Public Spending and Growth

by Patrick Minford and Jiang Wang

TWO VIEWS

• Endogenous growth, public spending and taxes: two views

• R&D, education and infrastructure spending by the state creates new factor input; hence growth. ‘ACTIVIST’

• Incentives generate risk-taking innovation by people/firms. Lower taxes raise incentives; hence growth. ‘INCENTIVIST’

ECONOMETRIC TESTS

• INCENTIVIST: Growth negatively related to the marginal tax rate on innovation.

• ACTIVIST: Growth positively related to state support of R&D and subsidy of the return on capital.

• Data: growth rate over three decades (1970s, 1980s, and 1990s) for 100 countries; 300 panel observations (some missing).

The negative impact of taxation on economic growth

Author Data coverage Main explanatory variables Comment

Barro (1991) 98 countries in the period 1960-1985

Human capital, govt. consumption, political instability indicator, price distortion

1% point of GDP increase in tax to GDP ratio lowers output per worker by 0.12%.

Koester and Kormendi (1989)

63 countries for which at least five years of continuous data exists for the 1970s.

Marginal tax rates average tax rate, mean growth in labour force & population

10% decrease in marginal tax rates would increase per capita income in an average industrial country by more than 7%.

Hansson and Henrekson (1994)

Industry-level data for 14 OECD countries

Govt. transfers, consumption, total outlays; education expenditure; govt. investment

Govt. transfers, con-sumption and total outlays have a negative impact on growth whilst government investment is not significant

Cashin (1995) 23 OECD countries over the 1971-1988 period.

Ratio of public in-vestment to GDP, ratio of current taxa-tion revenue to GDP, ratio of expenditure on transfers to GDP.

1% point of GDP increase in tax to GDP ratio lowers output per worker by 2%.

Engen & Skinner (1996)

US modelling together with a sample of OECD countries.

Marginal tax rates, human capital, investment.

2.5% point increase in tax to GDP ratio reduces GDP growth by 0.2% to 0.3%.

The negative impact of taxation on economic growth

Author Data coverage Main explanatory variables Comment

OECD - Leibfritz, Thornton & Bibbee (1997)

OECD countries over the 1965-1995 period.

Tax-to-GDP ratio, physical and human capital formation and labour supply.

10% point increase in tax to GDP ratio reduces GDP growth by 0.5% to 1%.

Alesina et al. (2002) 18 OECD countries over the 1960-1996 period.

primary spending, transfers, labor taxes, taxes on business, indirect taxes, govt. wage consumption (all in share of GDP).

1% increase in govt. spend-ing relative to GDP lowers the investment-to-GDP ratio of 0.15% and a cumulative fall of 0.74% after 5 years.

Bleaney, Gemmell & Kneller (2000)

17 OECD countries over the 1970-1994 period.

distortionary tax, pro-ductive expenditure, net lending, labour force growth, invest-ment ratio.

1% point of GDP increase in distortionary tax revenue reduces GDP growth by 0.4% points.

Folster & Henrekson (2000)

Sample of rich OECD / non-OECD countries over the 1970-1995 period.

Tax-to-GDP, govt. expenditure-to-GDP, investment-to-GDP, labour force growth, human capital growth

10% point increase in tax to GDP ratio reduces GDP growth by 1%.

Bassanini & Scarpetta (2001)

21 OECD countries over the 1971-1998 period.

Indicators of government size and financing, physical capital, human capital, population growth.

1% point increase in tax/GDP ratio reduces per capita output levels by 0.3% to 0.6%.

Below each parameter estimate, we report its robust standard error. Three asterisks denote statistical significance at the 1% level, two asterisks at the 5% level and one asterisk at the 10% level.lnGadj = ln(G+0.04) due to the possible negative value of G for some countries. Observations with growth less than –0.04 (-4% per annum) are omitted as outliers with special reasons (such as civil wars etc)

Dependent var is lnGadj

Cross-sectional regression

Random country effect (with fixed time effect)

Fixed country and time effects

lnr -0.4131001***0.0660606

-0.4089624***0.0675931

-0.27778890.2114343

constant -3.598126***0.119624

-3.665844***0.1186308

-3.47297***0.2984599

R2 0.1383 0.1596 0.1552

observations 295 295 295

Table 1

Figure1: Correlation between growth rate in real GDP per capita and tax rate

r

G Fitted values

.066607 1.4168

-.091544

.113574

lnr

lnGadj Fitted values

-2.70894 .348397

-8.70034

-1.87357

Table 2: Initial value of GDP per capita ― lnG0 Dependent var is lnGadj

Cross-sectional regression

Random country effect (with fixed time effect)

Fixed country and time effects

lnr -0.4585302***0.077744

-0.450433***0.0914043

-0.31334330.2046511

lnG0 -0.03885170.0463611

-0.03882470.0544193

-0.66156***0.175467

constant -3.343886***0.3531412

-3.200959***0.3710283

2.0365311.429603

R2 0.1403 0.1596 0.0177observations 295 295 295

Table 3: Initial value of human capital variable ― lnhc0 Dependent var is lnGadj

Cross-sectional regression

Random country effect (with fixed time effect)

Fixed country and time effects

lnr -0.3359003***0.0635351

-0.2824532***0.1006879

-0.15139850.1633693

lnhc0 0.0982599*0.0585831

0.191008**0.0859461

0.05577510.1579874

constant -3.605208***0.1316414

-3.506455***0.1541899

-3.141061***0.3213641

R2 0.1533 0.1874 0.1613observations 295 295 295

Table 4: Initial value of physical capital variable ― lnpc0

Dependent var is lnGadj

Cross-sectional regression

Random country effect (with fixed time effect)

Fixed country and time effects

lnr -0.2284145***0.0737142

-0.2201459**0.098377

-0.29422350.2092666

lnpc0 0.01257490.0456652

0.03101930.0414442

-0.01115470.1666652

constant -3.401858***0.3538077

-3.390082***0.30208

-3.147642**1.469304

R2 0.0676 0.1166 0.1018observations 295 295 295

Table 5: Ratio of investment/GDP ― lninvest

Dependent var is lnGadj

Cross-sectional regression

Random country effect (with fixed time effect)

Fixed country and time effects

lnr -0.1853692***0.0658581

-0.2010656***0.0806375

-0.2282440.2120415

lninvest 0.3552623***0.0785507

0.329507***0.0773946

0.2688975*0.1501412

constant -4.227441***0.2126837

-4.076769***0.1814519

-3.944269***0.4867979

R2 0.2060 0.2183 0.2167observations 295 295 295

Table 6: Adding lnG0, lnhc0, lnpc0 and lninvest Dependent var is lnGadj

Cross-sectional regression

Random country effect (with fixed time effect)

Fixed country and time effects

lnr -0.1906314***0.0626646

-0.1770296*0.0963501

-0.03103330.1550151

lnG0 -0.1534513**0.0782778

-0.274614***0.0921151

-1.328827***0.1914185

lnhc0 0.10606420.1085019

0.1394630.1128301

-0.05234430.1963471

lnpc0 -0.04259980.0595577

0.03381310.0710202

0.5468776***0.1519144

lninvest 0.3883018***0.0975898

0.371204***0.1254544

0.4716634***0.1684907

constant -2.790863***0.4433326

-2.349998***0.4931368

2.30601*1.386747

R2 0.1769 0.1859 0.0070

observations 295 295 295

Table 7: Results for the activist approach Dependent var is lnGadj

Cross-sectional regression

Random country effect (with fixed time effect)

Fixed country and time effects

KSUB -0.03288740.0281843

-0.0428637**0.0192169

-0.0651416**0.026076

HSUB -0.0172619**0.0080792

-0.00742870.007605

-0.00165910.010384

constant 0.0291336***0.0047852

0.0268741***0.0058173

0.0229654***0.0069056

R2 0.0567 0.0454 0.0164observations 122 122 122

Figure 2 Correlation between GDP growth rate and subsidy rate to investment

KSUB

G Fitted values

-.638728 .628834

-.091544

.080092

Figure 3 Correlation between GDP growth rate and subsidy rate to R&D

HSUB

G Fitted values

.029 1

-.039833

.113574

Table 8: Initial value of GDP per capita ― G0 Dependent var is lnGadj

Cross-sectional regression

Random country effect (with fixed time effect)

Fixed country and time effects

KSUB -0.03367050.0280481

-0.0439847**0.0194498

-0.050478*0.0278378

HSUB -0.01505290.0100392

-0.006690.0078479

0.00441650.0104769

G0 1.58e-073.25e-07

1.03e-073.65e-07

-1.36e-069.57e-07

constant 0.0263072***0.0090049

0.0257932***0.0067726

0.0308422***0.0088099

R2 0.0590 0.1510 0.0048

observations 122 122 122

Table 9: Initial value of human capital variable ― hc0

Dependent var is lnGadj

Cross-sectional regression

Random country effect (with fixed time effect)

Fixed country and time effects

KSUB -0.03213320.0260735

-0.0446653*0.0189903

-0.0621612**0.0252276

HSUB -0.00778160.0093707

-0.00107780.0079445

-0.00016380.0105279

hc0 0.0017704**0.0008503

0.0024214***0.0009453

-0.0035290.0029477

constant 0.01206230.009843

0.00912070.0082923

0.0438552**0.0186494

R2 0.0907 0.0991 0.0252observations 115 115 115

Table 10: Initial value of physical capital variable ― pc0

Dependent var is lnGadj

Cross-sectional regression

Random country effect (with fixed time effect)

Fixed country and time effects

KSUB -0.0739059**0.0345323

-0.0763963***0.00223675

-0.0492111*0.0258339

HSUB -0.0227255**0.0115609

-0.0163448*0.0090925

0.01995140.0158965

pc0 -3.46e-07**1.59e-07

-3.67e-071.61e-07

-1.00e-06**4.6e-07

constant 0.0417207***0.0097311

0.0401149***0.00743

0.0275565***0.0104124

R2 0.1684 0.2006 0.0282observations 91 91 91

Table 11: Ratio of investment/GDP ― invest Dependent var is lnGadj

Cross-sectional regression

Random country effect (with fixed time effect)

Fixed country and time effects

KSUB -0.02987950.0222218

-0.0398512**0.0170956

-0.064688***0.0249827

HSUB -0.00019290.0077747

0.00166560.0069521

0.00480420.0100272

invest 0.1299014***0.0192989

0.1271213***0.0217512

0.1330205***0.0530812

constant -0.0077780.0073933

-0.00664790.0077625

-0.01030570.0148337

R2 0.3676 0.3693 0.3524observations 122 122 122

Table 12: Adding G0, hc0, pc0 and invest Dependent var is lnGadj

Cross-sectional regression

Random country effect (with fixed time effect)

Fixed country and time effects

KSUB -0.05964850.0382799

-0.0678567***0.0209594

-0.04005690.0258894

HSUB -0.01143240.0103655

-0.00558150.0090662

0.01881480.0153496

G0 -2.56e-075.12e-07

-5.18e-076.29e-07

-2.98e-06*1.58e-06

hc0 0.00061720.0011025

0.00031540.0012224

-0.00169110.0029981

pc0 -4.22e-07***1.34e-07

-3.78e-072.61e-07

-1.39e-086.35e-07

invest 0.1003211***0.0265871

0.1270069***0.0300039

0.1134434*0.0596572

constant 0.01286210.0123803

0.00453070.01179

0.02302410.0280673

R2 0.3001 0.3406 0.0699

observations 89 89 89

CONCLUSIONS

• Tax elasticity of -0.4; meaning for initial tax rate of 40% and growth rate of 2.5%. Fall in marginal tax of 10% (ie to 36%) raises growth by 0.1% to 2.6%.

• No evidence in this data of effects of state spending on R&D or on subsidy to investment.

• Further work on structural models required.