“Okun’s law” is a much-loved rule of thumb — it links increases in the unemployment rate with decreases in output.? The red dots in the chart below illustrate?Brad DeLong‘s version of this rule, which relates the change in output over the past eight quarters with the change in the unemployment rate.? Most of these dots lie pretty close to the dashed line, which suggests a stable relationship.? Based on this rule, and the relatively mild decline in measured output, you might have expected the unemployment rate to have risen by 3.5 percentage points over the past two years, to about 8 percent.? But the dots at the top left show what actually happened-unemployment rose by something closer to 5.5 percentage points, and the unemployment rate is closer to 10 percent.? That’s a big difference.? And it has led?many commentators to ask whether Okun’s Law is broken.
But perhaps the problem isn’t Okun’s Law.?? Perhaps the problem is how we measure output growth.? In fact, there are two measures of output growth-the usual measure, which adds up total spending in the economy, and the alternative, which adds up total income.? In theory, the two should be exactly the same. In practice, they have been very different during this recession.? The blue dots show recent changes in this alternative measure of output.? These GDI numbers suggest that output growth actually declined much more sharply than had been widely understood.? Based on this alternative measure, the recent sharp rise in unemployment is no mystery at all. (Indeed, the 2008 data suggest that the real mystery may be why it didn’t rise faster, earlier.)
There’s a simpler way to show all this: Let’s map out Okun’s Law using this alternative income-based measure of output.? That’s what is shown in the figure below.? The rise in unemployment seems quite consistent with these alternative output data. If anything, the puzzle now appears to be why unemployment didn’t rise by more in early 2008, given the very weak state of this income-based measure of output.
What’s good news for Okun’s law, though, is bad news for the economy.? This alternative measure of output growth suggests that the recession may have been deeper, and longer-lasting than previously thought, although data for the fourth quarter aren’t yet available. While many economists believe the recession ended in the second quarter of 2009, this income-based measure of output kept shrinking in the third quarter, too.? And while the expenditure-based measure is back to its level from the third quarter of 2006, the income-based measure suggests that output is still 3.5 percent below that level.? That’s a pretty big hole to dig out of.
Which is the right measure of output to focus on?? It’s still an open question, but some?interesting recent research by the Fed’s?Jeremy Nalewaik suggests that we should be thinking harder about the income-based measure. And Jeremy has promised further new results, which he’ll present at the?forthcoming Brookings Panel.
and 
. You can also watch it on 
What is the difference in how total expenditures v total income is measured. Why are they not the same?
Productivity goes way up when your export jobs to slave-wage countries. Expentitures here don’t go back into our economy anymore. The multiplier effect is zero now.
Also, there is considerable problems in measuring unemployment in the US because of limited unemployment benefits in the US. Most people find it quite difficult to admit that they are unemployed and will do so only if there is some benefit (i.e. unemployment benefits). If the US instituted a small benefit such as a very small utility, gas or public transit ($5 perhaps) rebate that would automatically be removed once a person was employed, we could get a much better handle on current unemployment statistics and compare them with other countries unemployment statistics.
It looks to me like there is a time lag in the unemployment: when the curve is moving up and to the left, it is below the dotted line and when it is moving down and to the right it is above it. (This is just a guess: I can’t tell which direction the to high/right loops are moving in.)
Try replotting with various time lags: e.g. a 2 quarters lag would plot a point for GDP change 2006Q1-2008Q1 vs unemployment change 2006Q3-2008Q3. Perhaps one of these time offsets will give a much tighter correlation. (It would also be more useful: it would let you predict the future instead of predicting the present.) (Also: are the GDP figures adjusted for inflation? What effect would it have if you changed this?)
Interestingly, the dotted line is above/right of the (0,0) point, indicating some combination of increasing population and increasing GDP per capita (which in turn could be increased efficiency and/or inflation.) I think it would be better to plot GDP per capita (or maybe GDP per working population.) The unemployment axis is per-(working)population, so it would be cleaner (and directly show the trend in per-worker GDP increase) if the GDP axis was too.
Finally, I note the fitted line does not have a slope of one: 20% in the change-in-GDP axis corresponds to 7% on the change-in-unemployment axis. This suggests that either the most efficient (highest GDP production) workers are laid off first, or (more likely) that when unemployment is high, even those who still have jobs have decreased GDP output. (Or at least decreased relative to the secular trend towards increased GDP output over time.)
I am not an economist, so please excuse any error or misuse of terms in the above.
Damn, now I gotta go tell my friends and family I was wrong when I said “Okun’s law is broken.”
Dr J and MW both make excellent points. The fluctuations in the savings rate is a point worthy of consideration. Also, to point to what MW was saying, there is definitely a lag in unemployment that might indicate the numbers actually make more sense than they do when presented in the article. Of course, perhaps the “lag” is non-existant. Personally, I side with the originally posted (by MW) notion. Okun’s Law has historically been a very consistent measure, and the latest recession is not as much of a statistical anomaly as, say, the stagflation in the 70′s was (caused by bottle-necking of supply in the oil market, which has been run by ((basically)) a cartel since the flood!) you could easily argue that given the cybernation in conjunction with outsourcing that has occurred since the 70′s, that indeed unemployment tends to lag a bit against output. If this was weighed into the model described in the article- as MW stated- we might find a more uniform distribution which adheres to Okun’s law better.