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, Faculty of Agricultural and Natural Sciences. University of Pretoria.
| Type | thesis |
| Scope | Economics and economic history of apartheid |
| Method | descriptive quantitative |
| Period studied | 1910–2010 |
| Regions | South Africa, former homelands, Western Cape, Sub-Saharan Africa (comparative) |
| Themes | agriculture land, methods data sources, labour markets wages, bantustans removals influx, business and state, manufacturing industrial policy, macro fiscal finance debt, demography urbanisation, education human capital, comparative theory, sanctions divestment, transition post apartheid |
| Citation key | liebenberg2012a |
Argument¶
The thesis rebuilds a century of South African agricultural output, input and public R&D data (1910/11-2010/11) after correcting for the exclusion of black and homeland farmers from official commodity statistics, shifting statistical definitions and the ad hoc 'Thirtle' hybrid indexing. Using Fisher Ideal (Divisia-approximation) indexes on primary price and quantity data, it shows that earlier studies overstated output growth, input growth and especially multi-factor productivity (MFP) growth. Aggregate output trends turn out broadly similar to the official series, but input and productivity patterns differ substantially, and the share of output growth attributable to productivity is much smaller than earlier estimated. Public agricultural R&D and farmer support are placed in policy phases, with R&D spending stagnating from the 1970s. Racial discrimination in support and land policy, mechanisation and labour shedding are documented throughout.
Evidence and method¶
Agricultural census and survey reports from 1918 (OCS, BCS, CSS, Stats SA), Abstract of Agricultural Statistics, Handbook of Agricultural Statistics 1904-1950, Union Statistics for Fifty Years, unpublished handwritten ledger files of the Directorate of Agricultural Statistics, Survey of Agricultural Mechanization Sales and AGFACTS, commodity board data, government budget and appropriation reports 1910-2010, ARC records, SARB national accounts, FAOSTAT, and personal communications with departmental statisticians.
Place in the debates¶
- How reliable are official South African agricultural statistics and how far did they bias earlier productivity estimates (see Data reliability and econometric method). Argues official series (Abstract of Agricultural Statistics) are inconsistent, exclude homeland production, double-count milk, omit horticulture before 1958 and understate fertilizer use; the earlier Thirtle et al. (1993) hybrid index overstates MFP growth.
- Pace and timing of South African agricultural MFP growth (Thirtle et al. vs revised estimates) (see Data reliability and econometric method). Finds MFP growth of 0.86% a year 1945-2010 with the Fisher index against 1.84% with the Thirtle index; growth is much lower than earlier work and shows a slight contraction 1980-94.
- Effect of racial segregation policy and state support on black farming and national productivity (see Agricultural policy, productivity, farm size and land reform). Holds that support was biased towards white commercial farmers (Land Act 1913, Co-operatives Act 1920, credit and settlement exclusion) and that suppressed black output helps explain low measured productivity growth 1947-68.
- Causes of farm labour decline (mechanisation, tax incentives, labour legislation) (see Farm labour: mechanisation, labour shedding, wages and labour control). Attributes the post-1961 fall in employment to mechanisation of harvesting, supported by tax incentives on capital (citing Van Zyl et al.), with minimum wages and security of tenure legislation accelerating shedding in the 1990s.
- Returns to agricultural R&D and its timing (see Agricultural policy, productivity, farm size and land reform). Suggests post-war productivity gains partly reflect pre-war investments and settlement schemes; R&D investment stagnated from 1971/78 and productivity effects lag by decades.
Findings¶
-
Fisher index shows lower century-long output growth than the Thirtle hybrid index, so past studies overstated output and productivity. (p. 74)
the practice of splicing the reported Laspeyres output quantity indexes for the different base years to form a long-run index yields an index that overstates the growth in aggregate agricultural output
-
Official statistics systematically excluded black and homeland farmers until 1995. (p. 39)
Through to 1995, the commodity-specific data tables on agricultural output quantity reported in the Abstract of Agricultural Statistics excluded production from the former homelands, or former native reserves
-
State support, credit and settlement were racially discriminatory, disadvantaging black farmers. (p. 42)
the sustained and substantial government support to agriculture was biased towards white commercial farmers. Lacking a commensurate amount of public support, black farmers suffered as a consequence.
-
Area planted by black farmers peaked in 1923/24 despite the 1913 Land Act and then fell sharply after the 1940s. (p. 119)
Surprisingly, the area planted by black farmers has increased to 2.53 million hectares in 1923/24 in spite of the promulgation of the Land Act of 1913.
-
Farm employment peaked in 1961 and fell sharply thereafter owing to harvest mechanisation; minimum wages in 1992/93 raised labour cost 26.4%. (p. 127)
Since 1961 it decreased by 46.3 until 1982 as a result of the mechanization of harvesting.
-
Revised MFP growth is far lower than the Thirtle-method estimate over 1945-2010. (p. 220)
1945-2010 2.15 3.01 0.86 1.62 3.23 1.84
-
Fisher-based accounting attributes only about a third of 2010 output value to productivity growth, versus 62% under the Thirtle approach. (p. 223)
The remaining 33 percent (i.e., 48.6/148.6 = 0.33) is accounted for by the increase in productivity.
-
Public agricultural R&D spending stopped growing in real terms after 1971. (p. 179)
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.
-
Government spending on agriculture fell from 14.2% of total expenditure in 1947 to 2.3% in 1980 and 1.7% by 1994, with subsidies phased out by the 1990s. (p. 153)
from where it erratically decreased to reach 2.3 percent of total government expenditure in FY1980.
-
Tractor numbers grew rapidly to 1970s peak then declined, reflecting mechanisation drive and later retrenchment. (p. 90)
In 1918 there were only 231 steam and oil burning tractors on South African farms.
-
Black farmers' maize yields relative to white farmers fell from 39.9% in 1954 to 26.3% in 1964. (p. 206)
In 1954 yields obtained by black farmers were 39.9 percent of the average achieved by white farmers. Over the subsequent decade this ratio decreased to 26.3 percent in 1964
Evidence ledger¶
Sources used¶
| Source | Kind | Period | How used |
|---|---|---|---|
| Abstract of Agricultural Statistics (1958 onward; editions 1958, 1967, 1972, 1980, 1985, 1995, 1997, 2007, 2012) | official statistical series | 1936-2012 (editions); summary series back to 1910 | Baseline official output, input, employment, capital formation and volume-of-production series that the author audits, revises and rebuilds. |
| Agricultural Census and Survey reports, 1918-2007 (OCS, BCS, DOS, CSS, Stats SA; e.g. Agricultural Census No. 1, 1918) | agricultural census | 1918-2007 | Primary source for reconstructing national production accounts, farm structure, land use, capital stocks, employment, input expenditure and black/white sub-sector shares. |
| Handbook of Agricultural Statistics: 1904-1950 (DAEM 1961) | official statistical series | 1904-1950 | Supplementary production, trade and price statistics for the pre-1950 revised production accounts. |
| Union Statistics for Fifty Years: Jubilee Issue, 1910-1960 (BCS 1960) | official statistical series | 1910-1960 | Provided price and production data to correct poultry value and other pre-1960 estimates. |
| Unpublished ledger files and archives of the Directorate of Agricultural Statistics (gross value of production, volume of production, input prices, Farming Requisites Price Index) | state archives | 1910-2011 (hand-written ledgers before 1980; electronic from 1980) | Source of pre-1947 output estimates, homeland production adjustments and over 262 quarterly input price series used to build the Fisher indexes. |
| Survey of Agricultural Mechanization Sales (DAMS 1968-1994) and AGFACTS machinery sales data (1994-2010) | official statistical series | 1968-2010 | Used to disaggregate machinery expenditure into classes and to build capital stock and service-flow estimates, after being compared with census figures. |
| ARC Financial Database and ARC Annual Reports (1992/93-2010/11) | company records or reports | 1992-2011 | Funding sources, non-core income and staffing of ARC institutes (Tables 6.3, Fig 6.10). |
| Survey of Research and Experimental Development (CeSTII/HSRC GERD data, various years) | household or labour survey microdata | 1966-2008 | Gross expenditure on R&D series placing agricultural R&D in the national science context. |
| South African Reserve Bank long-run GDP and agricultural GDP series (SARB 2011) and Stats SA population/GDP data | national accounts macro | 1910-2010 | GDP deflator and national-accounts context for deflating all series and for Table 2.1 structural change. |
| Commodity organization and board data (Dairy Board annual reports, Cape Wools/Wool Board, SAGIS, Potato SA, tree census surveys DFPT/AVCASA/SAWIS, NAMC levy surveys, Milk Producers Organization) | employer association records | 1980-2011 mostly | Used to validate or correct milk, wool, area planted, tree/vine area, yields and levy-funded research figures. |
| Stats SA Producer Price Indexes and long-run building materials index | official statistical series | 2005-2011 (building materials long-run) | Substitute price indexes for input categories where Department of Agriculture prices appeared unreliable after about 2000. |
| Thirtle, Sartorius von Bach and van Zyl (1993) productivity dataset 1947-1997 | secondary literature compilation | 1947-1997 | Data supplied by Thirtle used to replicate and compare earlier index methods (the 'Thirtle index'). |
| Roseboom et al. (1995) ISNAR Statistical Brief No. 23 and Liebenberg et al. (2004) ASTI country brief (researcher and R&D data) | secondary literature compilation | 1910-2004 | Prior R&D spending, staff counts and qualification profiles extended and revised in the new panel. |
| Farming in South Africa (DOA 1952; 1960 Jubilee edition) and Union government reports | newspapers periodicals | 1952-1960 | Plant spacing for orchard area estimates, settlement scheme descriptions and early investment estimates. |
| FAOSTAT database (2011) | international trade or finance data | 1961-2009 | Regional comparison of land and labour productivity loci (Fig 7.6). |
| Personal communications with Department of Agriculture and industry officials (Blignaut, Strydom, FSSA, SAAMA, Anderson) | interviews oral history | 2010-2011 | Clarify how official estimates are constructed and verify data sources. |
| Land Settlement Acts and related legislation (Land Act 1913, Co-operatives Act 1920, Agricultural Marketing Act 1937, Soil Conservation Act 1946, Land Settlement Act 1912/1956) | legislation gazette | 1912-1996 | Context for discriminatory policy and subsidy regimes explaining black/white sector divergence and policy phases. |
New data¶
New annual national panel 1910/11-2010/11 of agricultural output prices and quantities (revised to include homeland and black-farmer production), input series for capital (23 classes, 1945-2010), land, labour (incl. proprietor/family, excl. domestic) and purchased inputs, plus a line-item database of public agricultural spending and R&D spending/staff since 1910/11.
Inference¶
Quantification: Econometric (causal design).
Time-series reconstruction and index-number construction (Fisher Ideal/Divisia, Törnqvist-Theil) from commodity-level price and quantity data, compared with the earlier hybrid 'Thirtle' index to quantify aggregation bias; partial and multi-factor productivity and growth-accounting by policy phase. Historical description of policy and institutions accompanies the quantitative work; no causal identification of R&D effects.
Key claims¶
| Claim | Basis | Strength |
|---|---|---|
| Official production statistics excluded black/homeland farmers and omitted or double-counted products (milk, poultry, vegetables, fruit), so revised livestock value is 3.6-10.7% above official before 1947. (p. 53) | Reconciliation of Abstract against agricultural census reports, Union Statistics, Handbook, ledger files and Dairy Board/Wool Board data. | Partial data |
| The hybrid Thirtle index overstates output growth (2.90% vs 2.75% per year Fisher, 1910-2010) due to changing commodity baskets and spliced Laspeyres indexes. (p. 74) | Fisher, Törnqvist-Theil, Laspeyres, Paasche and Thirtle indexes built from the same rebuilt price-quantity data (Tables 3.3-3.4). | Systematic data |
| Black farmers' share of field-cropped area in the homelands was 23% in 1910/11, peaked at 45.1% in 1923/24 and fell to 11.8% by 1975/76; their output share was much smaller. (p. 57) | Author's estimates from census nodes, ledger files and commodity organization data, trend-interpolated between census years. | Partial data |
| Official capital formation and fertilizer expenditure series were biased relative to census data (machinery up to 52.5% too high in 1981; fertilizer 7-44% too low since 1981). (p. 115) | Comparison of Abstract with census reports and Survey of Agricultural Mechanization; Fertilizer Association information. | Systematic data |
| Department of Agriculture input price data since about 2000 appear to be indexed rather than observed (high cross-correlation, inflation above Stats SA PPIs), biasing productivity upward in the old method. (p. 117) | Electronic price files from the Department; correlation tests on 268 price variables; comparison with Stats SA PPIs. | Systematic data |
| Fisher-based MFP growth is 0.86% per year 1945-2010 versus 1.84% with the Thirtle method; productivity accounts for about 33% of 2010 output value versus 62% in the Thirtle approach. (p. 203) | Constructed Fisher and Thirtle input, output and MFP indexes (Table 7.6) and growth-accounting arithmetic. | Systematic data |
| Real public agricultural R&D spending grew about 5.1% per year to 1952 and 7% to 1971, then stagnated, and intensity fell behind the US and Australia after the early 1970s. (p. 161) | New line-item database from Union/RSA budget Estimates and ARC records, deflated by SARB GDP deflator. | Systematic data |
| Productivity growth was fastest around 1934-1952, so post-1945 MFP analysis misses much of the gain from earlier public investment and settlement support. (p. 194) | Partial land and labour productivity loci from 1910 built from rebuilt output and land/labour series (Fig 7.4). | Partial data |
The author's caveats¶
Extensive: official statistics changed scope and definition over time and excluded homelands; early output and horticulture are partly imputed; census detail and coverage decline after the 1980s and the 1990s census samples only larger farms; homeland production and area are trend-estimated; land rental rates are arbitrary (4% of production value, 12 ha/LSU); no quality adjustment for labour or machinery; spliced Laspeyres price indexes for some inputs; input price data after 2000 unreliable; pre-1947 original food-balance records unavailable.
Gaps in the evidence¶
The output and input accounts for homeland and black farmers, pre-1958 horticulture and the pre-1945 inputs rest on interpolation and ledger fragments, so the black/white sector story is weakly measured. Farm-level or district microdata from the census (which exist in the archives but are no longer accessible) and quality-adjusted capital and labour data would test the MFP results better. The R&D-productivity link is only juxtaposed, not estimated; a regression with lags and controls would need the new series, which now exist.
Replicability: Data described but not public.