4 A review of methods and materials used in the present study
4.6 The database
4.6.1 Data limitations
Data availability was an important limitation when considering the scope of this thesis. As explained is section 1.7, the overriding interest of this thesis was to evaluate and examine TFPG at the farm level. Therefore, not only farm-level data were required. Most importantly, to examine the gains in TFP, data on the same group of farms over a number of years were required. Two sources of microdata on dairy farming were available: Dexcel and Ministry of Agriculture and Forestry (MAF). The Dexcel database gathers a vast amount of information on a large number of farms in different regions over a number of years. However, only a handful of farms are repeated over the years, making this database unsuitable for studies in productivity growth (at least at the farm level). The MAF database, in turn, requests less information from a limited number of farmers, often the same farms
over the years. Given that the dairy industry was interested in productivity gains at the farm level, the analysis was performed on the panel data (longitudinal data) gathered in the MAF database.
The Dexcel database could have been used to ascertain whether farms in any one region share the same technology with farms in other regions for any given year. The focus would be on the cross section and the outcome would be a measure of the (in)efficiency with which each farm is applying the technology. Similarly, the robustness of efficiency estimates to variable selection could have been assessed. Furthermore, the vast amount of secondary information on herd characteristics, investment, social factors and weather variables could have been used to examine the determinants of inefficiency. The SFA model proposed by Battese and Coelli (1995) would have allowed examining both, i.e., technological differences between farms in different regions and determinants of inefficiency (more on this in section 10.3).
The database obtained from MAF is used to monitor the production and financial status of farms in terms of their cash income and expenditure. Each year, MAF Policy published a “model” budget for different dairy regions. The “model” is based on the data from a survey of a number of commercial farms for each region10. MAF contracts with farm
consultants who select the properties based on a range of criteria (e.g., dairy farm within the required region, owner-operator and a commercial unit). The selection is therefore not entirely random, as the consultants tend to pick farms they know. The consultants visit the farms (in mid-May) and collect the financial information for the year ended (or just about to end) and for the new year starting. This is then collated together for the respective regions and provided to MAF. MAF then holds an “industry” meeting within each region that involves a cross section of people involved in the dairy industry (e.g., dairy company, bankers, accountants and some leading farmers) to discuss the information from the survey. The survey is then written up to include a published “model” budget for that region.
MAF Policy supplied farm-level data for the seasons 1996–1997 to 2004–2005 for the present analyses. Two hundred and ten dairy farms were surveyed over the nine year period but the database contained only 861 observations. A balanced (complete) panel would
contain a maximum of 1,890 observations (210 farms times 9 years), so the panel is not balanced, i.e., data for some farms on some years are missing. The time length of the MAF database was exploited. Hence, only farms that were surveyed at least in 1997 and 2005 were selected. The number of dairy farms present in both these years totalled 36: 4 in Northland, 8 in Waikato and Taranaki, 9 in Canterbury and 7 in Southland. Data from Northland were discarded because the number of farms was too small to enable the production frontier to be modelled. Therefore, the final panel was comprised of 32 farms. A total of 264 observations remained in the panel, so 24 observations were missing because some farms were not surveyed in all 9 years. Given the number of data in each region, it was decided to pool the four regions into two regions. Region I (RI) includes Waikato and Taranaki, whereas Canterbury and Southland comprises region II (RII). Each region has the same number of farms, 16, with 125 and 139 observations respectively. This aggregation was preferred because it better reflects the commonalities between Waikato and Taranaki—the traditional dairy regions—and the relatively newer dairy regions of Canterbury and Southland (Section 2.6).
The main characteristics of the whole sample are outlined below (Table 4.1, Table 4.2 and Table 4.3). Expenditures were converted into quantities by dividing by annual dairy farm expenses price index (1992/93=1000). Statistics New Zealand provides dairy farm expenses price index on a quarterly basis (Stats NZ). In order to match farm-level data, reported from June to May, the average of the period 2nd quarter 1992–1st quarter 1993 was used as the base year to convert expenditures into 1992/1993 NZ dollars. It was assumed that all farms paid the same prices for each item in any given period. If some farms paid higher prices for a quality input, dividing by the same price converts these inputs into a quality-adjusted input. The deflated expenditures were aggregated into different inputs.
Table 4.1 - Characteristics of the whole sample (average values per farm)
1997-2005 Average Std Dev All farms (264 observations) Max Min
Milk Production (total milksolids, kg) 140,509 114,062 725,000 30,000
Factor inputs
Cows in Milk (number) 385 273 1,600 104 Area (milking platform, ha) 143 102 555 33 Labour (total hrs per year) 5,044 3,195 22,180 2,250 Feed (all purchased feed, NZ$ 92/93) 69 80 666 6 Fertilizer (expenditure, NZ$ 92/93) 55 55 344 2 Intermediate inputs (health, breeding, shed, feed
and fertilizer expenses) 158 152 1,075 21 K2 (depreciation and interest on buildings and
machinery and expenditure on repairs and
maintenance, NZ$ 92/93) 76 65 470 17 K9 (expenditure on: repairs and maintenance on
buildings and machinery, fuel, electricity, rates and insurance, administration and miscellaneous, NZ$ 92/93)
76 61 423 19
Table 4.2 - Characteristics of the sample by region; average values per farm in Region I (Waikato-Taranaki)
1997-2005 Average Std Dev Waikato-Taranaki (125 observations) Max Min
Milk Production (total milksolids, kg) 64,704 24,364 132,000 30,000
Factor inputs
Cows in Milk (number) 202 62 372 104 Area (milking platform, ha) 67 23 153 33 Labour (total hrs per year) 3,143 784 4,897 2,250 Feed (all purchased feed, NZ$ 92/93) 28 18 98 6 Fertilizer (expenditure, NZ$ 92/93) 23 11 55 2 Intermediate inputs (health, breeding, shed, feed
and fertilizer expenses) 67 31 191 21 K2 (depreciation and interest on buildings and
machinery and expenditure on repairs and
maintenance, NZ$ 92/93) 40 16 92 17 K9 (expenditure on: repairs and maintenance on
buildings and machinery, fuel, electricity, rates and insurance, administration and miscellaneous, NZ$ 92/93)
Table 4.3 - Characteristics of the sample by region; average values per farm in Region II (Canterbury-Southland)
1997-2005 Average Std Dev Canterbury -Southland (139 observations) Max Min
Milk Production (total milksolids, kg) 208,680 119,897 725,000 52,000
Factor inputs
Cows in Milk (number) 549 284 1,600 158 Area (milking platform, ha) 212 98 555 49 Labour (total hrs per year) 6,753 3,561 22,180 2,266 Feed (all purchased feed, NZ$ 92/93) 106 95 666 12 Fertilizer (expenditure, NZ$ 92/93) 84 62 344 13 Intermediate inputs (health, breeding, shed, feed
and fertilizer expenses) 240 170 1,075 48 K2 (depreciation and interest on buildings and
machinery and expenditure on repairs and
maintenance, NZ$ 92/93) 108 75 470 24 K9 (expenditure on: repairs and maintenance on
buildings and machinery, fuel, electricity, rates and insurance, administration and miscellaneous, NZ$ 92/93)
107 70 423 33