A note on government statistics and unemployment

Update 3 Apr 2009: I am re-posting this post about unemployment figures from November 2008 because I think it relevant. The next unemployment number is coming out on Friday and I expect a large number. If you read this analysis, you can see that it is clearly too conservative because I was talking about 8.7% unemployment by July 2009. Clearly, unemployment will go much higher.

Here’s the original post, posted at 11:45 on 7 Nov 2008:

For those of you who have been reading this blog for a while, you know that I am often very skeptical of government statistics. This is largely because governments often change the methodology on how they calculate their data more than because I suspect outright fraud. However, as you can imagine, most of these changes end up presenting the economic data in a more favorable light. In the U.S., this is certainly true regarding inflation, unemployment and GDP. So, it makes it hard to compare numbers across time and get an accurate sense of what they suggest about the economy.

My goal is to get a sense of where we are in the business cycle and how this particular cycle compares to previous business cycle. Knowing that can help forecast what we should expect going forward in the next months and quarters. Often times the data just don’t allow you to make these forecasts.

As a result, I never use the data as presented to make economic predictions or calculations.

Generally, I do four things to the data:

  1. I focus in on year-to-year data. Doing so eliminates any seasonality you might get in focusing on quarter-to-quarter or month-to-month data. I hate month-to-month comparisons because they don’t tell you anything — it’s like missing the forest for the trees.
  2. I also try to focus on the change, the delta and not the actual number. This is great because it gets rid of the problems resulting from methodology changes. For example, with unemployment, I am not looking at the unemployment rate, but rather the change in the rate from one year to the next. So I can compare a 6.5% unemployment rate today with a 10.4% unemployment rate 26 years ago without worrying about how the government is distorting these figures. The change in unemployment in the last year today is 1.7% whereas in October 1982 it was 2.5%. That’s a much better comparison.
  3. I try to get non-seasonally adjusted data if I can. Most data series are seasonally-adjusted to help people who are looking at month-to-month comparisons. The problem with these adjustments is they may be wrong, giving you a distorted picture of what’s actually happening. Since I don’t use month-to-month comparisons, I don’t need seasonal adjustments.
  4. And sometimes I use average data. If you get a data series that jumps around a lot like jobless claims, comparing a single data point to another single data point can be misleading. So, I will average the data over a long enough period to smooth out those jumps.

In the end, what all of this allows me to do is get a sense of the trend and compare it to previous trends to know how things could look if we follow a similar path to previous business cycles.

For example, today we clocked in at 6.5% for U.S. unemployment. What does that mean?

Well, first of all, this is a seasonally-adjusted figure. The real unadjusted figure is 6.1% (9.469 million unemployed of 155.012 million in the labor force vs. the adjusted 10.080 million unemployed of 155.038 million in the labor force). But when you compare it to where we were 12 months ago, it is 1.7% higher on both counts. How does that compare to previous recessions?

In 2001, we reached a peak 12-month change of 1.8% in December. In 1991, the peak was 1.6% in May. In 1982, the July peak was 2.6%. In 1980, we had a July peak of 2.0% and in 1975 we had a May peak of 3.8%. Going back to World War II, the worst recessions in terms of employment all saw more than a 3% climb in unemployment in a year’s time. That suggests we have much further to go in this recession. If one looks at the 5.7% unemployment rate had in July 2008, 3% more would bring us to 8.7% and 2% more would get us to 7.7% by July 2009. Hopefully, that gives you a sense of where we are headed.

Unfortunately, these are very large numbers that most people are not prepared for. But, the data suggest this is where we are headed.

  1. Anonymous says

    Take these remarks within the context of I believe the govt employment statistics seriously under-state unemployment.

    I’m curious at to why no one makes the connection between rampant immigration (both legal and illegal) and accelerating job losses. No other country has such a liberal policy of ‘let them in at all costs.’

    Legal immigration = 1 million/yr +several hundred thousand temp visas (h1b, L1, etc.).

    If we simply halt all immigration for 1 year we can wipe out what are self inflicted job losses. I’m all for immigration in the good times, but it seems we are over-loading a sinking life boat.

    And please, don’t mention ‘the skills shortage’ myth, billionaires like Gates propagandize.

  2. Terry says

    Great methodological approach! It certainly does eliminate the garbage that goes with seasonality.

    Also, I appreciate your analysis of where unemployment is headed by comparing with earlier recessions. It’s a good way to forecast for the near- to mid-term, if structural aspects of the economy (eg–huge decline in manufacturing) don’t alter the basis for such a comparison.

    Now for the hard part: Employment/unemployment is generally considered a lagging or (at best) a coincident indicator of the economy. Have you tried any successful methods to extrapolate your outlook for employment for the future health of the economy, especially a turnaround?

    OK, it’s Sunday, you don’t need to answer that, but I’d be interested in yours or others attempts to use real labor data to look at the broader economy.

  3. Edward Harrison says

    Terry, I have taken a look at that and I found that the year-on-year change in the unemployment rate generally peaks around the end of recessions. Meaning the unemployment rate continues to rise after recession but its rise happens more slowly.

    For example, the rise peaked in Dec 2001 a month after the recession ended. In 1991, it peaked at 1.6% in Mar, May and Jun. The recession ended in March. In 1982, it peaked in July while the recession ended in November.

    I could site many other examples. But, the overall gist of the data is that even unemployment, which is a lagging indicator, stops rising as quickly — signaling an end to recession.

    If I saw the now 1.7% change in unemployment falling to 1.4% over the next few months, it would signal to me that things might actually be getting better.


  4. Grant says

    Just a month later, and look at the statistic now!

Comments are closed.

This website uses cookies to improve your experience. We'll assume you're ok with this, but you can opt-out if you wish. Accept Read More