Regular users of government data (like reporters) should know its limitations. Many of the series are highly erratic, meaning that any individual number contains a considerable amount of error and a limited amount of information. The hourly wage data very much fit this bill. In the real world, hourly wage growth doesn�t change very much from month to month. (How could it? � not that many people change jobs in a month; and wages don�t suddenly plunge or soar for workers keeping their jobs.) However, the monthly wage series does show large fluctuations in the rate of hourly wage growth. For example, in April, the average hourly wage reportedly increased by 10 cents, a 0.6 percent increase. Similarly, it reportedly rose by 8 cents in July, a 0.5 percent increase. Before anyone gets too concerned that these wage increases will lead to inflation, let me point out that wages rose by just 1 cent in May and 2 cents in August. The smart folks out there already guessed that the slow wage growth in May and August is directly related to the fast wage growth the prior months. If we imagine that there is some �true� rate of wage growth (e.g. 5 cents a month at the moment) then our monthly data will fluctuate around this true rate because the survey is imperfect. If it randomly shows too much wage growth in some months, as was likely the case in April and July, then it will likely show less than the true rate of wage growth in the following month (May and August). What does this mean? Serious people largely ignore single month wage data. Take a 3-month average. Does it look like wage growth is accelerating or decelerating compared to earlier periods? That is what you can get out of the data. When reporters make a big deal out one month�s wage growth, they are wasting their readers� time.
-- Dean Baker
