Appendix 6: Worker Estimation
Estimating Seasonal Worker Numbers
As one of the key objectives of the project was to provide a robust estimate of the number of seasonal migrant workers engaged in Scottish agriculture several approaches were taken in order to rigorously verify the number of total migrant farm workers. This included using all available sources of information and even looking at satellite images and Google street maps (identifying covered crops – polytunnels, and caravans as a way of verifying / estimating workforce scale). One key way of assessing the data provided through farm business survey and the JAC was to undertake multiple regression analysis to develop estimates of labour requirements per hectare of different crops, which were then compared to other data sources. The advantage of using many different approaches to the estimation process is that the robustness of each approach’s results can be assessed based by how close it is to other approaches.
Appendix 6a: Regression Analysis
Seasonal Migrant Workers and days worked relationship
In order to check the robustness of the survey data, Figure 37 shows that there is a very strong statistical (R 2 =0.94) relationship between the reported number of seasonal migrant workers and the reported number of work days they are employed for. The nature of the best fitting curve reiterates that as the (mostly soft fruit) business size increases there is proportionally more work days undertaken per worker - a result of increased prevalence of processing and packing and extended seasons through more varied cropping mix (something reported during the interviews and survey results).
Survey Regression Seasonal Migrant Labour Coefficient Estimation
A common approach to measuring the strength of association and effect of multifactor data is regression analysis, which is often called a regression model, or just model. Modelling data produces two useful set of statistics in verifying the number of migrant workers in Scotland. Firstly, it provides a parameter of impact for each variable included in the model which is useful for extrapolation to other datasets. Secondly, it provides a measure of model fit telling us how confident we can be in the model.
Figure 37: Minitab – regression analysis of number of reported migrant workers and reported work days undertaken by them
Regression models handle missing data in different ways, so it is important to consider the missing data prior to running models. In order to obtain a robust estimator of the number of migrant workers we first used an ordinary least squares ( OLS) model to estimate the number of workers that each farm would be expected to use – including those that noted that they used contracted labour provider workers to complete all, or some tasks. The full model, which includes variables that measure hectares of crops as predictor variables and total labour hours as the dependent variable, is shown in the following table.
Table 16: OLS Regression estimates with no missing data
|Vegetable for human consumption||-21 (94)|
|Field fruits||-114 (174)|
|Protected crops||751 *** (195)|
|Flowers & bulbs||10 (214)|
|Adj. R 2||0.50|
|p < 0.001, p < 0.01,p < 0.05|
When the above model is run using a method called Non-Negative Least Squares ( NNLS), we get the following parameter estimates: protected crops = 724, flowers and bulbs = 21.3 and potatoes = 3; field fruit, other crops and vegetables for human consumption all equal 1. These values represent work days needed to harvest one hectare.
In an attempt to accurately measure the error due to the missing data, a series of models were created that used imputation methods. Each model uses a slightly different method of data imputation that ranges from less strict to more strict. For more information see van Buuren and Groothuis-Oudshoorn (2012). The conventional approach is to take the average of the imputed model estimates. The results from the imputed models are shown below in Table 17.
Table 17: Regression models with different imputed values
|Model 1||Model 2||Model 3||Model 4||Model 5|
|Vegetable for human consumption||-68.86||83.68||-120.87 *||-44.73||-46.94|
|Protected Crops||493.38 **||699.10 ***||613.85 ***||682.98 ***||738.40 ***|
|Flowers & bulbs||-139.58||-92.45||-72.79||27.43||131.49|
|Adj. R 2||0.36||0.56||0.47||0.53||0.40|
|p < 0.001, p < 0.01, p < 0.05|
In order to provide estimates of per hectare labour requirements from the various regression models convention is to use the average value of each imputed model’s coefficient as final results. These are presented in Table 18along with the values for the non-imputed regression model present in Table 16 (all values are given in non-negative terms).
Table 18: Aggregated regression co-efficient and estimated results (non-negative estimates)
|Crop||Imputed (days/Ha)||No imputation (days/Ha)||Regression (Hours/Ha)||Scottish Hectares||Total Days|
|Flowers & bulbs||21.9||21.2||175.2||964||21,113|
|Vegetable for human consumption||0||0||1||-||-|
Appendix 6b: Standard Labour Requirements and published casual labour requirements
In order to fill in the missing data to corroborate regression estimates from the farm business survey and 2017 JAC, various labour requirements for main horticultural crops were gathered from various published sources. Standard Labour Requirements represent an estimate of the number of hours of work specific agricultural crops and livestock require on an annual basis. The Scottish Government publish  coefficients that include regular farm labour. Table 19 shows these coefficients and includes an estimate of the number of workdays.
Table 19: Scottish Government standardised labour hours per crop (and estimated workforce)
|Hours||Scottish Hectares*||Hours||Workers 4 months||Workers 8 months|
|Outdoor vegetables and salad||280||19,546||684,125||2,851|
|Top and soft fruit ( field fruit)||425||716||38,039||396|
|Vegetables under glass ( soft fruit)||7,000||1,404||1,228,448||12,796|
Andersons (2017) in their estimate of soft fruit labour use, for British Summer Fruits, used farm level data to calculate the median labour use by crop type on an “hours per tonne” basis. Then, using Defra estimates of total fruit output (tonnes) they derived an estimate of the total number of hours per UK crop. Using the assumption that a “ single seasonal worker is employed on average for 22 weeks at 40 hours per week (i.e. 880 hours)” they estimated the required UK workforce (permanent, part time and seasonal workers) for the chosen crops (strawberry, raspberry and other fruits – excluding blackcurrants) was 28,959. This is summarised in Table 20 where the number of hours per hectare and workers per hectare are calculated using the published data alongside Defra estimates of crop areas. It should be noted that blackcurrants were excluded in the Anderson (2017) study, “ as this crop is mainly grown for processing, with both husbandry and harvesting operations now largely mechanised; there is therefore a negligible requirement for seasonal labour”.
Table 20: Anderson’s estimated UK workforce and labour requirement by selected crops
|Hours per Tonne||Tonnes||Hours||Workers||Hectares^||Hours per Hectare`||Workers per hectare`|
|Other Soft Fruit*||375||9,400||3,525,000||4,006||933||3,778||4.3|
|Defra – Horticulture Statistics 2015 |
|*Excluding 2,511ha blackcurrants|
|` own calculation|
Defra 2016 horticulture statistics  suggest that on average (five year average) 5.44 tonnes per hectare are produced across the UK. Whilst a proportion is still harvested by hand for the fresh fruit market (with labour requirements similar to other soft fruits) the bulk of the crop is likely harvested at (excludes husbandry inputs) 0.5 hectares per worker per day.
Using Anderson’s coefficients (and assuming the average fruit seasonal worker is employed for 4 months – from the survey) with the area of crop extracted from the June Agricultural Census (see Table 21) suggests that there are 7,295 workers on fruit farms (excluding blackcurrants).
Table 21: Estimated Scottish workforce using Anderson’s estimates
|Scottish Area||Estimated Hours /Ha||Estimated Workers|
|Strawberries – in the open||77 Ha||4,036||402|
|Raspberries – in the open||127 Ha||3,914||650|
|Blueberries - grown in the open||25 Ha||3,778||122|
|Mixed and other Kinds of Soft Fruit||166 Ha||3,778||819|
|Protected crops - Area of which is strawberries||981 Ha||3,027||3,868|
|Protected crops - Area of which is raspberries||172 Ha||3,028||676|
|Protected crops - Blueberries||159 Ha||3,029||626|
|Protected crops area of which is other fruit||34 Ha||3,031||133|
ADAS (2014), in their assessment of the potential for growing a variety of horticultural crops in Wales, provided gross margin calculations for a number of crops grown in Scotland that could be used as comparator data. The data only includes production and harvesting labour requirements and therefore does not account for any processing, grading and packing that may occur.
Table 22: ADAS estimates of field labour hour per hectare (and estimated Scottish days) required for certain crops
|Production||Harvest||Total||Scottish Hectares*||Days (8 hrs per day)|
The John Nix Pocketbook (Redman, 2017) also provides estimates of the labour cost (and hence, use) for a number of crops, as summarised in Table 23.
Table 23: Nix Pocketbook casual field labour hours per hectare (and estimated Scottish days) required for certain crops
|Hours per Hectare||Scottish Hectares*||Days (8Hrs /day)|
|*June Agricultural Census|
The SAC Farm Management Handbook ( SAC, various) was used to provide gross margin data for selected horticulture crops that included estimates of casual labour use and cost. The estimates of labour used by crop are provided in Table 24.
Table 24: SAC Farm Management Handbook estimates of casual field labour hours (and estimated Scottish days) per hectare required for certain crops
|Hours per Hectare||Scottish Hectares*||Days (8Hrs /day)|
|*June Agricultural Census|