Skip to content

The political economy of institutions, stability and investment: a simultaneous equation approach in an emerging economy – the case of South Africa

Literature › Works

J.W. Fedderke; J.M. Luiz (2008). The political economy of institutions, stability and investment: a simultaneous equation approach in an emerging economy – the case of South Africa. University of Cape Town / University of the Witwatersrand (working paper).
Type working paper
Scope Economics and economic history of apartheid
Method econometric
Period studied 1954–1992
Regions South Africa
Themes macro fiscal finance debt, politics ideology elections, business and state, comparative theory, methods data sources, origins segregation
Citation key fedderke2008

Argument

Prior South African investment studies treat political instability as a measure of exogenous uncertainty. The authors argue instability is instead systematically determined by underlying factors. They set three accounts against each other: modernization (output), primacy of the political (political rights) and primacy of institutions (property rights, following North). Each is tested with Johansen VECM estimation of a joint investment-instability system using strong, weak and mixed exogeneity restrictions. The property-rights account performs best: property rights and output reduce instability, crime raises it, and instability, property rights and output all affect investment. Uncertainty therefore attaches to a wider institutional nexus rather than to instability alone.

Evidence and method

Annual South African time series 1954-1992, with a small sample that the authors say limits test power. Variables are the investment rate (net change in fixed capital stock), log real GDP, user cost of capital, a logged political instability index (weighted average of eleven repression/opposition measures), property rights and political rights indices, and criminal cases per capita. The indices come from Fedderke, De Kadt and Luiz (2001a). Tests are ADF unit-root tests and Johansen cointegration tests.

Place in the debates

Findings

  • Political instability is endogenously determined by property rights, output and crime; this reframes uncertainty as institutional. (p. 2)

    Political instability is no longer exogenous, but subject to endogenous structural determination. Uncertainty does not dissipate – but attaches to a constellation of institutions.

  • Prior South African studies consistently find a negative investment effect from uncertainty measures, usually proxied by political instability. (p. 5)

    the finding of a negative impact on the variables interpreted as uncertainty measures is pervasive.

  • The sample is short and constrained by data availability, which weakens statistical test power. (p. 10)

    This provides us with a data sample covering the period from 1954 to 1992. Sample size is restricted by data availability.

  • The strong tests are inconclusive, with implausibly large elasticities, such as property rights on instability (66.58), and weak error correction. (p. 14)

    while the implied mean elasticity associated with political rights is perhaps plausible (0.4), that associated with property rights is too dramatic (66.58) to be credible.

  • The weak tests also fail, with implausible elasticities and no error correction in the instability relation. (p. 14)

    the elasticities particularly on output in both the investment and the instability vectors are implausibly large (approximately 20 and 15 respectively)

  • The mixed test removes the implausible elasticities only under the modernization or property-rights exogeneity restrictions. (p. 17)

    the implausibly large elasticities found under the weak and some strong test specifications disappear, as soon as the weak exogeneity restriction of the institutional/property rights primacy or the modernization hypothesis cases are imposed.

  • The modernization specification fits but has almost no explanatory power for instability, failing to explain the 1990s decline. (p. 17)

    the observed declines in the political instability measure during the 1990’s for South Africa would simply not be attributable to the identified explanatory variables of this specification.

  • Only the property-rights specification performs satisfactorily; it gives standard investment-function results. (p. 18)

    Output growth acts as a positive trigger for investment flows (elasticity of 6), the user cost of capita is an investment deterrent (elasticity of 0.5), while improving property rights raise investment expenditure (elasticity of 3.2).

  • Property rights both dampen instability and raise investment, while output dampens and crime raises instability; support is for weak primacy only. (p. 18)

    Property rights prove to be important not only in dampening political instability (elasticity of 2.5), but in stimulating investment (elasticity of 3.2).

  • Political-rights primacy gets little support; property rights are not the sole determinant of instability. (p. 19)

    Our empirical results find support for both the modernization process, and the primacy of property rights, but relatively little support for the political primacy explanation.

Evidence ledger

Sources used

Source Kind Period How used
Fedderke, De Kadt and Luiz (2001a) index of political instability (INSTAB), weighted average of eleven measures of political repression and opposition secondary literature compilation 1935-1997 (sample used 1954-1992) Serves as the dependent variable in the instability equation and as the uncertainty proxy in the investment equation.
Fedderke, De Kadt and Luiz (2001a) property rights index (PROPRIGHT) secondary literature compilation 1954-1992 sample Measures formal property rights and is the key institutional regressor in the investment and instability relations.
Fedderke, De Kadt and Luiz (2001a) political rights index (POLRIGHT) secondary literature compilation 1954-1992 sample Proxy for political freedom, used to test the 'primacy of the political' explanation of instability.
Crime rate series CRIMPOP (criminal cases prosecuted per unit of population), from Fedderke, De Kadt and Luiz (2001a) secondary literature compilation 1954-1992 sample Crime variable in the instability equation, used as a measure of rule enforcement.
Investment rate series INVRAT (net change in fixed capital stock of machinery and equipment) national accounts macro 1954-1992 Dependent variable in the investment equation.
Real GDP series (LNRGDP) national accounts macro 1954-1992 Proxy for expected returns on capital and for economic development in the modernization test.
User cost of capital series (USER) official statistical series 1954-1992 Marginal cost of capital regressor in the investment relation.
Henderson (2004) MComm dissertation, compilation of prior findings on South African investment studies secondary literature compilation n/a Table 1 summarizing signs found in prior investment studies is adapted from it.

Inference

Quantification: Econometric (causal design).

Johansen VECM cointegration estimation on annual South African time series (1954-1992, about 39 observations). Over-identifying restrictions on cointegrating vectors and loading matrices implement strong, weak (weak exogeneity) and mixed tests of three theories of political instability (modernization, political primacy, property-rights primacy). Theories are compared by the plausibility of elasticities, signs and error-correction behaviour.

Key claims

Claim Basis Strength
Political instability measures matter for aggregate investment, and output is positively and user cost negatively related to investment. (p. 19) VECM cointegrating vectors on INVRAT, INSTAB, LNRGDP and USER (Tables 5-8). Partial data
Political instability is systematically related to property rights, crime and output rather than being exogenous uncertainty. (p. 3) Instability cointegrating vector in the mixed-test VECM (Tables 7-8) using INSTAB, PROPRIGHT, CRIMPOP and LNRGDP. Partial data
The property-rights-primacy specification performs best: it gives plausible elasticities, expected signs and genuine error correction. Property rights dampen instability (about 2.5) and raise investment (about 3.2). (p. 18) Mixed test, column (3) and (5)-(6) of Tables 7 and 8, with imputed mean elasticities. Partial data
The political primacy hypothesis receives little support: no explanatory power and implausibly large output elasticities. (p. 18) Strong, weak and mixed tests imposing weak exogeneity of POLRIGHT (Tables 5-8, col. 4 of Tables 7-8). Partial data
The modernization specification loses economic power: the instability relation dissipates and cannot account for the 1990s decline in instability. (p. 17) Column (2) of Tables 7-8 on elasticities; comparison with the observed instability path. Partial data
Rising crime raises political instability, and rising output dampens it. (p. 18) Elasticities on CRIMPOP and LNRGDP in the instability vector (Tables 7-8). Partial data
Strong and weak tests give implausibly large elasticities (e.g. property rights 66.58; output about 20) and poor error correction, indicating underspecification. (p. 14) Tables 5 and 6 VECM estimates and implied elasticities. Systematic data

The author's caveats

The sample (1954-1992) is restricted by data availability, and the small sample reduces the power of the statistical tests. Some variables (especially CRIMPOP) are likely partially integrated, so cointegration rank tests may overstate the number of vectors. The authors also acknowledge parameter instability and implausibly large elasticities in several specifications.

Gaps in the evidence

The data rest on a handful of subjective, author-constructed indices (political rights, property rights, instability) from a prior study, with no validation against independent measures such as archival records of strikes, protest and repression, or property-ownership data. Only about 39 annual observations are used to estimate a seven-variable VECM with two assumed cointegrating vectors, which is thin for identification. Sectoral or firm-level investment data, which existed for the period and were used in related work, along with disaggregated crime data, could test the argument better and address causality. Data sources for investment, GDP and user cost are not specified.

Replicability: Data poorly documented.