# Same data, different story

Same data, different story

I want to endow you with a healthy scepticism for how data is presented by telling different stories using the same data. Using 10 data points representing the salary of police offices in the period 2006-2015 I will simultaneously show that police officers were well paid, and not paid enough! I haven’t picked on police officers for any particular reason other than it’s a sufficiently relatable profession.

Mean or median?

The first fork in the road is to decide on the mean or median level of pay. Pay researchers tend to go for the median as a small number of high salaries can skew the mean. This is pretty uncontroversial right? Well it can make a real difference. Imagine you are measuring the gender pay gap. The median ignores the small number of high earners who happen to be disproportionately men. The official (the headline measure presented by the ONS) gender pay gap for full-time employees was 9.4% in 2015, but rose to 13.9% when the mean is used.

So we’ve settled on the median. Graph 1 shows the salary over the 10-year period. It rose £4,184 or 11.7%. Pretty good going. Certainly the line appears to be heading in the right direction.

The clever ones among you will be saying yes, but what about inflation? We can’t we tell if this is a real time pay rise or not.

Choosing an inflation index

We’ve now reached a second fork in the road. Which inflation measures to use? There are at least 4 in the running (CPI, CPIH, RPI, and RPIJ). This means we can shop around.

The inflation index chosen can make a big difference especially when dealing in a long time series. Graph 2 shows RPI and CPI inflation as well as Associate Professional and Technical occupations gross median annual pay. As the salary increase falls between CPI and RPI, a real terms pay rise is entirely dependent on whether you accept CPI as the most appropriate measure of inflation. Presenting the data in this way is quite useful because we don’t have to make a decision about which inflation index to use as we can show both.

*Gross annual median full-time pay : 2002 = base 100

Notice how the gap between RPI and CPI widens as time goes on? If we take our police officer’s pay in 2006 (£35,843) and ask how much this is worth in April 2015 prices (to make it comparable with the 2015 salary of £40,027) the answer will depend on the inflation indices used. For CPI the answer is £45,097 whilst for RPI the answer is £44,780. This is a difference of £317 which is not to be sniffed at! Whichever index is chosen, it is clear that police officers have had a real terms pay cut during this period of around £5,000.

Incidentally I was expecting the RPI indexed amount to be larger as RPI is always higher than CPI…. except it isn’t always. Actually during 2006-2015 RPI was higher than CPI every year except in 2009 where it plummeted into minus numbers. Negative inflation (an ungainly term that we should call deflation) was occurring much as happened recently with CPI as a result of the drop in the price of crude oil. What was to blame for the 2009 drop in RPI? In the fall out from the financial crash the base rate was slashed and mortgage interest rates fell to zero (With the first interest rate cut for 7 years occurring this week we might see a similar thing again). Remember all these indices show is a basket of goods be it mortgages or crude oil (or more likely the things that crude oil helps us make; apparently a roast dinner takes at least a pint of crude oil to produce). This basket of goods changes with the times and now includes craft beer, sweet potato, and Spotify. I digress but it’s interesting because we are reminded that this data comes from the real world, and in turn has real world consequences.

I mentioned that the gap between CPI and RPI widens over time. Imagine if the formula used to calculate the increase in your pension pot used RPI instead of CPI. With decades for the wedge to open up the difference over time would be significant. Most pension schemes have now changed from RPI to CPI.

So which one is right? There has been and will continue to be some debate, however the IFS carried out a review and said that RPI was not fit for purpose. RPIJ is an improvement on RPI used by the resolution foundation.

In graph 3 we have both CPI and RPI inflation as well as police pay. Most of the gap between pay and inflation has opened up since 2009 (If the time series went back further than 2006, eventually the CPI and RPI lines would fall below the other lines and show a real terms pay rise). As a point of comparison I have included some other occupations some of which have made more gains and some less than police in the same time period. Police pay has risen more than all Professional Occupations and but less than Receptionists, and Managers & Senior Officials. All of these occupations have seen a real terms fall in pay.

**Gross annual median full-time pay – 2002 = base 100

*notice how the RPI line dips below CPI in 2009 as mortgage interest rates drop which disproportionately affected the RPI index

By adding in a selection of other professions we can add some context. If you wanted to show that a profession (i.e. police officers) were badly paid, you might only include other professions that have had larger gains over the time series. What was pretty obvious to me (when I initially graphed 12 professions) was that most people have been squeezed over this time period.

Finally graph 4 shows the pay of police officer indexed to CPI inflation. At this point it’s worth looking back to graph 1 where police officers pay is now heading in the opposite direction.

The important thing to note is that none of these approaches are right or wrong. As long as the method and sources are properly referenced then anyone is free to interrogate and do their own analysis.

Notes

1. I’ve used data from 14.7a of the Annual Survey of Hours and Earnings. This survey takes a 1% sample of everyone on the PAYE system and the data is used extensively in pay bargaining. In fact, I have used this data to support arguments on both sides of the table in roles at an employers’ association and then a trade union.
2. You may notice some kinks in the lines of my graphs. There was a change in methodology in the ASHE survey in the years 2004, 2006 and 2011 which the ONS refer to as ‘discontinuities’ from which we should exercise caution when interpreting long term trends. The ONS continue to map the ‘discontinued’ data on the same time series graphs so we can take this to mean ‘in the absence of any better data this is still very useful to show in a time series but should be taken with a pinch of salt’. The gaps in the salary lines represent these discontinuities. In later graphs where I had the choice of two data points for the same year I picked one data point and thus the lines are continuous.
3. This analysis isn’t really about police officers pay but how presenting data differently can tell different stories. However, the pay story told here is fairly typical of many professions pay since 2006.
4. I’ve recently been inspired by Andy Haldane to make very colourful charts.