BUILDING US UP FOR A FALL? CONSTRUCTION AND STATE TAX REVENUE

Transcription

1 P O L I C Y b r i e f Washingon Research Council PB February 27, 2007 BRIEFLY Consrucion earnings as a share of Personal Income is a a 25-year high in Washingon Sae. A reducion in he share back o normal levels would resrain he growh in sae ax revenues, even if he overall economy were o remain srong. Washingon Research Council 108 S Washingon S., Suie 406 Seale WA voice: fax: BUILDING US UP FOR A FALL? CONSTRUCTION AND STATE TAX REVENUE In an earlier brief on Governor Gregoire's proposed budge for , we warned ha he sae may experience below average revenue growh over he nex four years (WRC 2007). This brief provides more deail on ha poin. The reason o expec below average growh going forward is ha sae revenue (paricularly from he sales ax) is sensiive o he share of consrucion in economic aciviy. Consrucion s share is a a 25-year high; a reducion in he share back o normal levels would resrain he growh in ax revenues, even if he overall economy were o remain srong. SALES AND USE TAX BASE GROWTH VERSUS PERSONAL INCOME The reail sales ax and is companion use ax ogeher provided $7.3 billion o he sae general fund in fiscal year This represened 56 percen of he $13.3 billion general fund-sae revenues for he year. In conras, for fiscal year 1975, sales and use axes oaled $662.3 million. Over ime, ax revenues change boh because of changes in he level of economic aciviy and because of changes in ax laws. For example, beween fiscal years 1975 and 2006 Washingon personal income, a measure of economic aciviy in he sae, grew from $22.4 billion o $229.4 billion; cerain goods were exemped from he sales ax, noably food purchased for off premises consumpion (groceries, for shor), manufacuring machinery and used cars, while he sae sales and use ax rae increased from 4.5 percen o 6.5 percen. When sudying he performance of ax sysems, careful analyss disinguish beween hese wo sources of change. Imporan ools o his end are consan ax base series, which hold fixed he definiion of wha is subjec o ax. Washingon Sae s Office of he Forecas Council (OFC) mainains consan base series for he sae s major axes. The OFC base reflecs sae ax law as of Char 1 makes use of he OFC consan base series for Washingon s sales and use axes. Shown is he raio of he general fund sales and use ax base o sae personal income (he ax base share, for shor) over 34 years, beginning wih he hird quarer of 1973 and ending wih he second quarer of Over he period, he sales and use ax base share grew from 65 percen of personal income o 75 percen. This dramaic growh allowed he general fund o easily weaher Iniiaive 345, approved by voers in November 1977, which exemped groceries from he sales and use axes, reducing he sales ax base by more han 10 percen. (By consruc-

2 Washingon Research Council ion, his change is no refleced in he consan sales and use ax base series and, herefore, does no appear on Char 1.) Even wih he loss of his much base, legislaors were able o reduce he sales ax rae from 4.6 percen o 4.5 percen in The subsequen fall in he ax base Char 1: Sales and Use Tax Base Share of Personal Income, 73:3 06:2 share was even more dramaic, however. By 1983 he sales and Independen Variable in Fied Equaion is Personal Income use ax base was 57 percen of personal income. This drop was 80% 75% largely he source of he rolling Tax Base Share fiscal crisis ha defined he John 70% Fied Equaion Spellman governorship. 65% In he laer half of he 1980s he ax base share grew modesly, 60% reaching a cyclical peak of 61 percen of personal income in he las 55% quarer of During he1990s, 50% he share declined, so ha by :3 75:1 76:3 78:1 79:3 81:1 82:3 84:1 85:3 87:1 88:3 90:1 91:3 93:1 94:3 96:1 97:3 99:1 00:3 02:1 03:3 05:1 i was 55 percen. I fell o 52 percen during he recession. Tax Base is Office of he Forecas Council Consan Sales and Use Tax Base. From ha low, i has grown o 58 percen of personal income. Personal income is a key variable in he sandard models used by economiss o explain he growh in he sales and use ax base (Holcombe and Sobel 1997, Chaper 5; WSTSSC 2002). The simple inuiion for his choice inerpres he sales ax as a ax on consumpion. As personal income rises, consumpion expendiures rise, and some of he incremenal purchases are of goods and services ha are subjec o he sales ax. The siuaion is acually a bi more complex han his, however, as he sales ax base includes business purchases as well as consumer purchases. (The Washingon Sae Deparmen of Revenue esimaes ha 36 percen of reail sales and use axes come from purchases by businesses.) Personal income is beer hough of as a proxy for he general level of economic aciviy, which deermines boh consumer and business purchases of axable goods and services. The appendix o his brief presens a simple model relaing he raio beween he sales and use ax base and personal income o he level of personal income and leas squares esimaes of he parameers of ha model. The r 2 of his regression is 0.705, meaning ha personal income alone explains 70.5 percen of he variaion of he sales and use ax base share. We plo fied values from his regression on Char 1. The fied values curve slopes downward indicaing ha he sales and use ax base share has ended o decline as personal income has grown. (I is imporan o remember, however, ha he base used in consrucing he curves on Char 1 includes groceries, which are now exemp in Washingon. Wihou groceries, he decline in he ax base share of personal income would be much less severe. See he discussion in he appendix.) Personal income by iself does no explain he spike in he raio beween he ax base and personal income experienced in he lae 1970s. Nor does i explain he run up in recen years. Page 2

3 Washingon Research Council CONSTRUCTION Consrucion largely explains hese deviaions. The level of consrucion aciviy has a powerful impac on sales ax revenues. The simple sory focuses on he fac ha he sae collecs sales ax on consrucion labor. (Washingon is one of he few saes o do so.) Thus a dollar of earnings in he consrucion secor is fully axed Char 2: Consrucion Earnings Share of Personal Income, 73:3 06:2 when i is earned as well as parially axed when i is spen. This 8.0% 7.5% suggess ha earnings from he consrucion secor should have a 7.0% larger impac on sales and use ax 6.5% revenue han oher earnings. Bu 6.0% he direc impac of consrucion 5.5% aciviy on sales and use ax revenues exends beyond consrucion 5.0% 4.5% labor. The sae collecs sales and 4.0% use axes on consrucion maerials as well as on complemenary 73:3 75:1 76:3 78:1 79:3 81:1 82:3 84:1 85:3 87:1 88:3 90:1 91:3 93:1 94:3 96:1 97:3 99:1 00:3 02:1 03:3 05:1 Source: Bureau of Economic Analysis furnishings such as furniure and appliances. Consrucion earnings can serve as a proxy for all of he axable aciviy associaed wih consrucion. Char 2 shows he raio of consrucion earnings o personal income (consrucion earnings share). The congruence beween he consrucion earnings share and he sales and use ax base share is sriking. Consrucion aciviy was very srong in he sae in he lae 1970s. Consrucion earnings, which had averaged 5.1 percen of personal income in 1974, peaked a 7.8 percen of personal income in he second quarer of 1979 and hen declined o 4.8 percen in Home consrucion is par of he sory. Housing permis hi a peak of 71,200 (annualized) in he fourh quarer of (Conras his wih he peak of he mos recen cycle, 57,700 in he fourh quarer of 2005, for an economy wih wice he employmen.). These were also he years when he Whoops nuclear plans were under consrucion. More recenly, consrucion earnings have grown from 4.9 percen Char 3: Sales and Use Tax Base Share of Personal Income, 73:03 06:2 Independen variables in fied equaion are personal income, consrucion earnings share and he unemploymen rae of personal income in he second 80% quarer of 2003 o 5.9 percen in he second quarer of % Tax Base Share The echnical appendix presens a 70% Fied Equaion model relaing he raio of he 65% sales and use ax base and personal income o hree variables: 60% personal income, he raio of consrucion earnings o personal 55% income, and he unemploymen 50% rae. We esimae he model saically and find ha i explains 96 73:3 75:2 77:1 78:4 80:3 82:2 84:1 85:4 87:3 89:2 91:1 92:4 94:3 96:2 98:1 99:4 01:3 03:2 05:1 percen of he variaion in he Tax Base is Office of he Forecas Council Consan Sale and Use Tax Base sales and use ax base share. Page 3

4 Washingon Research Council Char 4: Sales and Use Tax Base Share of Personal Income, 73:03 06:2 80% 75% 70% 65% 60% 55% 50% Page 4 As Char 3 shows, he model racks he spike in he ax base share during he lae 1970s and he collapse in he early 1980s. I picks up he increase in he lae 1980s, he decline in he 1990s, he recessionary dip and he curren upswing. 73:3 75:1 76:3 78:1 79:3 81:1 82:3 84:1 85:3 87:1 88:3 90:1 91:3 93:1 94:3 96:1 97:3 99:1 00:3 02:1 03:3 05:1 Tax Base is Office of he Forecas Council Consan Sales and Use Base To isolae he impac of consrucion on he sales and use ax base share, we use he esimaed coefficien for consrucion earnings share o calculae wha he sales and use ax base share would have been had he consrucion earning share remained fixed a 5 percen (is value in he fourh quarer of 2003) over he whole period. Tha hypoheical is graphed along wih he basic sales and use ax base series on Char 4. The verical disance beween he curves is he effec of deviaion of he consrucion earnings share from 5 percen. The consrucion earnings share sood a a bi more han 5.9 percen in he second quarer of Applying he curren 6.5 percen sae ax rae, he ax base difference shown on Char 4 from he firs quarer of 2004 hrough he second quarer of 2006 is worh $722 million. The difference in he second quarer of 2006 alone is worh $533 million when annualized. DISCUSSION Tax Base Share Wih Consrucion Earnings Fixed a 5% Consrucion aciviy has a powerful impac on sae revenues. And changes in sae policy ha reduce he level of consrucion aciviy will, herefore, have an impac on he general fund. Legislaors should keep his fac in mind as hey consider bills such as SB 5046, which, by increasing he liabiliy exposure of conracors, could reduce he level of residenial consrucion aciviy in he sae. The sae s deep fiscal crisis of he early 1980s semmed from he unwinding of a hisoric consrucion boom. This experience lef a widely held belief ha he heavy reliance on he sales ax makes Washingon s ax sysem more volaile han mos. This is no he case, as he las recession clearly demonsraed. The Washingon Sae Tax Srucure Sudy Commiee commonly known as he Gaes Commiee concluded: The sales ax, hough volaile, is less volaile han a graduaed personal income ax. There is no evidence ha a fla rae personal income ax would be less volaile han he sales ax (WSTSSC 2002, Vol. 1 pp ). The las few years have been kind o sae budge wriers, as upward revisions o he sae revenue forecas have followed one afer anoher wih numbing regulariy. Many have come o believe ha his is normal. This is a dangerous belief. While sae revenue forecass may have a conservaive bias, a large conribuor o he revisions was he failure by forecasers o anicipae he srengh of he consrucion in he sae. Consrucion has been a key driver of economic growh over he las several years, and as we have demonsraed, he level of

5 Page 5 Washingon Research Council consrucion aciviy has a powerful impac on sae revenues, even when he overall level of economic aciviy in he sae is held fixed. The share of consrucion earnings in personal income was percen in he second quarer of This is he highes level in 25 years, and could well mark he peak of his cycle. (The preliminary numbers released by he Bureau of Economic Analysis show he consrucion earnings share a percen in he hird quarer, bu given he ypical revisions o personal income and earnings numbers, we should no make oo much of his ye.) The sae s November Economic Forecas shows consrucion employmen as a share of oal employmen peaking in he firs quarer of From ha poin, consrucion s share of employmen declines coninuously hrough he end of he forecas period (he fourh quarer of 2009).Consequenly, he associaed November Revenue Forecas has revenue growh deceleraing significanly in he biennium from he pace se in (OFC 2006). The recenly released Preliminary February Economic Forecas conains an ominous piece of news: The number of housing unis auhorized by building permi plummeed 11,100 in he fourh quarer of 2006 o 42,200 from 53,300 in he hird quarer. Single-family permis declined 5,600 o 29,700 and muli-family permis fell 5,500 o 12,500. Toal permis in he fourh quarer were he lowes since he fourh quarer of 2003 and single-family permis were he lowes since he firs quarer of (OFC 2007) Economic and Revenue Forecas Council Execuive Direcor ChangMook Sohn noes ha his unanicipaed drop could be a sign of a significan down urn in housing, or, alernaely, i could jus be a emporary aberraion. The new economic forecas assumes ha he drop is a emporary aberraion, ha housing permis will rebound laer in 2007 so ha he relaive decline in consrucion will be genle. This assumpion of a rebound in permis will underpin he upcoming March Revenue Forecas. If permis do no rebound, ha revenue forecas will need o be adjused downward evenually. The las downswing in consrucion earnings share lased for nearly 6 years (from he second quarer of 1990 o he firs quarer of 1996) and he nex downswing could las well ino he biennium. The resul would be a long period over which sae revenues grew a less han heir long-erm rend. Legislaors should exercise cauion as hey wrie a budge for he biennium. Unlike he experience of recen years, above rend growh is unlikely o erase any srucural budge defici.

6 Page 6 Washingon Research Council APPENDIX: DETAILS OF THE ANALYSIS The sandard model used o esimae he relaionship beween ax base ( B ) and personal income ( Y ) is linear in he naural logarihm ( ln ) of he wo variables: ln( B ) = α + γ ln( ) + ε Y (1) Subscrips on variables indicae daes. (Our daa are quarerly.) The error erm in he equaion, ε, capures all he facors oher han personal income ha influence ax revenue. Wih his funcional form, he slope coefficien, γ direcly measures he "elasiciy" of revenue wih respec o personal income: A 1 percen increase in personal income resuls in a γ percen increase in ax revenue (Holcombe and Sobel 1997, Ch.5; WSTSSC 2002, Appendix C-21). Subracing he naural logarihm of personal income from boh sides of equaion (1) and applying he propery ha he difference beween he logarihms of wo numbers is he logarihm of heir raio yields he equaion: ln( B Y ) = α + β ln( Y ) + ε The consan α and he error erm ε in (2) are idenical o he consan and error in (1). The slopes β and γ are relaed by (2) β = γ 1 Leas squares esimaes of equaion (2) over he period 1973:3 o 2006:2 appear in column (a) of Table A-1 (on page 8). Fied values are ploed on Char 1 (page 2). The esimaed coefficien on he log of personal income is 0.105, implying ha he elasiciy of he sales and use ax base wih respec o personal income is The downward slope in he fied equaion on Char 1 manifess he fac ha he esimaed elasiciy is less han 1. Recall, however, ha he sales and use ax base we are using reflecs Washingon ax law circa 1973, a which ime food for off premises consumpion was subjec o ax. Holcombe and Sobel demonsrae ha including food in he sales ax base significanly lowers he elasiciy (Holcombe and Sobel 1997; see also Tannenwald 2004, p. 14). Naionally, he Bureau of Economic Analysis Naional Income and Produc Accouns show ha personal consumpion expendiures fell from 16.4 percen of personal income in 1975 o 11.7 percen of personal income in The curren sae sales and use ax would show a somewha greaer personal income elasiciy. Nex we expand equaion (2) o include he raio of earnings in he consrucion secor ( E C ) o personal income and he unemploymen rae ( U ) as explanaory variables: ln( B Y ) C = + β 1 ln( Y ) + β 2 ln( E Y ) + β 3 α U + ε (3) The resuls of esimaing his equaion for he 1973:3 o 2006:2 period are shown in column (b) of Table A-1, wih fied values ploed on Char 3 (page 3). Wih hese wo variables included, he r 2 is considerably higher (0.961 vs ) and he elasiciy wih respec o personal income is somewha higher (0.991 vs ).

7 Page 7 Washingon Research Council Six addiional regressions are repored in Table 1. Columns (c) and (d) show he effec of omiing eiher consrucion earnings share or he unemploymen rae from equaion (2). Inclusion of he unemploymen rae reduces he personal income s coefficien (compare [d] o [a] and [b] o [c]), while i also reduces consrucion earnings coefficien (compare [b] o [c]). Adding he consrucion earnings share raises personal income s coefficien (compare [b] o [d] and [c] o [a]). Columns (e), (f) and (g) repea he specificaions in (a), (b) and (c), bu esimae over a ime period (1982:3 o 2006:2) ha excludes he big spike in he consrucion earnings share of personal income. Coefficien esimaes (f) and (g) are no ha much differen from (b) and (c), indicaing ha he resuls for he impac of consrucion on sales and use ax revenue are no driven solely by he 1973:3 o 1982:2 daa. Noe, however, ha in he case of he simple specificaion where he ax base share is a funcion of personal income alone, he ime period does maer. The implied elasiciy of he sales and use ax base in column (e) is 0.940, as compared o for column (a). The lower elasiciy in he laer case appears o be an arifac of he consrucion boom of he lae 1970s. Finally, column (g) shows he resul of including an auo regressive error in he specificaion. Again, he effec on he coefficien esimaes is no large. On Char 4, he curve labeled Wih Consrucion Earnings Fixed a 5% plos he equaion Z = ( B Y ) b ( Y E ) b C where b is he coefficien on he naural log of consrucion earnings from regression (b). This adjuss sales and use ax base as a share of personal income o a consan 5 percen consrucion earnings share of personal income.

8 Page 8 Washingon Research Council Table A-1: Regression Resuls The Dependen Variable is he Naural Log of he Sales and Use Tax Base Div ided By Personal Income (Sandard Errors of Coefficiens in Parenheses) (a) (b) (c) (d) (e) (f) (g) (h) Consan ( ) ( ) ( ) ( ) ( ) ( ) ( ) Naural Log of Personal Income Naural Log of Consrucion Earnings Share ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) Unemploy men Rae ( ) ( ) ( ) ( ) AR(1) ( ) Adjused r-squared Sample 73:3-06:2 73:3-06:2 73:3-06:2 73:3-06:2 82:3-06:2 82:3-06:2 82:3-06:2 73:4-06:2 REFERENCES Holcombe, Randall G. and Russell S. Sobel Growh and Variabiliy in Sae Tax Revenue: An Anaomy of Sae Fiscal Crises. Wespor, CT: Greenwood Press. Economic and Revenue Forecas Council (OFC) Washingon Economic and Revenue Forecas. Vol. XXIX, No. 4. November Preliminary March Economic Forecas. February 26. Tannenwald. Rober Are Sae and Local Revenue Sysems Becoming Obsolee? Research Repor on America s Ciies. Naional League of Ciies and The Brookings Insiuion Meropolian Policy Program. Washingon Research Council (WRC) Governor Gregoire Proposes Budge wih $1.2 Billion Srucural Defici. PB January 3. Washingon Sae Tax Srucure Sudy Commiee (WSTSSC) Tax Alernaives for Washingon Sae: A Repor o he Sae Legislaure. Olympia, WA: Deparmen of Revenue. To receive advance noice of Washingon Research Council publicaions by send your address o