How reliable is the Halifax house price to earnings ratio?

Yesterday, I posted a chart illustrating the Halifax house price to earnings ratio. It showed that the ratio in July was 4.33. Furthermore, it is now converging to its long run level of about 4. The Halifax ratio was calculated using an average income of ₤36,576, and an average house price of ₤159,623.

However, one of my blog readers - Powerman - left a nice comment, asking where the Halifax got their estimate of income. He also left a link to the Office of National Statistics, which states that median male full time income is ₤521. This gives an annual total of ₤27,092. With this lower income level, the house price to incomes ratio rises to 5.9.

I took a closer look at the Halifax numbers. It turns out that they are using average annual income of a full time male.

The Halifax are being very naughty using average rather than median incomes. The distribution of income is very skewed. The vast majority of us earn around ₤27,092. There is also a tiny minority of people who pull in millions. When the ONS calculates a simple average, the minority of big earners pull the average up.

This is why the ONS prefers to use the median. That measure is defined as the mid-point, with 50 percent of the population earning less than that amount, and 50 percent earning more. If we use median incomes, we would know that half of the male full time workforce face a house price to earnings ratio of at least 5.9. Not even Northern Rock offered mortgages with those kinds of numbers.

The plain fact is that home ownership remains prohibitively expensive, particularly for young workers who tend to earn bless than the median income. The Halifax house price to earnings ratio neatly obscures this fact by using a biased measure (in a statistical sense).

Finally, I resent the fact that the Halifax are using male full time earnings to calculate their measure of this ratio. Using male income gives a further upward bias to this measure, since men earn more than women. It really should use a non gender specific estimate.