In comments on my post on the July Employment Release, reader Bruce Hall (August 3, 12:50PM) writes:
But, oddly, [Wisconsin] education and government employment increased in Wisconsin by more than 13,000 during the Jan-Jun period under uber-neo-con-anti-government Gov. Walker.
One can get this result by picking the last six month’s of data, and using not seasonally adjusted data from the Wisconsin Department of Workforce Development database. But it seems useful to examine the entire Governor Walker record. The following graphs highlight the misleading nature of Bruce Hall’s comment:
Figure 1: State and local government employment, seasonally adjusted (blue, left scale), and log employment normalized to 2011M01=0 (red line). NBER defined recession date shaded gray. Dashed line at 2011M01 (Walker administration begins). Source: DWD, NBER and author’s calculations.
State and local employment is now 2% below levels recorded in the month Governor Walker took office. Notice how the increase mentioned by Mr. Hall is relative to a distinctive trough.
Figure 2: Log local education employment, seasonally adjusted (blue, left scale), and 12 month log difference of not seasonally adjusted local education employment (red line). NBER defined recession date shaded gray. Dashed line at 2011M01 (Walker administration begins). Seasonal adjustment is calculated by estimating monthly factors on log series, over 2000M01-2012M06 period. Source: DWD, NBER and author’s calculations.
Local education employment (seasonally adjusted) in June is thus more than 2 6 percent below 2011M01 levels (in log terms). Since I have applied my own seasonal adjustment routine (although ARIMA X-12 doesn’t yield anything substantially different), one might want to see how the 12 month change in the not-seasonally-adjusted data looks. Education employment is over 4 percent below a year earlier.
Bottom line: Beware the suspiciously selective use of sample periods (six months is odd), and the use of not seasonally adjusted data. Sometimes one has to deal with nsa data, but then one can try to adjust (say by 12 month differences, which even the most mathematically challenged can do). I cannot speak about motives. But I can question inferences based on such odd selections.
Pwned.
Bruce Hall
Not seasonally adjusting data is especially inexcuseable given that ARIMA X-12 and TRAMO-SEATS are not only included with just about every econometrics package, but are freely available from the Census Bureau and Eurostat. Eurostat’s “Demetra” program even includes both ARIMA X-12 and TRAMO-SEATS all wrapped up inside a easy to use point-and-click GUI along with a good user’s guide.
http://circa.europa.eu/irc/dsis/eurosam/info/data/demetra.htm
Of course, you will need more than 6 months of historical observations to seasonally adjust data using Box-Jenkins type techniques.
Why do we begin analyzing Walker’s economic policy record the month he begins office, but we are still blaming Bush for Obama’s economic record?
I suggest blaming Jim Doyle for the at least the first 12 months of Walker’s term. That would mean we start blaming Walker around the start of 2012 or maybe even wait another year or two.
tj: Thanks for your deep and insightful comments regarding sample selection and seasonality.
I guess I forgot to put (sarc) on my comments.
Odd, again, how seasonally adjusted data is fine in one instance and not for Wisconsin.
As I have written:
Economics: once called the “dismal science” has progressed to the “incomprehensible science.” Two economists can look at the same data and derive opposing conclusions based on their preconceived notions. While the rest of the world sees failed policy, overwhelming government debt, and high unemployment, the economist says:
“… one could argue better output performance could have been achieved with more government spending…. None of the foregoing should be construed to be an argument that the recovery is as strong as it could be. In point of fact, we know that had policymakers implemented a larger stimulus package in early 2009, additional infrastructure investment, as well as more aggressive job creation measures (e.g., tax incentives for new job creation), we would likely have had more rapid growth.”
Gasoline on fire, eh?
Bruce Hall: I understood the sarcasm in your comment. But I am afraid your rebuttal on the statistical point is incomprehensible to me. You write:
You are the one using seasonally unadjusted data — or are you unaware of that? In Figure 1, I report seasonally adjusted data (which is standard practice). In Figure 2, I report seasonally adjusted data (blue line), where I have done the adjustment myself, using EViews. The red line is seasonal adjustment of a sort, where I do 12 month differences, which is standard practice again, where the seasonality is of a monthly regularity (so don’t do it for cases where the lunar new year celebrations are important, but I think I’m safe for Wisconsin).
Hence, I suspect others will also be confused. Please clarify.
I will admit that I may be mistaken… that the subject of deeper recessions should be followed by steeper recoveries… and that argument has been rejected previously by you.
I will admit that I may be mistaken… that the BLS seasonally adjusted data is more reliable than than Wisconsin’s data (or is it that Walker’s use of the data is not convenient?).
Now, as to me using data that is not seasonally adjusted, I can only point to footnote 3 from the BLS page that has the Wisconsin data:
http://www.bls.gov/eag/eag.wi.htm
(3) Number of jobs, in thousands, seasonally adjusted.
… but I will admit that I could be mistaken.
However, as to my statement that two economists can look at the same data and draw conclusions 180° apart, here’s an example:
http://cafehayek.com/2012/06/in-a-complex-system-bias-reigns.html
… but I won’t admit that I could be mistaken.
Bruce Hall: You cite in your original comment “over 13,000”. If you subtract government employment in January from June (seasonally adjusted, in the webpage you cite, you get 9,300. The number you are citing is consistent with the number from the DWD spreadsheet for seasonally unadjusted data, that I hunted down, and provided in this post.
In addition to the selective use of sample period to calculate the change, the point about seasonality is all I wanted to highlight.
So to recap, if you condition on balance sheet conditions, then the statement is wrong on deeper recessions are followed by faster recoveries; under current operations, BLS data is DWD data (unless you argue QCEW is Wisconsin data); finally regarding the Russ Robert post at Cafe Hayek, I agree with the need for humility. But I don’t think he really understands what CBO has done in its quarterly reports (they use a range). In any case, implicitly, he is saying we should ignore all meta-analyses. Well, okay. But what is the recourse? I can see your choice is pick and choose what you like.
By the way, any dispute that total state and local employment is down over 4% since 2011M01? And that local education is down over 6%?
“You cite in your original comment “over more than 13,000″.”
“But, oddly, education and government employment increased in Wisconsin by more than 13,000….”
But why quibble?
“…finally regarding the Russ Robert post at Cafe Hayek, I agree with the need for humility. But I don’t think he really understands what CBO has done in its quarterly reports (they use a range).”
Thanks for verifying my contention.
That’s the great thing about economics. The data can be used in so many ways to reinforce so much bias. Sort of like using trend analysis to show growth when you pick the lowest point of cyclical data over several cycles. The worst you can do is have a flat, no-growth line.
Regardless, I am a loyal reader and I do learn from your posts. I may look a little sideways at them, but that’s my own bias.
Bruce Hall: Thanks for highlighting the error in quoting. I have fixed.
I think you are highlighting the extent of the disagreement much more than it exists. In particular, I suspect the mean and modal estimate of the fiscal multiplier for academic macroeconomists and for business sector economists, when at the zero interest bound, is in excess of zero. (Regarding extreme bounds, I suspect there are a few people out there who believe the sun revolves around the earth.)
Menzie,
Government “stimulus” spending can always result in a short term boost to the economy… apparently quite short given calls for a QE3 or other efforts to pour out more borrowed dollars. The real issue is the long term drag on the economy. The CBO remarked last fall that the various stimulus efforts had some positive effect for 3… maybe 4 years… but will have a NET NEGATIVE effect over 10 years.
So one might ask, if a stimulus hammer is the only tool in your tool box, is every economic question a nail?
Bruce Hall: Yes, which is why you don’t want to do stimulus close to or above full employment (G.W. Bush 2006-07), but when there is slack, so that the negative impact occurs in a period of positive output gap.
It is beyond wonderful that highly intelligent people can look at the venal mess we call our government and sincerely believe that it is capable of reliably, rationally and honestly turning the heat up and down under the world’s largest economy.
Bruce Hall Government “stimulus” spending can always result in a short term boost to the economy… apparently quite short given calls for a QE3 or other efforts to pour out more borrowed dollars.
QE3 is not about increasing fiscal spending. QE3 is about tweaking the long run interest rate in order to (a) hopefully bring it down, and (b) fight off deflation.
The general consensus among Keynesian economists is that for most garden variety recessions the best strategy is to let monetary policy do the heavy lifting with minimal reliance on deficit spending. Or if there is going to be something on the fiscal side, then it should be something along the lines of a “balanced budget multiplier.” The ONLY reason some of us are calling for increased deficit spending is because we remember our IS-LM model and recognize that when the LM curve is flat movement from the IS curve (i.e., fiscal policy) is the only remaining policy tool. And with a zero interest rate and an economy flat on its back, it’s pretty safe to conclude that the LM curve is flat.
I wouldn’t want to rely upon IS-LM analyses if this was the 1990s; but we live in an economic world a lot more like the 1930s than the 1990s, so IS-LM ain’t looking too bad right now. As it is, my old Dornbusch-Fischer undergrad macro textbook is a lot more relevant than most graduate textbooks on my bookshelf.
c thomson: You might call the Wisconsin government under Governor Walker a “venal mess”, but I would not.
The problem has been rightfully perceived that a stimulus means the government distributes the yet-to-be-taxed dollars through lower interest rates or through bank rescues or through some imaginary set of “shovel ready” infrastructure jobs. It truly is as c thomson states a belief “that [government] is capable of reliably, rationally and honestly turning the heat up and down under the world’s largest economy.”
The best stimulus is for the government to let the private sector function in the creation of longer lasting jobs and through incentivizing business investment [something the banks have yet to do with the funds they received].
An interesting perspective:
“The trick is to front-load the stimulus as much as possible while putting in place policies that will tighten both fiscal and monetary policy next year. As terrible as our economic crisis is right now, we don’t want to repeat the errors of the past and set off a new round of stagflation.
For this reason, I think there is a better case for stimulating the economy through tax policy than has been made. Congress can change incentives instantly by, for example, saying that new investments in machinery and equipment made after today would qualify for a 10% Investment Tax Credit, and this measure would be in effect only for investments largely completed this year. Businesses will start placing orders tomorrow. By contrast, it will take many months before spending on public works begins to flow through the economy, and it is very hard to stop it when the economy turns around.
Stimulus based on private investment also has the added virtue of establishing a foundation for future growth, whereas consumption spending does not. As economist Hal Varian of the University of California at Berkeley recently put it, “Private investment is what makes possible future increases in production and consumption. Investment tax credits or other subsidies for private sector investment are not as politically appealing as tax cuts for consumers or increases in government expenditure. But if private investment doesn’t increase, where will the extra consumption come from in the future?”
Bruce Bartlett is a former Treasury Department economist and the author of Reaganomics: Supply-Side Economics in Actionand Impostor: How George W. Bush Bankrupted America and Betrayed the Reagan Legacy. He writes a weekly column for Forbes.com.”
http://www.forbes.com/2009/01/22/stimulus-keynes-taxes-oped-cx_bb_0123bartlett.html
Bruce Hall Oy! The rationale for fiscal stimulus is that the private sector is not willing to invest because there is no expected private sector demand. You’re assuming away the problem, which is why your conclusion doesn’t make sense. To be a bit more technical, the Wicksellian interest rate that the market “wants” is a negative rate. At the margin that’s the rate needed to clear investment demand and the demand for savings. But you really can’t have a negative interest rate…or at least much of one. So something has to give. That “something” is income. That’s how the market equilibrates investment demand and desired savings. Income shrinks. It’s called a recession. The reason some of us are calling for bigger fiscal stimulus is precisely because the market cannot move towards a normal output equilibrium position.
As to investment tax incentives, I’m all for them; however, in order for investment tax credits to be effective they must be temporary. Investment tax credits do not increase total investment over the long run, but they do pull forward the timing of investments. But if investment tax credits become permanent, then there is no reason to pull forward investment. That’s the problem with GOP investment tax credit schemes; they don’t have anything to do with fighting recession and are simply ways to shift the tax burden even further away from capital and more onto labor. The GOP criticized Obama’s investment tax credits because Obama didn’t want to make them permanent. And Obama was right.
Hal Varian is a first rate economist. And I used his microeconomics textbook. But Varian is a micro guy and not a macro guy. If Varian believes that making investment tax credits permanent is smart economics, then he is simply out of his league and quite wrong. If Varian believes investment tax credits increase total investment, then he’s mainly wrong on that count too. At least to a first order approximation, investment tax credits change the timing of the investment, not the total amount of investment. And while private sector investment is critical to future growth, that does not mean we don’t also need public sector investment. It’s a bit mathy, but it can be shown that an economy with both private and public sector investment will have a higher GDP than will an economy with only private sector investment. Like I said, it’s a bit mathy, but the intuitive reason for this result is that private sector investments must always be risk averse whereas public sector investments should be risk neutral.