South African Agricultural Production, Productivity and Research Performance in the 20th Century¶
Frikkie Liebenberg (2012). South African Agricultural Production, Productivity and Research Performance in the 20th Century. PhD thesis, Department of Agricultural Economics, Extension and Rural Development, University of Pretoria. University of Pretoria.
| Type | thesis |
| Scope | Economics and economic history of apartheid |
| Method | mixed |
| Period studied | 1910–2010 |
| Regions | South Africa, homelands (former native reserves, TBVC and self-governing territories), provinces (Western Cape, Free State, KwaZulu-Natal etc.) |
| Themes | agriculture land, methods data sources, labour markets wages, bantustans removals influx, macro fiscal finance debt, business and state, sanctions divestment, manufacturing industrial policy, education human capital |
| Citation key | liebenberg2012 |
Argument¶
The thesis rebuilds a century-long (1910/11-2010/11) set of South African agricultural output, input, public-spending and R&D series, correcting for the exclusion of black/homeland farmers from commodity statistics and for changing official statistical practice. It replaces the earlier hybrid Laspeyres/Törnqvist-Theil approach (the 'Thirtle index') with Fisher Ideal Divisia indexes built from primary price and quantity data. The corrected series show output growth broadly similar to the older estimates, but input growth and multi-factor productivity (MFP) growth that are much lower. The share of output growth attributable to productivity is therefore far smaller than earlier work implied. Public agricultural R&D spending is shown to have stagnated from the 1970s, and policy phases (segregation, state-sponsored industrialization of white farming, deregulation, liberalization) are linked to spending and productivity patterns.
Evidence and method¶
Agricultural censuses and surveys from 1918, Abstract of Agricultural Statistics, Handbook of Agricultural Statistics 1904-1950, Union Statistics for Fifty Years, unpublished ledger files of the Directorate of Agricultural Statistics, mechanization sales surveys, AGFACTS data, commodity board data, government budget and appropriation reports from 1910, R&D surveys, FAO data, SARB national accounts, and personal communications with statisticians.
Place in the debates¶
- Measurement of long-run South African agricultural productivity (Thirtle et al. 1993 vs revised Fisher/Divisia indexes) (see Data reliability and econometric method). Argues earlier hybrid-Laspeyres 'Thirtle' indexes suffer aggregation bias and data errors, overstating output, MFP growth and the productivity share of output growth; the Fisher-based estimates are lower and track policy phases better.
- Effect of racial segregation and homeland exclusion on measured agricultural performance and productivity (see Agricultural policy, productivity, farm size and land reform). Holds that exclusion of black and homeland farmers distorted official commodity statistics, and that discriminatory policy suppressed black farmers' contribution, helping explain low measured 1947-68 productivity growth.
- Causes of declining farm employment (mechanization, tax incentives, minimum wages and security of tenure) (see Farm labour: mechanisation, labour shedding, wages and labour control). Attributes the post-1961 fall in farm employment mainly to mechanization of harvesting and capital-favouring policy, noting labour legislation after 1992-97 accelerated labour shedding; cites Van Zyl et al. on depreciation tax incentives and says the continuing decline is not conclusively explained.
Findings¶
-
Agriculture's share of GDP fell sharply over the century while absolute output grew. (p. 15)
By 2010 the agricultural share of the total economy had shrunk to just 2.5 percent even though the agricultural economy had grown in absolute terms
-
Official commodity tables systematically excluded black and homeland farmers' production, though the national accounts included it. (p. 16)
this resulted in the contribution of the black farmers and later the economies of the homeland areas being excluded from the commodity specific data tables — but not from the national income accounts.
-
Government support was racially biased: the Land Act, Co-operatives Act and credit/settlement schemes favoured white commercial farmers. (p. 42)
The Land Act confined land ownership by blacks to dedicated native reserves, while the Co-operatives Act excluded black farmers from participating in farmer cooperatives.
-
Black farmers' share of planted area fell sharply after the Second World War, which the author links to discriminatory policy and low measured productivity. (p. 119)
the suppressing effect that the discriminatory policies increasingly had on the contribution of this section of the farming community is significant enough to explain the low estimates of productivity growth
-
The Thirtle index overstates aggregate output growth relative to the Fisher index. (p. 86)
2.75 percent per year for this study, versus 2.90 percent per year using the Thirtle indexing procedure.
-
Farm employment peaked in 1961 and then fell with mechanization of harvesting. (p. 127)
Since 1961 it decreased by 46.3 until 1982 as a result of the mechanization of harvesting.
-
The 1992/93 minimum wage for agriculture raised labour costs and cut labour use. (p. 127)
The introduction of minimum wage rates in the agricultural sector in 1992/93 resulted in an increase in the cost of labour of 26.4 percent.
-
Official fertilizer expenditure data underestimated actual spending after 1981, biasing input measures. (p. 129)
Since 1981 the estimated expenditure on fertilizers reported by the Department varies between 7 and 44 percent below the expenditure reported in the census.
-
Public agricultural R&D spending stopped growing in the early 1970s. (p. 180)
real spending grew by an average of 5.1 percent per year from 1911 to 1950 increasing to 7 percent per year from 1950 to 1971, at which point spending effectively ceased growing.
-
Fisher-based MFP growth is far lower than the Thirtle estimate over 1945-2010. (p. 220)
1945-2010 2.15 3.01 0.86 1.62 3.23 1.84
-
Productivity growth was highest in the late 1930s to early 1950s, before the post-war MFP window begins. (p. 214)
the annual average rate of change in productivity was the highest in the period spanning from about 1934 to 1952.
Evidence ledger¶
Sources used¶
| Source | Kind | Period | How used |
|---|---|---|---|
| Abstract of Agricultural Statistics, 1958–2007 editions (Department of Agriculture / DAEM / DAS) | official statistical series | 1936–2007 | Baseline published output, input, capital formation and employment series that the author audits, corrects and compares against. |
| Agricultural Census and Survey reports 1918–2007 (OCS, BCS, DOS, CSS, Stats SA) | agricultural census | 1918–2007 | Primary source for rebuilt production accounts, farm structure, ownership, land use, labour, capital and input expenditure. |
| Handbook of Agricultural Statistics: 1904–1950 (DAEM 1961) | official statistical series | 1904–1950 | Supplementary production, trade and price statistics to fill early-period gaps in the production accounts. |
| Union Statistics for Fifty Years: Jubilee Issue, 1910–1960 (BCS 1960) | official statistical series | 1910–1960 | Source of long-run production, price and economic data, e.g. poultry and egg prices used to correct pre-1947/48 poultry value. |
| Unpublished ledger files and price archives of the Directorate of Agricultural Statistics (DAS 2009) | state archives | 1910–2011 (prices from 1936/37 and quarterly from 1958) | Gross value of production, farming requisites prices and input expenditure aggregates used to impute prices and build input price series. |
| Survey of Agricultural Mechanization Sales (DAMS 1968–1994; AGFACTS 1995–2010) | official statistical series | 1968–2010 | Used to disaggregate machinery expenditure by class and to estimate tractor and implement stocks and sales. |
| Thirtle et al. (1993) productivity dataset 1947–1997 | secondary literature compilation | 1947–1997 | Earlier index data used to replicate the 'Thirtle index' benchmark against which the new Fisher indexes are compared. |
| Government budget and appropriation reports: Estimates of Expenditure (Union of South Africa 1910–1959; RSA 1960–2009) | commission or official report | FY1910/11–2009 | Line-item data on public spending on agriculture, subsidies, R&D and research staff, built into a new long-run database. |
| ARC annual reports and ARC Financial Database | company records or reports | 1992–2011 | Funding sources, institute-level external income and staff qualifications for ARC institutes. |
| CeSTII National Survey of Research and Experimental Development (GERD) | official statistical series | 1966–2008 | Gross expenditure on R&D used to place agricultural R&D within national science spending and compute intensities. |
| SARB long-run GDP and agricultural GDP statistics (2011), Stats SA national accounts | national accounts macro | 1910–2010 | GDP deflator and GDP/AgGDP series used for deflating and for structural-change tables. |
| Commodity organisation and industry data (Wool Board/Cape Wools, Dairy Board, Milk Producers Organisation, SAGIS, Fertilizer Association, Potato SA, tree census bodies) | employer association records | 1910–2012 (varies) | Benchmarks to correct wool, milk, fertilizer and orchard-area estimates and to estimate homeland production after 1970. |
| Stats SA Producer Price Indexes long-run series (2012) | official statistical series | 2003–2011 | Replacement price indices for Department of Agriculture input prices judged unreliable since the late 1990s/2003. |
| Personal communications with Department of Agriculture and industry officials (Blignaut, Strydom, Rankin, van der Linde, Anderson) | interviews oral history | 2006–2012 | Clarifies how official estimates were constructed (capital formation, input validation, fertilizer coverage, survey history). |
| Farming in South Africa (DOA 1952, 1960) and Farming in South Africa jubilee issue articles | newspapers periodicals | 1952–1960 | Historical summaries of settlement schemes, orchard plant spacing, tree areas and sector development. |
| FAOSTAT database (2011) | international trade or finance data | 1961–2009 | Regional output and labour data for comparing South African land and labour productivity with the rest of Africa. |
| Roseboom et al. (1995) and Liebenberg et al. (2004) research personnel data | secondary literature compilation | 1961–2010 | Researcher counts and qualification profiles used for staff trends and degree-status table. |
| Commodity levy and trust data (NAMC 2007–2009, trusts unpublished) | commission or official report | 1999–2008 | Industry contributions to agricultural research. |
New data¶
Entirely new annual national price and quantity accounts for agricultural output 1910/11–2010/11 (adding homeland/black-farmer, on-farm and omitted commodities); revised capital (23 classes, 1945–2010), land, labour and purchased-input series; and a new long-run database of public agricultural expenditure, subsidies, R&D spending and researcher numbers since FY1910/11 (75 functions).
Inference¶
Quantification: Econometric (causal design).
Index-number accounting for a national time series: output, input and multi-factor productivity indexes (Fisher ideal and Törnqvist-Theil approximations of Divisia) are built from commodity-level price and quantity data. They are compared with the earlier 'Thirtle' hybrid index to show aggregation bias, and phases are matched to policy periods. R&D and policy chapters are descriptive time-series history. No regression linking R&D to productivity is estimated.
Key claims¶
| Claim | Basis | Strength |
|---|---|---|
| Official output statistics excluded black and homeland farmers' production and omitted or double-counted products (milk, eggs, vegetables, fruit), so the author built revised national production accounts. (p. 45) | Cross-checking Abstract of Agricultural Statistics against agricultural censuses, ledger files, Dairy Board and Wool Board data. | Partial data |
| Homeland/black farmers' share of planted area fell from about 23% in 1910/11 (peak 45.1% in 1923/24) to 11.8% by 1975/76, while their production share was consistently lower. (p. 57) | Author estimates from census nodes, interpolated by trend, plus commodity organisation data after 1970. | Partial data |
| The hybrid spliced-Laspeyres (Thirtle) output index overstates aggregate output growth relative to the Fisher/Törnqvist-Theil Divisia index (2.90 vs 2.75 percent per year, 1910–2010). (p. 61) | Same underlying price and quantity data indexed by different methods (Tables 3.3, 3.4). | Systematic data |
| Department of Agriculture capital formation and input data are unreliable (inconsistent machinery expenditure, fertilizer underestimated by 7–44% since 1981, input prices apparently indexed rather than observed). (p. 115) | Comparison of Abstract with census figures, Survey of Agricultural Mechanization Sales, Stats SA PPIs and correlation tests on 268 price series (Table 4.8). | Partial data |
| Multi-factor productivity growth with the revised Fisher indexes is much lower than earlier estimates (0.86% p.a. 1945–2010 vs 1.84% Thirtle), so productivity explains about a third rather than 62% of 2010 output value. (p. 205) | Revised output and input indexes (chapters 3–4) compared across the two methods (Tables 7.5–7.6, Figures 7.8–7.9). | Systematic data |
| Public agricultural R&D spending grew about 5.1% p.a. to 1952 and 7% to 1971, then stagnated, with research intensity falling behind the US and Australia after the early 1970s. (p. 161) | New budget-line database from Union/RSA Estimates of Expenditure and ARC data, deflated by the SARB GDP deflator. | Systematic data |
| Government support and settlement programmes were biased toward white commercial farmers (Land Act 1913, Co-operatives Act 1920, credit and settlement exclusions), and black farmers' yields fell to about a quarter of commercial yields by 1964. (p. 32) | Legislation, census data on shares of area, planted crops and livestock (Table 2.5) and yield comparison using DAFF/SAGIS data. | Partial data |
The author's caveats¶
Acknowledges that pre-1958 fruit and vegetable production, nuts, specialty crops and ostrich output are poorly enumerated; that homeland output is mostly available only for census years and is interpolated; that early food-balance records are lost or spotty; that census detail and sampling have declined since the 1980s and 1990s, excluding small and black farmers; that input price data since about 2003 are suspect; that land rental rates rest on arbitrary 4% and 12 ha/LSU assumptions; that no quality adjustment was made for labour and capital; that settlement-scheme subsidies are not captured; and that analysis before 1945 for MFP and 1936 for inputs is impossible.
Gaps in the evidence¶
Farm-level or district-level microdata from the censuses, which the author says are no longer accessible, would permit direct estimation of black farmers' production and inputs rather than trend interpolation, as would spatially disaggregated land rental and quality data. The R&D–productivity link is asserted through phase matching and not estimated econometrically, though the series now exist to attempt it. Pre-1945 input data (labour quality, capital, settlement-loan subsidy equivalents) likely exist in archives but were not found, leaving the 1910–1945 period largely at partial-productivity level.
Replicability: Data described but not public.