Alternative title: Dog bites man.
The results of a new CNN poll are still very interesting, since the poll is much more thorough than the many other polls which showed, more or less, the same thing.
CNN has a write up where they note that the median respondent seems to think that much more money could be saved by cutting programs which he-she wants to cut. Those programs are foreign aid (as always) and, by a plurality, pensions and benefits for retired government workers.
The median respondent thinks that 10% of the budget is spent on “Aid to foreign countries for international development and humanitarian assistance.” 11% of respondents think that more than 50% of the federal budget is spent on such aid (more than the 5% who correctly answer “less than 1%”).
This is interesting, because it helps us evaluate Duncan Black’s (Atrios’s) hypothesis that people consider of military services directly provided in kind (US soldiers in NATO bases) foreign aid. In fact, 10% while absurd is much lower than numbers for foreign aid from other polls. Simple arithmetic would suggest that the median US adult thinks that well over 10% of the US budget is foreign military aid which fits Black’s hypothesis.
My guess looking at the poll however, is that the context matters more than the precise budget item in question. If the questions were asked in the order the results are presented, respondents were asked about Medicare, Medicaid, Social Security and Military spending before they were asked about foreign development and humanitarian aid. The sum of the median answers to those questions is 85%. This makes the median answer to share on foreign aid of 10% sound very high indeed. The median respondent (a purely theoretical construct which doesn’t correspond to a human being) thinks that 95% of the budget is spent on those programs.
Rather alarmingly the sum of median shares on the programs about which the pollster asked is 137%. Those programs did not include interest on the national debt, farm subsidies, NASA or the NIH.
Here I have a complaint with CNN — they present median answers but do not present mean answers. The logic is that with outliers (such as those who claim that more than half of the budget is foreign aid) the median is a less noisy measure of the central tendency — the mean squared difference between the sample median and the polulation median is much smaller than the mean squared difference between the sample mean and the population mean.
However, the sum of the medians is not the median of the sums. CNN chose not to report which fraction of respondents assigned more than 100% of the federal budget to the listed programs (it is definitely not 0%) or the median sum. If they had reported average answers, it would be very easy to assess the struggle of the average respondent with arithmetic.