Agricultural Statistics - A Need for Reform¶
Delia Hendrie (1976). Agricultural Statistics - A Need for Reform. SALDRU Farm Labour Conference, September 1976, Paper No. 38. SALDRU, School of Economics, University of Cape Town.
| Type | conference paper |
| Scope | Economics and economic history of apartheid |
| Method | policy analysis |
| Period studied | 1960–1976 |
| Regions | South Africa |
| Themes | methods data sources, agriculture land, labour markets wages |
| Citation key | hendrie1976a |
Argument¶
Official South African agricultural statistics, especially on farm labour, are too late, too infrequent, loosely defined and incomplete for research and policy. The census is published with a lag of more than two years, employment is counted on a single August date, and the categories of regular, casual and domestic worker are poorly defined. Wage estimates derived from the census are therefore unreliable. The paper compares this coverage with the US and recommends replacing the annual census with a less frequent census supplemented by quarterly stratified sample surveys. It argues this would improve timeliness and coverage without burdening farmers.
Evidence and method¶
Department of Statistics publications (Agricultural Census Nos. 39, 44, 45 and 46, the Statistical News Release of April 1976, and the 1974 Annual Report), summarised in tables of publication lags, employment 1960-1974 and labour statistics coverage. It also draws on a comparison with US farm labour statistics and the statistical-methods literature on sampling.
Place in the debates¶
- Annual full census versus sample surveys for agricultural data collection (see Data reliability and econometric method). Argues against the Department of Statistics' reliance on an annual police-assisted census. Favours a less frequent census plus quarterly stratified sample surveys.
- Reliability of official farm labour and wage data (see Data reliability and econometric method). Holds that census employment and earnings figures are too flawed to support firm conclusions on farm labour use or wage rates.
Findings¶
-
The latest published census was two to three years out of date. Farm wage movements after 1973 were largely unknown, though other sectors' African wages had risen sharply. (p. 4)
while in the mining sector this increase has been 189%.
-
Publication lags reduce the usefulness of census results for researchers. (p. 5)
This lag impairs the usefulness of the census results
-
The employment time series is incomplete: several years are missing and others lack a breakdown by type of employee. (p. 7)
Table 3 shows there is no employment data for the years 1966, 1967 and 1970
-
A single August employment count cannot capture agriculture's labour needs, because demand varies seasonally. (p. 7)
The demand for labour fluctuates enormously from one month to another.
-
The regular/casual distinction is ill-defined, so migrant or part-year workers cannot be classified properly. (p. 8)
whether the migrant farm worker employed for nine months of the year should be classified as a regular or casual employee
-
Implied domestic worker earnings are implausibly low, probably because part-time workers are counted as full-time. (p. 8)
This wage in 1973 works out at R6,86 for African employees
-
The census gives no direct farm wage rates, so only crude estimates are possible. (p. 9)
No data is given specifically on wage rates paid to farm employees
-
Farmers' arbitrary valuation of payments in kind undermines the accuracy of total farm wages. (p. 9)
This point is important since a large proportion of farm workers' wages are paid in kind
-
Proposed reform: a less frequent census plus quarterly sample surveys would raise both frequency and coverage of data. (p. 13)
A less frequent census supplemented by quarterly sample surveys would allow both the frequency of the agricultural data to be
Evidence ledger¶
Sources used¶
| Source | Kind | Period | How used |
|---|---|---|---|
| Agricultural Census No. 46 (1972/3), Report No. 06-01-10 | agricultural census | 1972/3 (also 1973 wage estimate) | Used to describe the census labour coverage (employment and earnings) and to compute the implausible average domestic wage of R6,86 per month. |
| Agricultural Census series (e.g. Nos. 39, 44, 45, 46) and publication-lag table | agricultural census | 1964/5-1972/3 | Publication dates are tabulated to show a lag of more than two years before census results appear. |
| Department of Statistics, Report on Agricultural and Pastoral Production 1970/1, Agricultural Census No. 44 | agricultural census | 1960-1971 | One of three sources for the Table 3 employment series 1960-1974, which shows gaps in the data. |
| Statistical News Release, 15 April 1976 (and 1973/4 release) | official statistical series | 1973/4-1974 | Supplies the 1974 employment figures in Table 3 and is cited as the only limited recent information on the sector. |
| Abstract of Agricultural Statistics | official statistical series | time series, unspecified | Described as one of the three census publications, containing national time-series data. |
| Production estimates of the Department of Agricultural Economics and Marketing | official statistical series | 1973-1976 | Cited as one of the few sources of information on recent agricultural developments. |
| Commission of Enquiry into Agriculture (Du Plessis Commission) | commission or official report | c. 1966-7 | Cited as evidence of data gaps: the Commission did its own primary research and used unpublished departmental files. |
| Unpublished files of the Department of Agricultural Economics and Marketing (used by the Du Plessis Commission) | state archives | unspecified | Mentioned as the unpublished material the Commission had to rely on. |
| Department of Statistics, Annual Report of the Statistics Council and the Secretary of Statistics for 1974 | commission or official report | 1974 | Quoted to show that a proposal to use sample surveys was rejected because the police assist the census. |
| U.S. agricultural labour statistics (farm population, farm employment, farm wage workers) | official statistical series | unspecified | Used as a comparison to show that fuller labour coverage is feasible; the source line is blank. |
| Fragmentary data on African wage increases in manufacturing and mining since 1973 | other | 1973-1976 | Cited as the figures behind the suggestion that farm wages have risen considerably, though no source is given. |
Inference¶
Quantification: Descriptive tables.
Critical audit of the official census statistics. It draws on tabulated publication dates and employment counts, a worked wage calculation, and a comparison with U.S. statistics, then argues for a sample-survey design. The reasoning is argumentative and methodological.
Key claims¶
| Claim | Basis | Strength |
|---|---|---|
| Publication of agricultural census results lags by more than two years, making them of little current use. (p. 5) | Table 1 compares census dates with publication dates from the Department of Statistics. | Partial data |
| The farm employment time series is discontinuous, with only eight of fifteen years (1960-1974) offering a breakdown by employee type. (p. 7) | Table 3, compiled from Agricultural Census Nos. 44 and 46 and the 1976 News Release. | Systematic data |
| A single-date (August) employment count cannot show agriculture's reliance on labour because demand fluctuates seasonally. (p. 7) | Description of the census design in Table 2, plus the author's knowledge of production types; the extent of the August bias is not measured. | Asserted |
| The census categories of regular, casual and domestic employees are poorly defined. The implied 1973 African domestic wage of R6,86 per month is implausibly low. (p. 8) | Census No. 46 earnings and employment data divided to give an average, with the author's judgement that the result is implausible. | Partial data |
| Wages in kind are valued arbitrarily by farmers, so total wage estimates are unreliable. (p. 9) | Description of the census questionnaire, which gives no valuation guidance. | Asserted |
| Major labour data gaps exist (age, sex, education, job category, dependants), as shown by the Du Plessis Commission's need for primary research and by the fuller U.S. coverage. (p. 9) | The Du Plessis Commission's practice and the U.S. comparison table. | Illustrative or anecdotal |
| Replacing the annual census with a less frequent census plus quarterly stratified sample surveys would improve frequency and coverage at little extra cost. (p. 13) | Sampling theory and the author's argument, with no costing or pilot data. | Asserted |
The author's caveats¶
The author notes that the 1964 and 1965 employment figures must be treated with caution because such large fluctuations seem unlikely. She also notes that the wage estimates are crude, that Table 1 dates are approximate, and that the U.S. table is incomplete.
Gaps in the evidence¶
The paper asserts that the August count is unrepresentative and that in-kind valuations are arbitrary, but offers no alternative data to test this, such as monthly employment series or farm-level surveys. It gives no quantitative comparison of census wage estimates against independent wage surveys. Pilot sample-survey results or costings would test the proposed reform, and such data, like farm surveys and later census revisions, could exist for comparison today.
Replicability: Data described but not public.