Assume
that we want to target only the worst 5% of cases. Let us look initially at a map of
the number of males unemployed in the pre-1984 County of Humberside.
The blue
areas show the worst 5% of areas. The largest blob is Hull to the north of the
Humber Estuary. The northern most blob of blue is Bridlington.
The distribution of worst cases is very similar to the distribution of population with
concentrations in the urban centres. Numbers unemployed is known to have an urban
bias because the numbers are constrained by the number of economically active males. This
is why ratios are widely used in comparative studies. They are believed to
standardise data against the base population.
However, if we look at a map of % unemployed males, it is quite obvious that we now
have a rural bias. This was pointed out by a Polish geomorphologist, Choynowski, in 1959.
It is easier to see the urban and rural
biases if we look at the scatterplot below of % unemployed against numbers unemployed.
Here
we note two extreme ratios which stand out as outliers. If we were able to point to
such outliers and locate them on the above maps, we will see that they are the ones with
the two largest squares in the ratio map. If we could also retrieve the topographic
maps for 1971, we will soon find that these squares had male borstals in 1971. So
the outliers are due to errors in data entry.
Such errors in data entry are not uncommon
in large data sets. We need procedures for detecting and correcting them. Instead,
data are aggregated up for for wards, districts and counties, where errors remain averaged
and hidden in many analyses.
If we ignore the 2 extreme values, we see two trends reflecting the existence of rural and
urban unemployment. They are measured respectively by % unemployed and number
unemployed .
