aly concerns 'lbout th o il recessio n was that highpaying jobs in til oil fi elds would be r placed by poorly paid jobs in elv ices. Selv ices hav , in fact, been the focus of much of the these states' job growth in tile 1980s. Eve n so, the verall contribution o f illdustly mix to wage levels remains positive. Ta ble 8 makes a similar divis ion, but fo r high-growth v rsus low-growth industri s. O nce mo l' , the actu al growth rates in thjs table match colu mns presented in Ta bles 4, 5, and 6; in this case, howeve r, th variable of interest is jo l s per ca pita. Tndustty mix, measured in thjs way, played a ve ry small ro le befo re 1985, favo ring th positive side of z ro during 1969-79 and the negative side during 1979-85 . Values turned strongly positive afte r 1985, but not e no ugh to compe nsa t fo r the inte rnal facto rs that turned strongly nega tiv at th ame tim e. Federal Reserve Bank of DaUas  Table 6  Contributors to the Growth of Real Per Capita Income, 1985-88 (Percent per year) Percentage-Point Contribution : Earnings per Capita Income per Capita  Earnings per Worker  Jobs per Capita  Property Income per Capita  Transfers per Capita  - 1.0  - 1.0  -.9  .4  .5  Texas Louisiana  - 1.1 -.8  -.9 -1 .6  - 1.1 - .4  .3  .5 .7  Population : Group 1 Group 2 Group 3 Group 4 Group 5  - 1.2 - 1.0 - 1.6 - .7 - 1.1  - .8 - 1.0 - 1.9 - 1.2 - 1.7  - 1.1 -.7 - 1.0 - .6 -.8  .4 .2 .4 .6  .4 .5 .8 .7 .8  Oil Dependent Less Dependent  - 1.5 - .7  - 1.7 -.4  - 1.0 - .9  .6 .2  .6 .4  Total  .5  .4  NOTE : Totals may differ from the sum of the parts because of rounding errors.  Wh y did intern al factors p rfo rm so poorl y after 1985? Tabl es 7 and 8 show nega tive contributio ns of - 1.8 percent per yea r fro m ea rnings rates and -2.4 perce nt fro m th employmentpo pulatio n rati o. Th pl"i'm en]! reason for this poor showing is that th regio nal slowdown ca me just as a lo ng nati o nal exp ansio n b ga n to accelerate, Ev nts in T xas and Lo uisiana, driven by oil prices and a r gional constru ctio n cycle, wer badly out of synchroni za tio n w ith the U.S, business cycl . To th ext nt th "inL m al" factors ar res iduals, they suffer much wors fro m compari son with the U ni ted States in this p eri od than in the ther, ea rli r peri ods,  Income Growth and Structural Change Th dates chosen as end points in this analys is are ith r p ak y ars in the national or r g io nal busin ss cycl o r yea rs of continued expansio n, They were selected to minimi ze cyclica l ffects and to highlight th long-term Economic Review -  November 1991  structural hange ta king pJ ac in the o uthwest. This emphaSiS i s particul arly impo rtanL in the late 1980s, fo r xa mpl , as restru cturing was xtensive, Th e reg io n did not have the lux ury of an cono mic recov ry thaL proceed d quickly by resto ring wo rkers to their o ld pos itio ns; Lhe number of working ri gs, banks, and build ing sites was perm an ntl y reduc d. Beyond the dislocatio ns o f human and physica l ca pital, as Ga rnick pointed out in his discussi o n of th p rfo rman ce o f the outhwestern United StaLes in dl 1980s, there w r odler r 'Isons the reg io n's p rfo rmance was poor com par d w ith the rest of th nation, For exa mple, d1 ga p in compensa tio n of technica l and managerial work rs and dle r st of dle labo r force grew w ider in dl 1980s, This div rgence hurt the Southwest in comparison w ith New England and the M ideast r gio n, w hich suppo rt l arger (tho ugh declining) numl , rs o f technica l and manage rial work rs. Furth r, much of th x pansion o f the 19805 was I d by fin ancial activity, sp cia ll y nonbank 9  Table 7  Division of the Effect of Industry Mix and Internal Factors on Growth of Earnings per Worker (Percent per year)  Actual  1969- 79 Industry Mix  Internal  Actual  1979- 85 Industry Mix  Internal  Actual  1985- 88 Industry Mix  Internal  Total  1.4  .9  .5  - .3  -.2  -. 1  - 1.0  .9  - 1.8  Texas Louisiana  1.4 1.3  .9 .8  .5 .5  -.2 -1.0  -.2 - .3  0 - .6  - .9 - 1.6  .9 .8  - 1.7 - 2.3  Population : Group 1 Group 2 Group 3 Group 4 Group 5  1.4 .9 1.2 1.6 1.6  .8 .9 .8 1.0 1.1  .5 .1 .3 .6 .5  -.2 -. 1 - 1.1 - 2.0 -.7  - .3 .1 -.2 - 1.7 .1  .1 - .2 -. 9 -.3 -.8  - .8 -1 .0 - 1.9 - 1.2 - 1.7  .9 1.3 .7 .5 .3  - 1.7 - 2.4 - 2.5 - 1.8 - 2.0  Oil Dependent Less Dependent  1.5 1.2  .9 .8  .7 .4  -.7 .1  - .2 - .1  -.5 .3  - 1.7 - .4  .5 1.1  - 2.2 - 1.5  NOTE: Totals may differ from the sum of the parts because of rounding errors.  financing. Texas, in particular, was struggling to sav e its banking industry and w as in no pos ition to benefit from a peri od o f fimncial inno vatio n. Th e compensa tio n an I bo nu s s fo r junk-bo nd p'lckage rs, arbitrage rs, and related fin ancial pro fessionals fl owed elsewhere. Tab l 9 shows correlatio n coefficients compul d across th e thirty-six metropolitan areas and relat s stru ctural change coeffici nts w ith ea rnin gs growlh resulting fro m eith r industlY mix o r intern al facto rs. This correlatio n simp ly inclica tes how metro po li tan economies felt structural change. COlT latio ns ar coml ut d se parately fo r the ea rnings rate and the mployment- popu latio n ratio . The coe fficients indica te that stru ctural change affect d ea rrrings rat positiv Iy fro m 1969 to 1979 and nega tively from 1985 to 1988. Altho ugh (as seen in Table 7) the industry mix compo n nt made a pos itive contributio n to ea rnin gs rat s fro m 1985 to 1988, this contribution was consistentl y diminished by stru ctural change. Stru ctural change improved th e emplo yment mix in favo r o f hig h-growth indu stri es fro m 10  1985 to 1988. Tabl e 8 ind ica tes positive ga ins fo r the regio n, and the p ositive co rrelatio n indica tes th at th biggest favo rable ga ins ca me wh r strLl clLJral change wa s m o t x tensive. On th Odl r hand , th e in t rn al abili ty o f th e m tro politan areas to g n r ate n w jo bs was hurt badly during this peri od ; the co rrelati on coe ffi cient o f - 0.725 indi ca tes that th e mor stru ctural change OCCUlT d , th e mo re bad ly dam aged w er th e intern al facto rs. Another way to judge the effects of structu ral chang is to comlYlre dle outcom S fo r ea rnings growth for oil dependent and less dependent m tropoJitan <I I' as in Tables 7 and 8. The o ildepend nt cities underw nt far more structural adjustment, and the result is som w hat more stable behav io r by less-dependent cities a we mov from o ne peri od to dl ne){t. Th g neral paltern o f slowdown and decline in rea l ea rnings during dle 1980s is still shar d by bodl groups, however. As w e look at the ro le of intern al factors o nl y in Tab les 7 and 8, it is difficult to point to a subperi od fro l11 1969 to 1988 tbat might be Federal Rese lve Bank of Dallas  Table 8  Division of the Effect of Industry Mix and Internal Factors on Growth of the Number of Jobs per Capita (Percent per year)  Actual  1969- 79 Industry Mix  Internal  Actual  1979- 85 Industry Mix  Internal  Actual  1985- 88 Industry Mix  Internal  Total  1.5  .1  1.4  .4  -.2  .6  -.9  1.5  - 2.4  Texas Louisiana  1.5 1.5  -. 1 .8  1.6 .7  .5 - .1  -.5 1.0  1.0 - 1.2  - 1.1 0  1.2 2.9  - 2.2 - 3.0  Population: Group 1 Group 2 Group 3 Group 4 Group 5  1.5 1.2 1.3 1.8 1.1  0 -. 7 .5 .4 .8  1.6 1.9 .8 1.4 .3  .5 1.2 -.8 - .2 .3  -.5 - .7 .3 .7 .2  1.0 2.0 - 1.1 - .9 .1  - 1.1 - .7 - 1.0 - .6 -.8  1.2 .9 1.9 2.5 2.6  - 2.3 - 1.5 - 2.9 - 3.0 - 3.4  Oil Dependent Less Dependent  1.9 1.1  -. 1 .3  2.0 .8  - .4 1.1  - .2 .2  -.2 .9  - 1.0 -. 9  2.1 .9  - 3.1 - 1.8  NOTE: Totals may dlHer from the sum of the parts because of rounding errors.  described as "busin ss as usua I. " When business was good, it was velY good. When busin ss was bad, it was v lY bad inde d. But, in my judgment at I ast, it s ems tila[ such an in-belween period is likely to find ti1ese Souti1west rn cities strugglingo n the basis of intern al facto rs alone-to keep up with th rest of the natio n. The nex[ section underscor s the importance of this prol lem.  Internal Factors and Actual Earnings Growth Finally, over the long run , if we exam in indu stry mix versus inc m al ffe ts , Which is mol' closely relat d to actu al in om growth? Borts and t in exa m ined employme nt growth in manufacturing across va ri o us stat s. Consider their conclu ion .3 Maturity and decline are long-run phenomena and are not produclS of tJ1e business cycle. These phenomena are not produced to any significa nt degree by states having heavy concenu'a tions o f decl ining industries. Maturity and decline arise (as a rul e)  Econom.ic Review - Novembel' 1991  because the state's industries have grown at lower rates ( in fact, nega tive rates) than their n'ltiona l counteq "IrtS. Our explanations o f interstate differences in growth rates must x pl ain interst.ate differences in il/lernal rates. A theory of growth may usuall y disrega rd ... th e composition of a Sla te's industries .... [Borts and Stein's italic .]  My conclusion was the sa m ,ev n though J expand d significa nLly o n th scope of til ir analys is in s v ral ways, I moved outside tile manufacturing industri es to encompass all industJy secto rs and x pand d tile industry mix conce pt to include not jusL high- versus low-growth industri.es, but high- versus low-wage industry as well. Table 10 shows the k y r suits. Tabl e 10 displ ays th correlation ac ross meLropolitan areas of actu 't! growth in arnin gs rat s w ith the mix  r:  See Borts and Stein, p. 47.  11  Table 9  Correlation of Structural Change with Real Regional Earnings: Thirty-Six Metropolitan Areas, 1969-88 Earnings per Worker  1969- 79 1979- 85 1985- 88  Mix  Internal  Mix  Internal  .477 - .279 -.573  -. 118 .043 - .134  - .015 .011 .496  .325 -.403 -. 725  and intern al compo n nts, as well as a tllal growth in the mployment-po pulati o n rati o w ith its mix and inte rn rtl compo nents. In both cases, indu stry mix ass um d a secondalY ro le. Inte rn al factors were consistently and strongly I' lared to actu al rates f observed growth in ea rnings. This res ult was tru e fo r all tim e periods and both measures of industry mix. T here is a po irive correh tion fo r industry m ix as it relate to the ea rnjn gs rate for both th e 1979-85 rind 1985-88 peri o d. This findin g may I' fl ct the extrao rdin ar y m ix change and structural shocks under way in the regio n in the 1980s and the co ntinu d inability of th o il , ba nking, and co nstru cti o n indu stries to rev ive to levels close to pri o r pea ks. Howeve r, thi s findin g does no t d imini sh the si ze and co nsistency o f intern al facto rs as a determin ant of th e actu al in come growth . In hct, this posi tive corr latio n is indi cative of the critica l nature of th s in te rn al fac to rs and their impor ta nc ven in times of w idesp rea I change.  Sununary and Co nclusions For the for caster o r business-cycl analyst, in lustry mix is a criti ca l compo nent o f growth (Fom by and H irschberg, 1989). For the econo mi c d ve lo pment specialist concern ed w ith paltern s of lo ng-term growth and regio nal development, the conc pt is a fa r less mea njngful. Over the lo nger term , int m al factors sp cific to th e loca li ty increasingly make the difference between growth and d cl in . 12  Jobs per Capita  Imp rove ments in industlY mix or shi fts to high-wag , hig h-growth , o r high-va lu -added industries are not mea ningless, but th eir ro le is cyclica l. uch improvem ents may smooth the I' gio nal busin ss cycle, mak th e reg io n less susceptibl e to exogenous shocks, or improve the reg io n's occupatio nal mix . There is liltl evid enc, how ve l', that the mix w ill have a grea t impact o n the I' g ion 's lo ng-term growth of in come o r empl oy m nt. Data from th 1980s provided som vid nce that, amo ng the thirty-six J11 tro politan rlreaS in Texas and Lo uisiana, industry mix was positiv Iy co rr lated w ith actu al growth o f rea l ea rnings per worker. H ow v 1', this I' suit s med to b mor o f a co nce ion to th e massive stru ctural change und I' way throughou t the regio n than a signal of the impor ta nce of indu stry mix. Mix was not impo rtant in affecting actual outco m s fo r job growth in any period, nor did mix affect ea rnin gs rates before 1979. Intern al hctors, o n the other hand , were consist ntl y and stro ngly CO LT lated w ith actu al growth o f both arnings rates and empl oyment, and this w as tru e fo r all tim p eri ods. Ev n when indu stry mix prov ed to be Significa nt, it detracted nothing fro m th pow I' of th ese inlern al factor. A strong element of I' g ional luck n cessa rily enters into th growth of a city, state, o r region . T he w hims of the o il market, changing consumer taste, the outcome of defense spending, and t chno logica l chang w ill affect th growth o utcom . Texas and Lo uisiana weI' fo rtuito usly positi o ned by nature to tak advantage of the o.il Fede ra l ReseJve Bank of Dallas  Table 10  Correlation of Actual Earnings Growth with Mix and Internal Effects: Thirty-Six Metropolitan Areas, 1969-88 Earnings per Worker Mix  1969- 79 1979- 85 1985- 88  .115 .388 .705  Jobs per Capita  Internal  Mix  .885 .891 .834  .297 .058 .028  Internal  .759 .758 .740  boom of the 1970. , much as New England and Ca liforni a weI' spec ial beneficiaries of the defense bu ild-up and high-tech growth of the 1980s. However, as we m ove to the long run and the regiona l w heel of fortune inev itabl y turns, mo re fu ndamental factors take control- the quality of the labor forc , du cational system, infrastructure, transportation system, environment, and so forth . Those regions that ar prepared to cope flexibly with a w ide range of possible futures w ill inev itably fare best in a changing world.  Economic Review - November 1991  13  References tein (964), Economic Growth in a Free Market (New  Borts, George H ., and] ro me L.  Expansio n in th T nnesse V'tll ey," Growtb  and Cbcmge 20:62-70.  Yo rk : Co lumbia U ni versity Press). Fomby , Tho mas B. , and l os ph G. Hirschberg (989), "Texas in Transitio n: D p ndence on O il and the Na ti o nal Econo my, " F d ral Rese rv Bank of Dall a Econ omic Review, January, 11-27. Ga rnick , Daniel H . (990), "Accounting fo r R gio nal Diff r nces in Per Ca pita Tncom Growth: An U pdat and Extension ," Sumey q/ Current  Business 70:29-40. Ga rnick , D 'mi I H ., and H owa rd 1. Fried nberg (1982), "Accounting ro r Differences in Per Ca pita Income Growth , 1929-79, " urvey q/  Current Business 62 :24-33.  Gruben, Willi am c. , 'md D onald W. Hayes (1991), "For casting the Lo uisiana E o nomy, " l~ ed ral R serve B'lnk of D allas Economic Review, March, 1- 16. Tsa r I, Walter ( 960), Methods q/ Regional A nc/~ysis: A n introduction to Regional cience (Ca mbridge, Mass.: M.I.T. Press) . Jackson , Gr gory, G org Mas ni ck , Roge r Bolton, usa n Bartlett, and Jo hn Pitkin (1981), Regional  Dive1"Sity: Growth in the United tates, 19601990 (Bosto n: Auburn House). Lawrence, Hob rt Z. ( 984), Can America Compete? (Was hington , D .C.: Th Broo kings Institutio n) .  Gilm r, Robert W., and Allan G . PulSipher (989), "Structural Change in o uth rn Ma nufacturing:  14  Federal Reserve Bank of DaUas  Keith R. Phillips Economist Federal Reserve Bank of Dallas  The Effect of the Growing Service Sector on Wages in Texas  F  ro m 1978 to 1989, the av rage rea l wage in Texas increas d slowly. During the same period, employm nt in th Texas selv ice sector incr ased tro ngly, w illie employment in th goods sector declin d. Because selv ice jobs are generaUy thought t p ay mostly low wages, analysts ha ve suggested that. the groWdl in the. Iv ice s cror has been impo rtant in suppressing overall wag ga ins in d1e star . Wages in dle stare also have b come more d isp rsed across workers, o r in other terms, more un qual. Beca use the servic s tor is often perceiv d as providing many low-wag jobs, a moderate number of high-wag jobs, and I' laLively few middle-wage job , qu stions have b en raised about the ffect of til selv ic sector on wage in qua[j ty. Th growth and equ ali ty o f wages has an impo rta nt impact o n the welfare of th state's I' sid nts. Ga ins in til average rea l wage signify g'lins in the workers' standard of living. Wag in qu ality is also impo rtant. If th 'Ive rag wage is in creasing sol ly beca use of ga ins in a sm all number of high-paid jobs, then the incre'lse w iU do little to help th majo rity of wo rkers o r to aid social problem such as pov rty. Thi s stud y attempts to exp lain the effect of th e ri sing service secto r on chang s in av rage wag s and w ag dispersio n in Texas. r estimate the ffect of th shift in empl oym ent shares b tween goods and se rvices o n change in th stat 's av rage wag and w'lge di sp rsio n . r also exa m in wage gro wth and dispersio n in these two secto rs. With res p ct to wage growth , T find that the shi ft to se lv ice jobs only slightly dampened wage growth in Texas during the 1980s. The priJ1Cipai reason fo r slow wage growth was wea k wag expa nsio n in both th goods and service se to rs. 1 also find that the shift to servic -secto r jobs, Economic Review -  Nove mber 1991  ho ld ing dispersio n constant in each s cto r, had littl e f~ ct o n wage in qu ali ty in lh stat . Whil e overall w'lge inequ ality incr ased, the incr ase res ulted almost entirely from a larg increase in wage inequality in the goods secto r and a I' !-t tiv Iy smaU increas in wage inequality in the servi ce secto r. ' During th 1980s, the goods s cto r experi enc d a signi fi ca nt decline in the pro porti o n of wo rkers ea rnin g middl e wages and an incr ase in th propo rti on of wo rk rs ea rning both low wages and high wag s. During the sa me peri od, the pro po rti o n o f workers ea rnin g midd le and high wages increased in the service secto r. Th e pro po rtio n of workers ea rning low wag s in the service sector d c1 ined .  Defining the service sector Befo re analyzing grow th in the se lvice s ctor, it is important to defin e s rvice-sector jo bs. It is common to c1 as ify jobs in accord ance w ith a  I wish to thank William  C. Gruben. Joseph H. Haslag. and  Stephen P.A. Brown for helpful comments and discussions. I also would like to thank James L. Hedges for excellent p rogramming assistance. I  I do not attempt to analyze the root causes of changes in wage g rowth or dispersion. Suc/1a study would examine the factors affecting boll1 the supply of and demand for labor. For an example of a study tha t looks at supply and demand factors for the U.S. labor market. see Bound and Johnson (1989). Also. because of data restrictions, I look solely at wages, not total compensation. At the national level. total worker compensation has been growing at a faster pace than wages partly due to steep increases in the cost of health insurance.  15  set of g uid lines esta blished by the Office f Management and Budget (OMB) ca Ued th e standa rd industrial c/asstfi"calion (SIC). The broadest classification includes th fo ll owing ten industry divisio ns: agriculture; mining; co nstruction; m a nu(-~l cturing; transportation, ommunic ltio n , and public utilities (TCPU); w hol e ale trade; I' tail tra Ie; finance, insurance, and rea l stat (FIRE); s rvices; and government. The first fou r div isio ns in the list are industries that primar il y produce goods, whil th next six industri es primari ly prod uce s rvices. 2 Often Lhere is confusion beLween th narrowly defin d SIC service divisi o n and the larger g roup of service-producing indusLri s. ]n this 'Hticl , r define se/v ices as industries in which the primary utput is not a tangible good . These industri es includ e TCPU, who lesa le trade, reta il trade, FIHE, serv ices, and governm ent. Other services is th e SIC divi sion defin d as s rvic s, w hich includ es variou s indu stri es such as hea lth, busin ss, and ntertainment s rvic s.  Understanding the growth of service-sector jobs The perce ntages of Texas and U. . servicesecto r employment have incr ased dlro ughout this ntury (Fed ral Hes rv Bank o f Dall as 1986). Before 1960, most of the growth C'lme from a decreaSing share o f employm ent dev t d to agri culture. An impo rtant reaso n for the declin in the share of empl oyment in agriculture was stro ng  produ ctivity g rowLh, w hi ch aUowed farm rs to produ ce more outp ut w ith ~ w l' work rs. Sin c 1960, however, agri culture'S share of empl oyment sta bili zed , whi le the share of oth I' goods-produ cing indu stries declin ed . Much of this decline was du e to ga ins in manufacturing produ ctivity. Like the pI' vio us produ ctivity ga ins in agri culture, th in creased produ tivity in manufacturing all owed mol' production w ith fewer work rs. Anoth er importa nt fa ctor for th e ri se in th share of servi ce employment w as ri sing co nsum er incomes, which in reased the demand fo r servi ces such as h alth ca r , education , and financia l and recrea tional s rvi ces. A lso, a ri se in the number of wom en in the work force increased the dem'md for serv ices such as child ca re and restaurant services . A ltho ugh the incr ase in demand fo r consumer se rvic s w as an important part of the shift toward service mploym nt, mu ch of th shift stemmed fro m an in reased d m and for produc I' services . A d scribed in tanback , Bea r e, Noyell e, and Karas k (981), "Large firms ope rating in national and internati on al mark ts need the help of izab le, speciali z d sta ffs for pl anning 'md control , as w II as for innovatio n, marketing, and other management fun ctions that require mol' specialists both in-house and out-oF-hous . .. .It is in correct to visuali z the U nit d tates economy as turning its back o n th produ ction of goods .... Hather the natur of how produ ction takes place is und rgoing rapid hange, w ith more and more depend nce on the rol e of produce r services."  The service sector in Texas  16  2  Another way to classify service-sector jobs Is by occupation. For example, receptionists in manufacturing industries are considered to be in the service sector when classified by occupations, but they are considered to be in the manufacturing sector when classified by industry. Th e focus of this study is the effect on wages of the growth in industries that primarily produce services. Another interesting topic, which t do not address, is the effect of the growth in service occupations on wages.  3  The industry classifications shown in Table 1 are at the three-digit SIC level. The data in Table 1 end in 1987 because changes in the SIC code classifications beginning in 1988 significantly affect the comparability of the data at the three-digit level.  From 1978 to 1987, the LwenLy iJldustri s that showed the highesL gain in their employmenL shar in Texas w ere all from th service sector. As T'lbJ 1 shows, these growth industri s include low-wage industries such a re taUl-ants, middle-wage industries such as insurance compani es, and high-wage industries such as phYSiCians' offic S.3 The data in the table are from repo rts filed by employers subjecL to state unemployment insurance laws and include annual wages paid to full- and pa rt-tim e work rs. Although ho urly wages are a mol' comparable measure of compensation 'tcross industries, hourly wages are not available in the Cov I' d Employm nt Federal Reserve Bank of Dallas  Table 1  The Top Twenty Growth Industries in Texas, 1978-87  Industry Eating and drinking places Miscellaneous business services Health and allied services, not elsewhere classified Grocery stores Personnel supply services Legal services Computer and data processing services Real estate operators (except developers) and lessors Offices of physicians Savings and loan associations Real estate agents and managers Department stores Services to dwellings and other buildings Air transportation, certified carriers Insurance agents, brokers, and services Private household services Hotels, motels, and tourist courts Accounting, auditing, and bookkeeping services Mortgage bankers and brokers Child day care services  Change in Employment Share  1987 Average Annual Wage  Industry Division  .385 .261  $ 7,810 $19,765  Retail trade Other services  .185 .182 .149 .128 .121  $ 7,236 $12,300 $11,830 $37,619 $32,863  Other services Retail trade Other services Other services Other services  .084 .083 .078 .078 .069 .062 .063 .056 .055 .054  $16,559 $46,082 $22,802 $19,735 $11,801 $ 8,670 $33,070 $23,645 $ 8,472 $10,412  FIRE Other services FIRE FIRE Retail trade Other services TCPU FIRE Other services Other services  .053 .045 .043  $25,440 $28,064 $ 7,312  Other services FIRE Other services  NOTE: The change in employment share represents the industry's percentage share of employment in 1987 minus the percentage share in 1978. FIRE is the standard Industrial classification division finance, insurance, and real estate, and TCPU is the division transpor-  tation, communication, and public utilities.  SOURCE: U.S. Bureau of Labor Statistics, Covered Employment and Wages.  and Wages data set. (See the Appendix/or em expla-  nation q/dala sources.) The industries in Tabl 1 reflect the growing demand for both produ ce r and consumer services. Many of these industri s produce exportabl se rvi ces such as business, lega l, and air transportati on s rvices. Th se industri es ca n act like traditional manufacturing and agricu ltural industri s that expo rt Lheir o utput and , in I' turn, bring o utside ca pital into th regio n. Also, many of the industri s, such as child ca re and grocery stores, are I' gio nally ba s d. Whi le the overa ll se rvice secLO r ha s expan led strongly, rates o f expansio n vari d wide ly Economic Review - Novembel' 1991  across the broad selv ice ca tego ries, as shown by th e empl oyment data Fro m the Nonagricultural Establishment Survey (Figure 1). The I ader of growdl was the olher services category, wh ich includes rapidly growing industries such as hea lth, busin S5, and legal services. This ca tegory increased at a ro bust 5.3 perc nt annual rat during the deca d . The TCPU seCLo r W'IS generally dle wea kest selvice category, although it 11'IS grow n stro ngly ince 1986. Gov l'llment employ ment, w hich is dom inated by educa tio nal selvices, expanded throughout the decade. Trade grew strongly until the state's downturn in th mid1980s, and its recovery has b n slow; empl oy17  Figure 1  Figure 2  Texas Service-Sector Employment (Seasonally Adjusted)  Texas Goods-Sector Employment (Seasonally Adjusted)  Thousands of workers  Thousands of workers  1,800  .....................,...  1,600  Trade •••.•.•.•••••• ........... Other Services  ..........  1,400  1,000  800  1,200 1,000  1,200  .........................../ .........600  800  Construction 400  600 400 200  Transportation, Communication, ... ... ~~~.:..~~~I~ .~~I!~~i~~........... .. ..... ··• · .. · .. •  200  Mi~·I~:·······..······..·....................................  ..........................................  Finance, Insurance, and Real Estate  O+-~~~~~~.-.-.-.-.-.-.-.-.-r-r-r-"  '70 '71 '72 '73 '74 '75 '76 '77 '78 '79 '80 '8 1 '82 '83 '84 '85 '86 '87 '88 '89 '90  O+-~~~.-.-.-.-.-.-.-.-.-.-.-.-.-.-r-~ ~~~~~~nTIn~w~_~~  • • w.u~  SOURCE OF PRIMARY DATA : U.S. Department of Labor, Bureau of Labor Statistics, Nonagricultural Establish ment Surveys.  SOURCE OF PRIMARY DATA : U.S. Department of Labor, Bureau of Labor Statistics, Nonagricultural Establishment Surveys.  me nt o nly rece ntly has urpa sed its previous peak. FIRE grew until mld-1986 and has not yet shown signifi ca nt signs of a turna round. Wh il e many service industries increa ed strong ly, many industries in the goods sector declin d . As Figure 2 shows, nonagri cul tural employm nt in th e T xas goods secto r d dined during the 1980s. Manu facturing and mining mployme nt pea ked nea r the beginning of 1982, and construction mployment pea ked in mk l-1984.4 Whil e gr w th in the e rvic sector's shar of mp loyme nt in Te'xas wa o mew hat weak r than in the natio n du ring th arly 1980 , fro m  1978 to 1990 th und rlying tr nd was s imil ar (Figure ). The servi e ector's hare in Texas in creased from 71 pe rce nt ill 1978 to 79 percent in 1990, w hil in the nation it in creased from 70 p rcent to 77 pe rcent.  , Although the main focus of this article is on 1978- 89, I have plolted th e employment data from 1970 to give some historical perspective. Because I chose to extend the period before 1978, I was unable to use the ES-202 data, which include agricultural employment. Instead, I used the nonagricultural establishment survey data ; therefore, Figure 2 does not include agricultural employment. • Because of oversampling of certain population groups in the March 1990 survey, the data presented in this analysis were adjusted using the March supplemental weights provided by th e U.S. Department of Commerce.  18  Industrial change and wage growth in Texas Did the mploym nt shi ft to services du ring the 19805 hav an important impa t on wag growth in Texas? Wage growth was w ak, w hile the shift toward servic s acc I rated. But w hil the changing ind ustria l com position of th Texas work fo rc red uc d wage growth, the effect was relative ly small. To account fo r th effect that pa rt-time workers have o n average annual wages, I analyze data fro m the Ma rch Current Population UIV ys (CPS) fro m the U.S . Bureau of the Ce nsLl .5 Fro m 1978 to 1989, wages fo r year-rou nd, full-time (YRFT) work rs in Texas wer about 17 perc nt lower in th s Iv ice sector than in the goods sector ( Table 2). Even w ith this d iffe r nce, however, the vid nce from the CP shows that the shjft to s rvices only slightly slowed wage growth among all YRFT worke rs. (Fo r em explal1ation q/ tbe i'l1:fl.alion measure used to dqf7a te wages, see tbe Fed eral Reserve Bank of Da llas  Figure 3 Service-Sector Share of Nonagricultural Employment Percent 80 78 76 74 72  ....... .........................................  70 68 66 64 62 60 +-~~~~~~~r-r-r-r-~~~~-.-.-.-. 70 7 1 72 73 74 75 76 77 78 79 '80 '81 '82 '83 '84 '85 '86 '87 '88 '89 '90  SOURCE OF PRIMARY DATA: U.S. Department of Labor, Bureau of Labor Statistics, Nonagricultural Establishment Surveys.  bo.x lilled "M easuring h~jlalio l1 in Texas. ")  To analyze the effect of the shift to services, I se parat th growth rate in tota l wages into wag growth in th goods sector, wag growth in the service secto r, and th e shift in employment shares betwe n the sectors. From 1978 to 1989, rea l annu al wages for YRFT wo rkers increased at an annu al rat of 0.7 percent in the goods and service sectors and 0.6 p rcent for all workers (Fig u,re 4) If th shares of worke rs in the goods and se rvice s ctors had remained constant over this pe riod , the rea l annual wage growth would have be n 0.7 percent, o nly 0.1 percentage po int higher than it actu ally was .  percent incre'lsed from 39.5 percent to 42. 1 perce nt. Wages received by the middle 60 percent of workers declined from 53.1 percent to 51.3 percent. Figure 5 shows the increase in wage inequality. Th p rcentage of workers, sOlted by d1eir wages, is plotted o n the ho ri zo ntal axis, whil th e percentage of total wages received by the workers is plotted on the verti ca l axis. Th distributions show n are typica Uy ca ll ed Lorenz cu rves. If wages were equal acros wo rkers, the Lorenz curve wo uld be equal to th 45-degre line. As Figure 5 shows, the Lorenz curve was slightly furth er from the 45-degr e line in 1989 than in 1978. Wages beca me mo re unequally distributed in both the goods and selvic sectors, although d1e change was largest in th e goods sector. As Figures 6 and 7 show, the Lor nz curve for th goods s ctor shifted signifi ca ntly mo re than the Lore nz curve for the selvice sector. Although in 1978 wage disp rsio n was cI 'Irly grea ter in the se rvice se tor, th e relatively large in cr ase in dispe rsion in the goods sector ca used disper io n in the two sectors to b quite similar in 1989. In 1978, the Lo renz cUlve fo r the goods sector was entir Iy within the Lorenz curv for the selv ic sector (h gu re 8). By 1989, however, th e two curves crossed , impl ying that wages Figure 4 Average Annual Wages for Year-Round, Full-Time Workers in Texas Thousands of 1989 doliars 31 30 29 28  Wage dispersion in Texas  27 26  The perc pri on of s Ivice jobs being eith er high-wage jobs, such as docto rs and lawyer , or low-wage jobs, such as busboys 'Ind janitors, has ca us d speculation that th e growth in the service s ctor has led to grea t r wage inequality . h om 1978 to 1989, wag s for YRFT wo rkers in Texas becam less equall y distribut d. Willie the shar of wages rece iv d by th lowest-paid 20 p rcel1l of the work force declined from 7.4 percent to 6.6 pe rcent, the share rec iv d by the top 20 Economic Review - Novembel' 1991  25 24 23 22 21 78  79  '80  SOURCE OF PRIMARY DATA: U.S. Department of Commerce, Bureau of th e Census, Current Population Survey, March surveys.  19  Table 2  Employment and Wages for Year-Round, Full-Time Workers in Texas Annual Wages (1989 dollars)  Number in Sample 1978  1989  1978  1989  2,083 (56.7)  2,453 (60.1)  $24,123  $25,864  Employment Shares Goods industries  36.3 (61 .7)  30.9 (66.8)  $26,580  $28,841  Service industries  63.7 (54.2)  69.1 (58.4)  $22,719  $24,530  NOTE: The number In parentheses Indicates year-round , full -time workers as a percent of total workers. SOURCES: U.S. Department of Commerce, Bureau of the Census, Current Population Survey, March 1979 and March 1990.  in the goods sector were no longer unambiguou Iy mor equ ally distributed than in the Iv ice s ctOI' (F igure 9). Altho ugh the Lo renz curve i us ful in describin g wage inequ ality, its use is limited in meas uring h w changing dispersio n in diffe r nt cto rs affects the total di .p rsion (se Lam 1986). To appo J'[io n the increase in tota l wage d ispersio n in to factors r lating to the goods and s Iv ice s cto rs, I 'Inalyz th va ri ance o f dle natural log f wages (va rlnwg). An incr ase in th var/17wg signifi s an increase in wag in qu ality. Chang s in this m asure ca n be broken down asil y into th contributio ns from dle goods and selv ice s ctors.6 As me'lsur d by dle varln wg, fro m 1978 to  1989 wage dispersion in creas eI 27.3 P r ent in th e goods s ctor, 16.4 perce nt in th s rvice s ctor, and 19.3 p rcent ove rall. Thes increases are consist nt w id1 the shjft in the Lorenz curves describ d ea rlier. Whil w age disp rsio n incr ased fro m 1978 to 1989, v ry li ttle of the elisp rsio n ca n be attributed to the shi ft in mployment hal' s from th goods s ctor to th . ervi e sector. This fact is shown by first fa to rin g o ut the va ri anc of tota l wages in to th va ri ance of wages in the goods and se rvice s cto rs, the empl oyment shares in each secto r, and th differ nc in mea n wages in the two sectors. As shown in Lam (1986),  +  N, 8  Ng+ N, 6  OR  +  N,  , + NH  I\T  vaT I11W() , O s  Champernowne (1974) analyzes six measures of inequality and finds that different measures are more sensitive to different types of Inequality. He shows that the variance of th e natural log is sensitive to inequalily that is due to very low wages.  20  vad nw o  where N is th number o f wo rkers, )1 is th m an, and th subscripts II and s I' pres nl th goods and s Iv ice secto rs, r sp ctively. Federal Reserve Bank of Dallas  Measuring Inflation in Texas To measure changes in real wages, or in other words, changes in purchasing power, a measure of inflation is needed. Little information is available about inflation on the state level , and using a U.S. price index could misrepresent regional price changes. I define a Texas consumer price index (CPI) by combining the movements in the CPls for Houston and Dallas- Fort Worth , indexes that are produced by the U.S. Bureau of Labor Statistics (BLS) . One problem in using either regional or national CPls produced by the BLS is that the indexes were constructed differently before 1983. The main difference in the indexes is that, beginning in January 1983, the home ownership component ofthe CPI was changed to a rental equivalence method . This method better estimates the actual cost of shelter services consumed by homeowners. Before 1983, a method that inflated the costs of home ownership was used. This effect was especially important during the late 1970s and  early 1980s, when mortgage rates were at historically high levels. l Although the rental equivalence method of measuring home ownership cost in the CPI began in 1983, the BLS experimented with this measure before that date. The CPI-U-X1 was the direct antecedent of the current method used to calculate the CPI. Subtracting annual changes in this measure from annual changes in the U.S. CPI gives an estimate of the bias in the inflation estimates due to the improper measurement of home ownership costs. Subtracting this bias estimate from annual changes in the Texas CPI should give a better estimate of inflation in the state. As Figure A shows, the adjusted CPI indicates that inflation was high in the early 1980s yet significantly lower than without the housing cost adjustment. 1  For more detailed information about the switch to the rental equ ivalence method, see U.S. Department of Labor (1983) . For a comparison of the currently reported CPI and the CPI-U-X 1 for 1969-82, see Economic Report of the President (19 83).  Figure A  Inflation in Texas Percent change 16 14  •  Published Consumer Price Index Adjusled Consumer Price Index  1978  1979  1980  1981  1982  1983  1984  1985  1986  1987  1988  1989  SOURCE OF PRIMARY DATA: U.S. Department of Labor, Bureau of Labor Statistics .  Economic Review -  Novembel' 1991  21  Figure 5 Wage Distribution for Year-Round, Full-Time Workers in Texas (Lorenz Curve)  Figure 6 Wage Distribution for Year-Round , Full-Time Workers in the Texas Goods Sector (Lorenz Curve)  Cumulative percent of wages  Cumulative percent of wages  100  100  90  90  80  80  70  70  60  60  50  50  40  40  30  30  20  20  10  10  o  10  20  30  40  50  60  70  80  90  100  Cumulative percent of workers  o  10  20  30  40  50  60  70  80  90  100  Cumulative percent of workers  NOTE: The 45·degree line represents total wage equality.  NOTE: The 45·degree line represents total wage equality.  SOURCE OF PRIMARY DATA: U.S. Department of Commerce . Bureau of the Census. Current Population Survey. March surveys.  SOURCE OF PRIMARY DATA: U.S. Department of Commerce . Bureau of the Census. Current Population Survey. March surveys.  By holding the employment shares in the above q uatio n fixed between 1978 and 1989 and then comp'lring this change in disp rsio n to th  actu al chang , it is possible to isolate th e effect of the shi ft in employm nt shares. 7 The sh ift in employme nt hares accounted for o nl y 0.9 p rce nt of the increase in lOta l wage va riance. The change ill the means in the two secto rs actually decreased th va riance by 2.4 pe rce nt, whil e th change in va riance in the goods and servic sectors accoun ted for 99.4 perce nt of the actual net increase in the va ri ance. s Tn summa lY, the contenti on that wages in Texas have become less eq ually distribu ted is suppo rted by both the shifting out of the Lorenz cUlve and the increase in the va riance of the natural log of wages. However, the co ntention that the incr ase in wag ineq uali ty was ca used by the shi ft of employment from th goods s ctor to the service sector is not suppo rted. Rathe r, increases in the d ispersion in both the goods and  22  7  This decomposition concentrates solely on the direct effects that the change in employment shares 11as on total wages. It is likely. however. that the changing employment shares also have an effect on relative wage growth in each sector. This effect reduces the usefulness of the simulation because it is unlikely that wages would have changed the way they did if employment shares had remained constant. Thus . while the decomposition gives useful insight into the dynamics of a change in wage dispersion. the information gain is limited by interrelationships among the component variables.  8  An interaction effect accounted for the remaining 2. 1 percent of the increase in the varlnwg .  Federal Reserve Bank of Oa Uas  Figure 7  Figure 8  Wage Distribution for Year-Round, Full -Time Workers in the Texas Service Sector (Lorenz Curve)  Wage Distribution for Year-Round, Full-Time Workers in Texas, 1978 (Lorenz Curve)  .  Cumulative percent of wages  Cumulative percent of wages  100  100  90  90  80  80  70  70  60  60  50  50  40  40  30  30  20  20  10  10  O~~~--,---,---,---,---,---,---~--~--.  o  10  20  30  40  50  60  70  80  90  100  o  10  20  30  40  50  60  70  80  90  100  Cumulative percent of workers  Cumulative percent of workers NOTE: The 45-degree line represents total wage equality.  NOTE: The 45-degree line represents total wage equality.  SOURCE OF PRIMARY DATA: U.S. Departmenl of Commerce, Bureau of the Census, Current Population Survey, March surveys.  SOURCE OF PRIMARY DATA: U.S. Department of Commerce , Bureau of the Census, Current Population Survey, March surveys.  s to r ca us d virtuall y all th ove rall wag di spe rsio n.  wage alegori s revea ls that a significa nt declin in the pro porti o n of middle-wage ea rn ers at l ast partiall y ca used Lhe in cr as d w'lge dispersio n amo ng Y"RFT workers in the goods s Clor, as the dislributio ns in Figure 10 show. 9 Bel ween 1978 and 1989, th e p rcent'tge of goods-secto r workers ea rning middl wages (lh e box d area in Figur 10) declined, w hil the perce ntage e'lrning high and low wages increased . The perce nl'lge o f workers arnin g middl e wag s in the goods secto r  serv ic  in crease in  A closer look at wage inequality Th conclusio n that th e shifl LO services ha nOl had a larg effecl o n wage dispersio n does nm disprov th contention thal low- and highw age jobs in the s rvi ce secto r are repl acing middl -w 'lg jo l s in th goods sector. Rath r, it says lhat lh shift itse lf, ho lding s ctor dispersio n constant, W ' IS not a larg part o f lhe increas in wage disp rsio n. The in r ased di spersio n in both th e goods and servi e s cto rs is consist nl w ilh bOLh a r du clio n in middl e-wage jobs in th goods seClor and growlh o f low- and high-wage jo bs in rhe se rvi ce s Clor. An exa min ati o n o f ea rnin gs across the till· e Economic Ueview -  November 1991  • Any definition of middle-wage jobs is arbitrary and depends on factors such as the worker's age and whether the worker lives in a metropolitan or rural area. In this study, I define middle wages as the range between $1S.000and $40,000, in 1989 dollars.  23  Figure 9 Wage Distribution for Year-Round , Full-Time Workers in Texas, 1989 (Lorenz Curve) Cumulative percent of wages 100 90 80 70 60 50 40 30 20 10 o~~.---.---.---.---,---,---,---,---,---.  o  10  20  30  40  50  60  70  80  90  100  Cumulative percent of workers NOTE: The 4S-degree line represents total wage equality. SOURCE OF PRIMARY DATA: U.S. Department of Commerce , Bureau of the Census, Current Population Survey, March surveys.  the distributions in Figure 11 show that from 1978 to 1989 mer was a slight decrease in the perc ntage of service workers earn ing low wages and an increase in the percentage receiving middle and high wages. The percentage of workers I' ceiving low wages declin d from 35 percent to 33.4 per ent, dlose I' c iving midd l wages incr ased from 54.7 p rcent to 55.2 percent, and thos receiving high wages increased from 10.3 percent to 11.4 percent. Thu ,whil th Lorenz clllve 'md the uadnwg showed an increase in wage in qua lity in dl seIvice sector, it appears that the increase pardy was due to stronger groWdl in the propoItion of workers ea rning high wages d1'l1l in the proportion of workers ea rning midd le wages. 10 It is also int resting to note that by 1989 a greater percentage of workers in dle service sector r ceiv d middle wages than thos in th goods sector. The changing distribution of wages in the seIvi.ce s ctor offs t much of dle decline in the p roportion of middle-wage jobs in the goods se tor. Overall, the percentage of workers receiving middle wages d creas d from 56.8 percent in 1978 to 54.7 perc nt in 1989 (Figure 12). Changes in the seIv ice ctor also offs t most of dle increase in dle proportio n of low-wage jobs in the goods sector. The p rcentage of workers receiving low wages increased only slighdy, from 30.7 percent to 30.9 percent, while the p rc ntage of workers r c iving high wages increased from 12.5 percent to 14.4 p rc nt. Conclusion  decreased from 60.6 percent to 53.5 percent, workers earning low wages increased fro111 23 percent to 25.3 percent, a nd workers arning high wages increa ed from 16.3 percent to 21.2 percent. The data, however, do not support dle contention dlat me groWdl in services has b en conc ntrat d in low- and high-wage jobs. Instead,  '0  24  Although the relative strength of high-wage jobs likely affected the increase in the varlnwg and the shift in the Lorenz curve, many other factors are also important. The analysis of the distribution of workers across broad wage categories ignores distributional changes within the wage classes and is dependent on the definition of the classes. Also, the Lorenz curve and the varlnwg measure variation around a given mean, while the distribution of workers across wage classes is not mean invariant.  Obs Ivers have given mu ch attention to the decline in middle-wage jobs in the goods sector and me rise in seIv ice-sector employment. While th selv ice sector has had a positive impact on employment growth in Texas, peopl often view growdl in dle ervice sector as providing mostly low-wage jobs. Data PI' sen ted her show that in Texas YRFf workers in tJle seIv ice sector g t paid about 17 percent less than YRFT workers in the good sector. How ver, the growing share of selv icesector jobs during the 1980s had little dir ct effect on ove rall wage growth in dle stat. The primalY source of slow wage growth wa weak wage growdl in both th goods and selvice sectors. Another criticism of the growing selvice sector is dlat it provide velY D w midd le-wage jobs, Federal Reserve Bank of Dallas  Figure 10  Distribution of Wages in the Texas Goods Sector Percent of workers 25 •  1978  •  1989  20 Middle Wages  15  10  5  0.,..---..... 0-5  5-10  10-15  15-20  20-25  25-30  30-35  35-40  40-45  45-50  50-55  55-60  6CHl5  65-70  70-75  75-80  8CHl5  >85  Annual wages for year-round, full -time workers, in thousands of 1989 dollars SOURCE OF PRIMARY DATA: U.S. Department of Commerce, Bureau of the Census, Current Population Survey, March surveys.  Figure 11  Distribution of Wages in the Texas Service Sector Percent of workers 25 •  1978  •  1989  20 Middle Wages  15  10  5  0-5  5- 10  10-15  15- 20  20-25  25- 30  30-35  35-40  40-45  45- 50  50-55  55-60  6CHl5  65-70  70-75  75-80  80-85  >85  Annual wages for year-round , full -time workers, in thou sands of 1989 dollars SOURCE OF PRIMARY DATA: U.S. Department of Commerce, Bureau of the Census, Current Population Survey, March surveys.  Economic Review -  November 1991  25  Figure 12  Distribution of Wages in Texas Percent of workers 25 •  1978  •  1989  20 Middle Wages 15  10  5  o 0-5  5-10  10-15  15-20  20-25  25-30  30-35  35-40  40-45  45-50  50-55  55-60  60-65  65-70  70-75  75-80  80-85  >85  Annual wages for year-round, full -time workers, in thousands of 1989 dollars SOURCE OF PRIMARY DATA : U.S. Department of Commerce , Bu reau of the Census, Current Population Survey, March surveys.  many low-wag jobs, and a moderate amount o f high-wage jobs. Two measures of wage clisp r 'io n, the va ri ance of tll natural Jog of wage and tb e Lo renz curve, show that w ages fo r Yl~FT workers were m ore unequall y distributed in the servi ce sector than in the goods secto r in 1978. H oweve r, by 1989 the Lo renz curv could no lo nger veri fy that wages we re mo re unequ al in the servi e secto r than in the goods secto r. During the 1980s, both measur s o f disp rsio n sh owed that ove rall wage dispersio n increased. T he increase was ca used by a relati v ly stro ng incr ase in wage disp ersion in th e goods s cto r and a relati ve ly moderate increase in w age disper. ion in the service sector . Th shift to  26  service-s ctor jo bs, ho ld ing dispersion co nstant w ithin se tors, was r sponsible for little o f the in cr ase in w ag dispersion. A n an alys is of the distributi o n of w orkers across wage cIa e revea l that the large in reas in wage disp rsio n in the Texas goods sector during the 1980s was ce nter d in the loss of workers earning middl w ages. During this peri od, how ver, the selv ice sector saw an increase in dl percentage of w rkers ea rning middl wages . The increase in th share of ervic work rs ea rning middl wages an I a decrease ill the share of workers ea rnin g low w 'lges p arti ally offset dle opposit Ffects oc LllTing ill the goods sector.  Federal Reserve Bank of Dallas  Appendix  used three sources of employment data in this study. I used the Nonagricultural Establishment Survey data from the U.S. Bureau of Labor Statistics (BLS) to define growth in the goods and service sectors, as Figures 1-3 show. These data are based on a survey of establishments that record the number of employees on the payrolls during the pay periods that include the twelfth day of the month . The survey is a large sample, covering approximately 40 percent of all nonfarm employment. The survey results are bench-marked to the BLS's Covered Employment and Wages data, which represent nearly all establishments with more than one employee . While this establishment survey data set covers a broad sample of employers, its usefulness is limited by the fact that its wage data cover only a limited number of industries. I used the Covered Employment and Wages data set for the analysis in Table 1. This data series , often called the ES-202, includes the number of jobs in each month and the quarterly wages paid. The benefit of the ES-202 is its wide coverage of both employment and wages, representing almost all companies in the state that have at least one employee . A weakness of the ES-202 is that it does not account for part-time workers. Changes in hours worked or the percent of  Economic Review -  November 1991  part-time workers can have an important impact on quarterly and annual wages across industries and over time. The third source of employment and wage data is the Current Population Survey (CPS) from the U.S. Department of Commerce, Bureau of the Census. I used this data set for most of the analyses in this article. An advantage to using the CPS data set is that it contains records of individuals so that wage dispersion can be measured . The individual records also indicate ifthe person works'yearround , full time , so that this sample can be analyzed separately. There are several main differences between the ES-202 data and the CPS data series . The ES-202 has wide coverage, while the CPS is a relatively small sample. The CPS counts persons employed , not jobs . If a person has several jobs, he is counted only once in the CPS data, while the ES-202 would record each of the individual jobs. The CPS counts workers at their place of residence , while the ES-202 counts jobs at the place of work. The CPS excludes workers younger than 16 years old, while the ES-202 does not. '  1  For more information on the comparability of the ES·202 and the CPS data series , see U.S. Department of Labor (1989) .  27  References Bound , John , and Gorge Jo hnson (989), "Changes in th Structu re of Wages During th 19805: An Ev'duatio n of A ltern ative Explanatio ns, " NBEH Working Paper Series, no. 2983 (Ca mbridge, Mass.: National Bureau of Economic H s arch, May). Champern ow ne, D. G. (974), "A Compa riso n o f Measures o f Inequality o f Income Distribution ," The i:.con omic ]ou rna l, D eceml er, 787-809.  Lovem'ln , Gary W. , 'Ind Chris Tilly (988), "Good Jo bs o r Bad Jobs: What Does th e Evidence Say?" Federa l Hese rve Bank o f Boston New England Economic Review, January/ Feb ru ary,  46-65. Sta nbac k, Thom'ls M., Jr. , Peter J. Bearse, Th ierry J. Noyell , and Robert A. Karasek (981), ervices: 71.?e New Econo111Y, Co nse rva Li o n o f Human R so urc s S ri s, no. 20 (Totow'l, N ..J. : A ll enh elcl , Osmun).  Econ om ic Report q/tiJ e President (983)' U.S. Governm ent Printing Office, Washington,  D.C. ,226. F cI ral Reserve B'lI1k of Da lhs (986), "The Growth of S rvices in th Texas Economy," 1985 A nnual Report (D all as: Fede r'tI Reserve Bank of Dall as, 4-17).  U.S. Depa rlment of Labor, Bureau of Labor Sta tistics  (989), Employment a nd Wages Annual Averages, .1989, Bu ll tin 2373, October, 530-33. - - - (983), "Changing th e H o m ow n rship Co mpon ent o f th Co nsumer Price Index to Rental Equivalence," CP! Detailed Rep ort, Janu ary, 7-13 .  Lam, Dav id (986), "lhe D ynami cs of Po pul atio n Growth , Differential F rtiliLy, and Ineq uality," America 17 Econom ic Review 76 (5, December):  110-16.  28  Federal ReseJ'vc Bank of DaUas  Other Publications Available from the Federal Reserve Bank of Dallas To re 'e ive o n o f o ur olher publi ca ti ons, write th e Public Affa irs Dep'lltm ent, Feden l Heselve Bank o f D'tll as, tatio n K , Da ll as, Texas 75222, o r phone (214) 698-4436,  Annual Report Published in F bruary and aV'li lable free of ci1'lrge.  Crossroads A quarterl y newsJetter devoted to developm nts in fil1'lnci al se rvices and special to pi cs relating to the Federa l Heserve System. Free.  Tbe Soutbwest Economy A bimonthly newsletter about econom ic conditions and business developments in the o uthw st. Hece nt issues have cove red 'Igri culture, b'lnking, energy , high techno logy, manufa cturing, and intern atio nal trade. hee.  Trade-Weighted Value of the Dollar A monthly news letter prese nting a meas ure of the trade-we ighted va lue of the U .S. d oll ar aga inst 101 trading partn ers. Th e publica ti o n also prese nts subindexes for sp cifi c co un tries and geogr'lphical areas. Annua l subscripti o ns are $48.  Texas Industrial Production Index A mOnLhl y updat o f a m asu r of th e Texas industri al secto r's o utput. Free.  Quarterry Survey of Agricultural Credit Conditions in the Eleventh Federal Reserve District A statistic il release compiled quarterl y from a survey of ElevenLh District agri cultural bankers. Fr e.  Houston Business Briefs A newsletter about th Gu lf CO'lSt econo m y, w ith an emphas is o n Housto n. Published at six-week interva ls. E'lch issue contains an economic fea ture and a repo rt abo ut the 1110st rec nt HOllston Beig Book survey . Fr e.  Financial Industry Studies A sem i'lnnu al se ri es that pI' sents resea rch on economic and fil1'ln cial topics, w ith an emph asis on rh e finan ch l industry ar regiona l, natio nal, and intern 'ltio nal I vels. Free.  Economic Review -  Novembel' 1991  29