What is the appropriate level of aggregation for productivity indices? Comparing district, regional and national measures¶
B Conradie; J Piesse; C Thirtle (2009). What is the appropriate level of aggregation for productivity indices? Comparing district, regional and national measures. Agrekon. Agricultural Economics Association of South Africa.
| Type | journal article |
| Scope | Wider apartheid literature |
| Method | descriptive quantitative |
| Period studied | 1952–2002 |
| Regions | Western Cape, Karoo, Breede River Valley, Olifants River Valley, Swartland, Boland |
| Themes | agriculture land, methods data sources, labour markets wages |
| Citation key | conradie2009 |
Argument¶
The paper argues that national agricultural TFP indices hide large regional and district variation, so aggregation destroys information relevant to policy. Using Tornqvist-Theil TFP indices built from farm censuses for 31 Western Cape magisterial districts and nine regions, it shows very different trajectories. The Karoo declined, the irrigated Breede and Olifants valleys grew fast, and Swartland results were driven by one district. The differences reflect output mix, irrigation, electrification and climate. The authors conclude that localised productivity measures are needed and that national figures omit the black smallholder sector.
Evidence and method¶
Ten Farm Censuses (1952, 1956, 1960, 1965, 1971, 1976, 1981, 1988, 1993, 2002) for 31 Western Cape magisterial districts grouped into nine regions, plus the Abstract of Agricultural Statistics (1999). Outputs are aggregated from fruit, vegetables, wine, livestock, poultry and field crops. Inputs are labour hours, land, intermediate inputs, animal herds and machinery. National deflators are used where local prices are lacking. The comparison benchmark is the national TFP index of Thirtle et al. (1993).
Place in the debates¶
- appropriate level of aggregation for agricultural productivity measurement / induced innovation (see Data reliability and econometric method). Argues that aggregate national indices mask regional heterogeneity, citing Olmstead and Rhode's regional critique of Hayami-Ruttan induced innovation, and favours disaggregated district and regional indices with local policy advice rather than one-size-fits-all policy.
Findings¶
-
The national TFP index for 1947-91 showed growth only from the mid-1960s, which provides the benchmark for the regional comparison. (p. 4)
TFP grew at 1.26% per annum, but before 1965, there was no TFP growth at all. Then, growth was a healthy 2.15% per annum until 1981
-
Western Cape TFP growth over 1952-2002 was moderate and statistically similar to the national average, but its timing differed. (p. 4)
over the full period TFP for the Western Cape grew at 1.22% per annum, which is not significantly different from the national average
-
Karoo TFP declined over the period, with no intensification and falling output, reflecting a sheep-dominated arid region. (p. 7)
Table 2 shows that TFP declined at 0.72% per annum between 1952 and 2002. The increase in inputs is not significantly different from zero, so there is no intensification
-
Karoo agriculture shed inputs but not fast enough to offset falling output after the wool boom. (p. 7)
inputs fell at 0.85% per annum from 1971 onwards and output at 1.32%
-
The Breede River Valley had very high TFP growth after 1976, driven by canned fruit exports. (p. 7)
Together, these changes give a TFP growth rate of 3.51%, which is way above the figure for the province or the country.
-
The Breede gains came from improvements in irrigation water productivity rather than from irrigation itself. (p. 7)
The main change that enabled this impressive productivity increase was not irrigation per se, as this dates back to the 1920s.
-
In the Olifants valley, take-off is attributed to irrigation and rural electrification. (p. 8)
Here, the take off is due to the introduction of irrigation and also rural electrification, as these amenities became available.
-
The 1993 dip in Swartland is a rare instance in which policy (wheat deregulation) shows up in the data. (p. 8)
This follows in the wake of deregulation of the wheat market, so it is one instance where policy effects seem to have a measurable impact.
-
Data quality is a major limitation, including implausible census records and the use of national deflators. (p. 11)
the information available on rapidly growing sectors such as poultry, pigs and dairying is simply not correct.
-
National commercial-farm statistics ignore the black smallholder sector, so land redistribution could move national TFP in misleading ways. (p. 11)
it makes little sense to only record commercial agriculture whilst ignoring the substantial black smallholder sector.
Evidence ledger¶
Sources used¶
| Source | Kind | Period | How used |
|---|---|---|---|
| Farm (Agricultural) Censuses of 1952, 1956, 1960, 1965, 1971, 1976, 1981, 1988, 1993 and 2002, magisterial district level data for the Western Cape | agricultural census | 1952-2002 (10 census years) | Supplies the district-level physical and value data on outputs and inputs from which the Tornqvist-Theil TFP indices are built. |
| Abstract of Agricultural Statistics (1999) | official statistical series | post-1950 (published 1999) | Supplementary source of agricultural information for the period after 1950. |
| Thirtle, Sartorius von Bach & Van Zyl (1993) national TFP index for South African agriculture 1947-91 | secondary literature compilation | 1947-1991 | Provides the national TFP growth benchmark against which the Western Cape, regional and district results are compared, and the method for capital and farm-produced inputs. |
| National price deflators (for machinery and other inputs) | official statistical series | 1952-2002 | Used to deflate current-value inputs such as machinery and equipment where district-level prices were unavailable. |
| Constructed price deflators matched to input categories (for poultry and other value-only outputs) | other | 1952-2002 | Deflates outputs reported only in current values so that they approximate physical series comparable across years. |
New data¶
The authors constructed district-level and regional input, output and TFP indices (Tornqvist-Theil) for 31 Western Cape magisterial districts and nine regions from ten farm census years, 1952-2002, by aggregating and deflating census data; the indices are reported only in Table 2 and Figures 1-2.
Inference¶
Quantification: Statistical analysis.
Tornqvist-Theil (superlative) TFP indices are built from census data, with growth rates estimated by regressing log TFP on a constant and time trend over sub-periods. They are compared across province, regions and districts and with the national aggregate of Thirtle et al. (1993). The causes of the differences (crop mix, irrigation, electrification, weather, deregulation) are inferred informally from the output mix and the index patterns, without formal tests.
Key claims¶
| Claim | Basis | Strength |
|---|---|---|
| Western Cape TFP grew about 1.22% a year over 1952-2002, not significantly different from the national average, but with different timing (decline 1965-71, growth after, decline at the end). (p. 4) | Table 2 trend regressions on the farm census based indices, compared with the Thirtle et al. (1993) national figures. | Partial data |
| The Karoo experienced TFP decline (about 0.72% a year) with no intensification, driven by falling output in a sheep-dominated arid region. (p. 7) | Table 2 and Figure 1 for the Karoo region, with the output mix explanation asserted from the region's sheep dominance. | Partial data |
| The Breede River Valley had very high TFP growth (2.22% a year overall, 3.51% after 1976), attributed to canned fruit exports and improvements in the productivity of irrigation water. (p. 7) | Table 2 and Figure 1 for the region; the attribution to irrigation water productivity and canned fruit is argued without separate data. | Partial data |
| The Olifants River Valley follows a similar but weaker pattern to the Breede, with growth linked to the introduction of irrigation and rural electrification. (p. 8) | Table 2 and Figure 1; the causal link to irrigation and electrification is asserted. | Illustrative or anecdotal |
| Swartland's regional TFP growth is driven by Piketberg's switch to higher-valued fruit, while the dominant district Malmesbury-Moorreesburg shows no significant TFP growth. (p. 8) | District-level indices in Table 2 and Figure 2; the crop-switching explanation is conjecture. | Partial data |
| The 1993 fall in Swartland inputs and outputs followed wheat market deregulation, a case where policy effects show up in the data. (p. 8) | Figure 2 pattern in 1993 plus timing of deregulation; no policy data or test. | Illustrative or anecdotal |
| Aggregation hides large differences between and within regions, so localised productivity measures are needed for policy advice. (p. 9) | Comparison of national, provincial, regional and district indices from Table 2 and Figures 1-2. | Partial data |
The author's caveats¶
The authors acknowledge that census data for poultry, pigs and dairying are incorrect (e.g. 412 chickens recorded in Worcester), that many prices and quantities are missing so national deflators had to be used, and that no quality adjustment for chemical and mechanical inputs is possible. They also note under-recorded home slaughtering, unrecorded fruit sold informally, sporadic or missing records for some outputs, diversification of farm income, and only ten observations per district, with pre-1952 data too incomplete. They add that the data cover only commercial agriculture and ignore the black smallholder sector.
Gaps in the evidence¶
The paper does not give the input and output weights, the data tables or the district series, so the indices cannot be checked, and no test shows that the regional and district differences exceed measurement error. The explanations (irrigation, electrification, crop switching, weather, deregulation) are not tested against rainfall, irrigation, electrification or policy data, which exist in archives and official series. With only ten census points per district and no information on the black smallholder sector, the results are fragile and the apartheid-era institutional context is untreated.
Replicability: Data described but not public.