Conservative unemployment analysis: an update
Earlier today Left Foot Forward published a short analysis I undertook of the Conservatives’ claim that “the rise in unemployment since the beginning of this recession has outstripped that in the last three recessions”. My paper points out that we entered this recession with lower unemployment rates than the 1980s and 1990s, so actual rates are lower now than they were then. I also considered the speed at which actual rates have increased, and found that it’s pretty much the same as previous recent recessions, but that unemployment rates accelerated a little faster in the 80s. Finally I looked at annual change in unemployment rates, and again found faster monthly increases in the 80s than currently. The Office of the Shadow Chancellor has now been in touch with me to respond to my queries, which I sent last week, and I have posted details of their response below.
The explanation of the index that I’ve been provided with is set out below – the notes are my own:
“1) The base data is the monthly ONS unemployment rate from 1971 through to May 2009 (latest available at time chart produced).”
Note: As the chart is marked “Labour Force Statistics September 2009” I had previously assumed that the data covered the period until June 2009. This is because the September release includes data for the May-July rolling quarter, and monthly data for May and June. I have updated my analysis, but there is no significant change.
“2) The annual percentage change in the unemployment rate was then calculated for each month.”
Note: I have assumed that this means that annual change in unemployment was calculated by deducting the unemployment rate of 12 months previously from the month in question (e.g. annual change this month would be the rate of October 2009 minus the rate of October 2008). The chart in ‘Get Britain Working’ is labelled ‘start of each recession’, so the wider analysis in my paper considers the 12 months from the start of each recession, rather than the 12 months from which unemployment started to rise (the two dates are not always the same). I have therefore updated my analysis of the speed of increase in actual unemployment rates (which showed a slightly faster increase in the 80s than currently), and it is not significantly affected by this change.
“3) The four recessionary periods were then isolated – the start month for each was the first month unemployment increased following the on-set of recession plus 13 consecutive months”
Note: I have assumed that this means the starting point for the index was the first point that annual change in unemployment increased, but I may be wrong about this.
“4) Each of the four periods were then indexed in order to show the comparative rates of change of the unemployment rate.”
Note: I assume that this means comparative rates of change in rates of ‘annual change in the unemployment rate’ – comparative rates of change in the actual unemployment rate are shown in my paper, where the 80s saw slightly faster increases than currently.
“This is the standard approach used when comparing macroeconomic trends during different time periods.”
Note: It is absolutely true that indexing is a standard approach, but I do think it’s fair to question whether measuring unemployment by considering rates of annual change is a standard approach – elsewhere in ‘Get Britain Working‘ more straightforward analysis is used, for example a chart considering regional unemployment looks at how actual unemployment rates have increased over the year. I have been analysing the unemployment data since the start of this recession, and haven’t yet come accross anyone else presenting the information in this type of index.
Following the explanation, and in accordance with the notes I have set out, I have replicated the index – but still can’t re-create the chart. For those who care about the detail the problem that I have is that in the 1990s the rate of annual change during the first month included in the analysis was 0.1. When you index rates for the next 12 months to this rate (which basically means calculating the monthly percentage increase from a base of 0.1) you find a steep curve – if the next month the rate of change is 0.3, this is an increase of 300%, if the month after is 0.5 this is an increase of 500% etc. However, on my index the 1980s show a much more gradual increase than on the Conservative chart. This is because the starting rate of annual change (in accordance with the time periods set out above) was 0.4. So, if the next month is 0.7 and the month after 0.8, these are increases of 175% and 200% respectively. Despite the fact that the 1980s had higher overall rates of annual change than the 1990s, the use of an index in this way shows the 1990s to have experienced steeper rises than the 1980s. Clearly I am undertaking different calculations to the Conservative team, so I look forward to finding out more.
But putting the matter of replicating the chart aside, my wider point is still also outstanding. I remain convinced that the best way to consider how unemployment has changed across recessions is to look at monthly changes in actual unemployment rates. I therefore find it hard to understand why considering how annual change in unemployment has increased during recessions relative to annual change in unemployment at the start of each recession could be the best way to compare unemployment trends.
To reiterate, I may very well be wrong about what the chart is intended to show, or how the Conservative analysis has been undertaken. I have therefore asked for further clarification which I will post it here if I receive it.