Andrew Leigh is reporting on 1999 research showing that many high income earners wrongly place themselves in lower income deciles and many low income earners place themselves in a higher income decile than is justified by their actual income (also cross-posted at Core Economics).
In the past (p.16) I have used this data to suggest that some people who agree to survey propositions that above-average income earners pay more tax – as 41% of people are in the latest Essential Research survey – may get a nasty shock when they find the taxman raiding their wallets.
While I still think this is likely to be the case, asking people to put themselves into the correct income decile is a big ask. I would expect more general questions such as average, below average, or above average would yield more accurate results. Using data from the 2005 Australian Survey of Social Attitudes and comparing it to 2006 census household family income data I found that accuracy improved but significant discrepancies remained.
(The image is not entirely clear: the three horizontal axis labels are below median <$52,000; median $52,000-$77,999, and above median <$78,000)
There are some difficulties in making these comparisions. On my calculations the median census household family income bracket is $52,000 to $62,000 a year. The AuSSA bracket was $52,000 to $77,999 a year, so some above median income households families earning $62,000 to $78,000 a year are deemed to be correctly classifying themselves as ‘average’. Also, respondents’ impressions are probably based on inferring income from apparent living standards, which due to factors such as household size, borrowing, and accumulated assets have an at best approximate relationship to income in a given year.
But in rough terms, it appears that most average and above average household income respondents correctly classify themselves, while those on below average incomes are less accurate. However given that there would be many retired people in this group, who have paid off their homes, are no longer supporting children, and receive generous treatment by the tax and welfare system, their living standard perceptions may be reliable enough.
Andrew Norton 
I guess it’s a case of glass half-full vs half-empty, but your results seem to show that even when we use only 3 income groups, only about 60% of the population can correctly determine which they fall into. Would you say 60% was a good score if this was (say) an examination of basic literacy or numeracy?
LikeLike
(Sorry, forgot to say well done for crunching the numbers!)
LikeLike
Andrew – Given the inherent difficulties in inferring the incomes of other people 60% right is not too bad in my view. Unlike literacy or numeracy, it is not something the average person needs to know.
LikeLike
Andrew – And we would probably get better results still with a more relaxed view of ‘average’ (or median, as I used), given that there is a reasonable amount of clustering around the middle. When someone answers ‘average’ to a question like this they probably mean reasonably typical of the general community, while knowing that this covers a range of actual income levels.
LikeLike
Well done indeed for crunching the numbers but can I suggest you avoid 3d column graphs? They tend to obscure rather than reveal information – in fact that’s exactly why this one is “not entirely clear”.
BTW, a must-read for anyone wanting to present data in understandable ways is Tufte’s famous “The Visual Display of Quantitative Data”. You’ll do graphics differently after reading it – I know I did.
LikeLike
DD – I can avoid 3D, but I find it easy to read myself. I’ve been using Google Documents for blog figures because it easily saves into jpeg files, but it has less flexibility than Excel in design.
LikeLike
If you only need simple pictures and want them to be in .jpg or .bmp format, another way to do it is to do it on whatever you think looks best and is easy to use and then take a picture of the screen area — I often do this for powerpoint presentations.
There is free and really easy to use program called MWSNAP. You can get it here if you want.
LikeLike
Conrad – Thanks. That should make it easier to use Excel images (not that they are the most attractive, but that’s the format my data is usually in to begin with).
LikeLike