Skip to content

Wagner's Law: An Econometric Test for South Africa, 1960-2006

Literature › Works

Emmanuel Ziramba (2008). Wagner's Law: An Econometric Test for South Africa, 1960-2006. South African Journal of Economics, Vol. 76:4. Blackwell Publishing / Economic Society of South Africa.
Type journal article
Scope Comparative and methods
Method econometric
Period studied 1960–2006
Regions South Africa
Themes macro fiscal finance debt, methods data sources, comparative theory
Citation key ziramba2008

Argument

The paper tests Wagner's law, the claim that government expenditure grows because national income grows, for South Africa over 1960-2006. It uses the Pesaran-Shin-Smith ARDL bounds test for cointegration and the Toda-Yamamoto procedure for Granger causality. The author finds a long-run relationship between real per capita government expenditure and real per capita income, but causality runs in both directions. Wagner's law requires unidirectional causality from income to expenditure, so the paper tentatively concludes that it finds no support. The relationship is described as neither purely Keynesian nor Wagnerian.

Evidence and method

Annual data for 1960-2006 (47 observations). Income data come from the South African Reserve Bank. Government expenditure and population come from IMF International Financial Statistics, deflated with the GDP deflator (2000=100). Table 1 uses SARB Quarterly Bulletin data on spending components. The paper reports Ng-Perron unit root tests, a bounds-test Wald statistic and Toda-Yamamoto causality tests.

Findings

  • Government size grew in real terms and relative to GDP, peaking in 1976. (p. 2)

    Real government resource use increased from R48.2 billion in 1960 to R223.4 billion in 2003. Over the same period resource use increased from 16.6% of GDP to 22.2%.****

  • Government resource use peaked at 26.5% of GDP in 1976, and growth was not uniform over the period. (p. 2)

    resource use reaching a peak of 26.5% of GDP in 1976

  • Social protection's share of spending doubled between 1983 and 2005, with the largest increase in 1994, which the author links to the first democratic government's priorities. (p. 2)

    Between 1983 and 2005 government expenditure on social protection doubled its share in total expenditure, with the major increase being recorded in 1994.

  • Defence's share of spending fell from 14.17% in 1983 to about 5% by the late 1990s (Table 1). (p. 3)

    1983 22.63 6.16 17.72 14.17

  • Both variables are non-stationary in levels and stationary in first differences, i.e. I(1). (p. 8)

    The unit root test results show that both variables are integrated of order one, I (1).

  • The bounds test rejects no cointegration (Wald statistic 8.184, p=0.017), indicating a long-run relationship. (p. 8)

    Ln(G/N) = d0 + d1ln(Y/N) + e3t 8.184 0.017

  • Granger causality runs in both directions between income and government spending. (p. 8)

    The results suggest that the null hypothesis of no Granger causality can be rejected in both directions. Thus, there is evidence of bidirectional causality.

  • Causality test statistics are 2.767 (p=0.096) for G→Y and 3.928 (p=0.048) for Y→G, so the G→Y result is significant only at the 10% level. (p. 9)

    Ln(G/N) Ln(Y/N) 2.767 0.096 G→Y Ln(Y/N) Ln(G/N) 3.928 0.048 Y→G

  • The overall conclusion is that Wagner's law finds no support and the relationship is neither Keynesian nor Wagnerian. (p. 9)

    the relationship between government expenditure and national income is neither Keynesian (G→Y) nor Wagnerian (Y→G).