Skip to content

Federal Institutions and the Democratic Transition: Learning from South Africa

Literature › Works

Robert P. Inman; Daniel L. Rubinfeld (2008). Federal Institutions and the Democratic Transition: Learning from South Africa. NBER Working Paper Series, No. 13733. National Bureau of Economic Research.
Type working paper
Scope Economics and economic history of apartheid
Method mixed
Period studied 1950–2006
Regions South Africa, Western Cape, KwaZulu-Natal
Themes transition post apartheid, politics ideology elections, macro fiscal finance debt, sanctions divestment, unions strikes, comparative theory, inequality poverty unemployment
Citation key inman2008

Argument

The paper builds a game-theoretic model of how a white elite and a Black majority could agree a peaceful transition to democracy. Under a federal constitution the elite-run Western Cape provides redistributive services (education, health, welfare) cheaply, while the majority-controlled centre sets the elite tax rate. This creates a "hostage" game with grim-trigger strategies, and in equilibrium the tax rate stays below the revenue-maximising rate (Proposition 1). The model is calibrated to 1994-2006 fiscal data. The authors then estimate a growth dividend from the end of sanctions and union pressure, and conclude that both elite and majority gain from democracy, federal or unitary, so that federalism mainly supplied the credible ex ante commitment.

Evidence and method

Calibration of a stylised model with figures on intergovernmental transfers (FFC reports, Division of Revenue Bills 1995/96-2005/06), demographic and income estimates (elite 9.6m adults, majority 25m), teacher training years, homeland and security spending (DBSA, Dept of Information), Ugandan school capture estimates (Reinikka and Svensson), election results, and a small time-series regression on Penn World Tables 6.1 growth, openness and investment 1950-2000. Qualitative support from Waldmeir (1997), Steytler and Mettler (2001) and Muthien and Khosa (1998).

Place in the debates

Findings

  • Calibrated economic sizes of the elite and majority populations and incomes at the time of transition. (p. 28)

    The initial, pre-democracy elite and majority voting age populations were N0 = 9.6 million and M = 25 million residents.

  • Growth regression: sanctions and COSATU each reduce growth, and growth in the period of both is negative. (p. 36)

    From 1985 to 1993, when both sanctions and strong unions were in force , the annual growth of the apartheid economy became negative, -1.222 percent per annum

  • Sanctions and strong unions operated through trade openness and domestic investment. (p. 36)

    Sanctions had their primary effect through restricted trade (col. 3), while strong unions limited both trade (col. 3) and the rate of domestic investment (col. 4) during apartheid.

  • Majority residents gain from either form of democracy, federal slightly more. (p. 37)

    The lifetime welfare gain for a young majority resident is 596,000 Rand ( $99,300) by moving to a federal democracy and 576,000 Rand ($96,000) by moving to a unitary democracy.

  • Young elite residents also gain, with a high internal rate of return, because the growth dividend outweighs higher taxes. (p. 38)

    Elite residents with a real rate of time preference less than .122 (* $ .89) prefer democracy to apartheid.

  • Mandela-era redistribution used only part of the redistributive potential of democratic federalism. (p. 33)

    President Mandela chose to use only 2/3's of the redistributive potential available within democratic federalism.

  • The Mbeki-era budgets pushed toward the limits of sustainable federalism; the q*-regime would be unsustainable. (p. 26)

    In a q* regime, South Africa’s federal democracy is no longer sustainable.

  • Estimated annual tax cost of apartheid's homeland and security spending per elite adult. (p. 35)

    Total expenditures are therefore 60 Billion Rand, which when allocated over the 9.6 million elite adults, averages to 6,250Rand per elite adult resident.

  • Authors' judgement that federalism may not have been strictly necessary, though it gave a credible promise. (p. 38)

    we see with 20-20 hindsight that they may not have been necessary to the original democratic agreement. (Unitary democracy appears nearly as beneficial.)

  • Transition required a growth dividend to make all veto players better off. (p. 39)

    It is the growth dividend that has proven so valuable for South Africans.

Evidence ledger

Sources used

Source Kind Period How used
Financial and Fiscal Commission (FFC), The Allocation of Financial Resources Between the National and Provincial Governments, FY 1996/97 and FY 1997/98 (Tables 2, 3, 6b) commission or official report FY 1995/96-1997/98 Provides provincial intergovernmental transfer data (Table 2) and the service targets (teacher ratios, clinic visits, social grants) used to compute the mandated service level q = .53.
Minister of Finance, Division of Revenue Bill, Various Years, Part 4: Provincial Allocations official statistical series FY 1998/99-2005/06 Supplies per-capita grants for the Western Cape and majority provinces in Table 2, used to infer the Mbeki-era q levels and test sustainability.
Penn World Tables 6.1 (GRGDPCH, OPENK, KI) international trade or finance data 1950-2000 Annual real GDP per capita growth, openness and investment series for the Table 5 regressions estimating the growth effects of sanctions, COSATU and democracy.
Development Bank of South Africa, Annual Report, Various Years official statistical series last years of apartheid (1977-1993) Homeland budget averages (26.3 bn 2000 Rand) used to estimate apartheid-era redistributive services (q = .16) and apartheid tax costs to the elite.
South African Department of Information, Perskor, South Africa, Various Years official statistical series 1977-1993 Average military, police and justice spending (33.7 bn Rand) used to compute the apartheid security tax burden on elite adults.
South African Reserve Bank Time Series KBP6244J official statistical series post-2000 Used to estimate a 2.5 percent annual growth rate after 2000 for projecting future income growth.
2004 Western Cape election results (www.elections.org.za) election returns 2004 Vote shares (ANC 45%, elite parties 51%) used to set the elite province's majority-resident share, so that λ = .184.
Constitution of South Africa (1996) and Interim Constitution (1994) legislation gazette 1994-1996 Provisions on housing, health, education and the provincial assignment of services used to define the binding service constraint (q-regime) and the federal structure.
Reinikka and Svensson estimates of capture in Ugandan local schools secondary literature compilation 1990s Uganda Source of the elite capture parameters nL = .20 and nH = .85 used in the calibration.
Author's Data Appendix for Table 3 (demographics, incomes, teacher training years, salaries, m, Z) author survey or fieldwork c. 1994-2001 Underlies the calibration of the elite and majority populations, incomes, teacher training years, the common salary S and the trained-provider share m.
South African ten-year bond real returns 1996-2006 stock market financial data 1996-2006 Sets the discount factor δ = .97 (r = .03) for the welfare calculations.
Waldmeir (1997), Anatomy of a Miracle, and Muthien and Khosa (1998) secondary literature compilation 1990-1996 Qualitative accounts of the negotiations, the NP's secession proposal, provincial boundary-drawing and the payments to the IFP/Zulu nation (Z).
Financial Times, September 21, 2007, p. 4, 'Boost for Zuma's Leadership Campaign' newspapers periodicals 2007 Cited as evidence of mounting ANC rank-and-file pressure for more redistribution, bearing on whether the q or q* regime applies under Mbeki.

New data

The authors built a calibration dataset for the South African political economy (Table 3) and a spliced per-capita provincial transfer series for 1995/96-2005/06 (Table 2). They also constructed a regime-indicator growth dataset for 1950-2000 from PWT 6.1 (Table 5). Details are in a Data Appendix 'available upon request'.

Inference

Quantification: Econometric (causal design).

Primarily a formal game-theoretic model (a repeated policy game with grim-trigger strategies) calibrated to South Africa using parameters from various sources. It is supplemented by a time-series regression of annual growth on regime dummies and a present-value welfare simulation of apartheid versus democracy.

Key claims

Claim Basis Strength
Federal fiscal institutions give a self-enforcing, credible protection against majority exploitation of the elite, sustaining redistributive taxation below the revenue peak (Proposition 1). (p. 25) Theoretical proof, plus calibration to the 1994-2006 provincial transfer data (Tables 2-4). Partial data
Mandela-era budgets (q = .53) satisfy the credibility constraints and used only about two-thirds of the redistributive potential of federalism. (p. 33) FFC/Division of Revenue transfer data (Table 2) compared with model-predicted bounds (Table 4) from calibrated parameters. Partial data
Mbeki's rising service mandates (q = .71 to .77) push federalism to the limits of sustainability, and a q*-regime would make it unsustainable. (p. 32) Table 2 provincial spending converted to q, compared with the Figure 1-2 constraint bounds. The result is sensitive to the calibrated D (1,720 vs 1,550 Rand). Partial data
Sanctions and COSATU union power substantially lowered apartheid-era growth, and democracy restored positive growth, creating a growth dividend. (p. 36) Table 5 OLS regressions on Penn World Tables 1950-2000 with regime dummies (SANC, COSATU, DEM). Partial data
Both majority and elite residents gain economically from the democratic transition, federal or unitary, despite much higher elite taxes; young elite internal rate of return is about .12. (p. 37) Present-value simulation (Table 6) from calibrated incomes, taxes, growth rates (Table 5) and a 70-year horizon. Partial data
The elite's human-capital advantage in providing public services (teacher training years 17 vs 14 vs 7) gives it the bargaining asset underpinning federalism. (p. 28) Teacher years-of-schooling used as productivity proxy (Data Appendix, not shown); salary and cost assumptions. Illustrative or anecdotal
Federalism was essential to the NP's willingness to accept democracy and to the peaceful transition. (p. 26) Waldmeir (1997), Steytler and Mettler (2001) accounts of the negotiations and Mandela/de Klerk quotes; no systematic data. Asserted

The author's caveats

The authors admit that no South African data were available to estimate the elite's exit response to taxation, so β is calibrated to a plausible revenue peak and checked for robustness up to β = .00025. They treat λ = 4123 as a lower bound because recent budgets may have been constrained. They note homeland spending is a generous estimate of services delivered, and sensitivity to D is tested within one standard deviation.

Gaps in the evidence

Key parameters (capture rates from Uganda, the elite exit elasticity, protest costs D inferred from a union growth coefficient) are borrowed or calibrated rather than measured for South Africa. Direct data on actual provincial leakage or capture in the Western Cape, on elite emigration and tax responses, and on negotiation-era archival evidence (NP and ANC positions, Kempton Park records) exist and could test the mechanism. The growth regressions use a short national series with overlapping regime dummies and no controls, so they cannot separate causes; richer Reserve Bank and sectoral data could improve this.

Replicability: Data described but not public.