Seasonal Adjustment (Unemployment)

[Just thinking]

What you are proposing is a form of seasonal adjustment. Not the best way to do it because you lose so much data but seasonal adjustment nonetheless.

Not really ... just looking at trends, using just raw numbers. BTW, I have no problem (as I said earlier) with 4-week averaging, that's fine too.

Also, notice it makes little difference whether you use the peaks or valleys from the top graph ... the overall shape comes out pretty much the same.

 

[Just thinking]

But how are you supposed to figure out from the raw data whether the economy is better right now than it was last May?

You could compare the number of claims vs. hirings, which comes from determining the actual number of unemployed (U3 or U6) numbers each month. You also, of course, have the U3 and U6 numbers themselves. Also, just as with adjusted numbers, you can predict the weekly number of actual claims (at any time of the year), and see how far off the mark you are. If, in May, you're pretty much in line but in December the numbers come out much higher than expected, then I'd guess that December is showing a worse economy.

 

[Just thinking]

And that's why you adjust the numbers, essentially by subtracting off the 30,000 expected Santa's so that if everything happens exactly as the long-term seasonal history suggests, the net result is zero.

That's the best argument I've read so far ... and it makes sense. But is it needed (the adjustment) if you also have at your disposal the true U3 and U6 numbers? From what I understand them to be (U3 and U6), they will incorporate the fewer hired Santas and show up as larger values.

 

[Francesca R]

Yes ... the lower graph's overall shape comes very close to what one can draw going from peak-to-peak on the upper graph

But the level would be persistenty 150K higher or thereabouts, consistent with perma-recession.

which pretty much supports my initial claims.

False but great wordplay.

 

[Francesca R]

I have no problem (as I said earlier) with 4-week averaging, that's fine too.

4 week averaging does not iron out seasons. (Maybe it would for lunar initial claims, dunno.). You would still observe a depression every year-end. Until you figured out that you ought to always mentally adjust the Dec-Jan numbers because they say more about what time of year (season) it is than they do about the state of the labour market.

Which is why seasonal adjustments are done. Happy yet?

 

[Puppycow]

I've never seen the NSA numbers graphed like that.
It's interesting to note that the spike always coincides exactly with the line dividing between years. Something I didn't know.

 

[lomiller]

Not really ... just looking at trends, using just raw numbers. BTW, I have no problem (as I said earlier) with 4-week averaging, that's fine too.

Also, notice it makes little difference whether you use the peaks or valleys from the top graph ... the overall shape comes out pretty much the same.

As I said by looking at just the peaks you are ignoring most of the raw data, and the data you are looking at is seasonally adjusted to the end of December. Looking at the valleys would make no difference, it would just shift the time of year you are looking at.

The reason it's a form of seasonal adjustment is that the peaks and valleys occur at the same time every year, so by picking them off the graph you are comparing that time period to the same time period in other years, which is exactly what seasonal adjustments do except they do it for every data point not just a couple of selected ones.

 

[drkitten]

You could compare the number of claims vs. hirings, which comes from determining the actual number of unemployed (U3 or U6) numbers each month.

[...] Also, just as with adjusted numbers, you can predict the weekly number of actual claims (at any time of the year), and see how far off the mark you are.

Congratulations. You've just re-invented the process of seasonal adjustment.

And the reason that you do it is because a single seasonally adjusted number that takes into account all relevant factors like the U3 numbers against the predicted U3 numbers is more easily published and cited. If you just need a quick snapshot of how the economy is doing (for example, if you're looking at a collection of economic indicators to determine if we're in a recession), you use the single unemployment number. If you need to look at the current employment in detail, you'll ignore the SA numbers and look at its components.

Basically, you're asking why people bother with the DJIA instead of looking at the prices of all 30 of the components -- or the S&P 500 index, or the Russell 2000. There's certainly no more information in the Dow than there is in the component prices. In fact, there's a lot less. But by the same token, there's a lot of noise in the individual components as well, and it's much harder to track 30 numbers than one.