Thursday, January 9, 2014

Noah Smith And His Commenters Miss Boat On Fiscal Policy In 2013

I cannot make the link seem to work, but earlier this evening at Noahopinion, he posted on Bad Event Studies, arguing that the supposedly contractionary fiscal policy implied by the sequestration that started after March, 2013, may have provided a bad event study to test the relative strengths of monetary and fiscal policy, particularly market monetarism versus Keynesianism.  He proposed various combinations of outcomes, and in the end basically said we do not know because people may have been acting in earlier years on expectations about what was going to be done in 2013 in monetary and fiscal policies.  However, throughout, he and all the commenters accept the idea that fiscal policy was contractionary during the year.

This is wrong.

As is so often the case, what was going on at the state and local levels was completely ignored.  From June 2009 until the end of 2012, roughly 750,000 government jobs were lost.  Of those about 36,000 were at the federal level, while the rest were at the state and local level, with a good 500,000 of them at the local level.  Quite a few people in fact noticed during this period that the biggest drag on the economy was this downward pull from coming from the fiscal "policy" of the state and local governments aggregated, which of course is not somthing consciously controlled, and hence we do not think of it as "policy."

In 2013, all that changed.  In February, local governments began to hire and by midyear state governments did also.  I was only able to get numbers on this from a Washington Times report of all places in September, which drew on BLS numbers I could not find, but they reported that the combined increase in employment by the two together from February through August was 74,000.  Given that housing prices have continued to rise since then and they play a major role in this, it is near certain that this number was higher by the end of 2013.

I did find from a BLS report in December the total of federal government jobs lost for the year through November. That was 92,000 and can be blamed on the sequester.  Looks like pretty much of a wash.

So, the bottom line is that looking at fiscal policy in its entirety for 2013, it was neither stimulative nor contractionary.  The contraction coming from the federal sequester appears to have been about offset by expansion at the state and local levels.  There is simply no way to use 2013 to test any of this for this reason.

9 comments:

  1. "So, the bottom line is that looking at fiscal policy in its entirety for 2013, it was neither stimulative nor contractionary."

    Since when are the effects of fiscal policy measured purely in terms of their effects on government employment?

    In February the CBO estimated the “equivalent” of about 750,000 fewer full-time jobs would exist by the fourth quarter due to the sequester. The CBO’s estimates did not include a breakdown into federal versus non-federal jobs but I think it’s fairly obvious that the CBO did not believe a large proportion of the jobs lost would be in the federal government. Rather, they would be due to the indirect effect of the decline in federal government outlays.

    According to the CBO’s “Budget and Economic Outlook: Fiscal years 2013 to 2023″, Table 1-2, the Defense Department, nondefense discretionary and Medicare will be cut by 7.9%, 5.8%, and 2.0% respectively. The Defense Department employed about 770,000 civilians in FY2010, the Department of Health and Human Services employed about 70,000, and all other departments, except the Social Security Administration (which is completely exempt), employed about 2.00 million people. Even assuming the federal jobs lost is proportional to the size of the spending cuts in a given department, that yields about 61,000 civilian Defense Department jobs, about 1,400 Health and Human Services jobs lost, and about 116,000 other federal jobs lost, or the equivalent of about 176,000 federal government jobs lost.

    Thus, assuming this estimate is correct, less than a quarter of the estimated “equivalent” jobs lost will be in the federal government. Furthermore, federal agencies will probably avoid furloughs longer than 22 days because they fall under “reductions in force” rules. Without extended furloughs or layoffs, few federal government jobs will actually be lost as your cited estimate of 92,000 federal jobs lost due to the sequestor demonstrates.

    Furthermore, about 70% of the fiscal changes at the federal level in 2013 were in taxes, not spending. In November 2012 the CBO estimated that the maximum level employment effect would be a decrease of about 200,000 jobs and 640,000 jobs (80% 0f combined payroll and UI effect of 800,000 jobs lost) for the high income tax increase and payroll tax increase respectively:

    http://www.cbo.gov/sites/default/files/cbofiles/attachments/11-08-12-FiscalTightening.pdf

    That's a total of about 840,000 jobs lost due to tax increases alone, and I seriously doubt that the CBO thought many, if any, of those would be government jobs.

    ReplyDelete
  2. Excellent comment, Mark, which I have already responded to over on Noahopinion. Anyway, let me repeat the gist here.

    Your data sources at CBO do not take any account of state and local fiscal policies. They are strictly focused on the federal level, as they should be. They work for Congress, and Congress is responsible for federal fiscal policy, not state and local fiscal policies.

    Therefore, your numbers do not account for the offsetting effects at the state and local levels at all.

    I also think your point that lots of fed spending goes to the private sector is not a big deal. The same occurs at the state and local levels as well. Those increased hirings at the state and local levels are being accompanied by increased hirings of private sector employers who do work for those governments.

    Where I shall grant that you have a solid point is on the tax side. So, there were tax increases at the fed level. I did not investigate at the state and local level, but it is my strong impression that there has been little tax cutting at those levels. So, it is highly likely that there on tax effects at the fed level have not been offset. This would suggest that indeed probably the net effect of the total fisc was contractionary in 2013, but still not be nearly as much as all these estimates would indicate that fail to account for what has gone on at the state and local levels.

    ReplyDelete
  3. "Your data sources at CBO do not take any account of state and local fiscal policies. They are strictly focused on the federal level, as they should be. They work for Congress, and Congress is responsible for federal fiscal policy, not state and local fiscal policies."

    The IMF clearly states that their fiscal measures are for "general government".

    In international government finance statistics the phrase "general government" essentially means all levels of government: central, state, regional or provincial; and local. You can find a detailed description of the meaning of "general government" in Chapter 2 of the IMF's Government Finance Statistics Manual:

    http://www.imf.org/external/pubs/ft/gfs/manual/

    If you still have any doubts that the IMF's estimates of US general government finances do not include state and local government, I refer you to "Statistical Table 3. Advanced Economies: General Government Revenue and Expenditure" on page 71 of the IMF Fiscal monitor:

    http://www.imf.org/external/pubs/ft/fm/2013/02/pdf/fm1302.pdf

    Note that US general government expenditures were 38.3% of GDP in calendar year 2013. According to the CBO, Federal outlays for FY 2013 were 20.8% of GDP:

    http://www.cbo.gov/publication/44716

    So obviously this measure includes a lot more than just the federal government.

    At Noahpinion you said the following:

    "Government employment is likely to be a decent proxy for overall direct employment changes due to spending changes."

    This might be true at the state and local level but it's definitely not true at the federal. BEA measures of government spending show that about 60% of all government spending occurs at the federal level. However according to the BLS currently there are about 21.9 million employed in all levels of government in the US, of which only 2.7 million are employed by the federal government. Furthermore employees of the federal government make up only about 2% of all employed workers although federal spending is nearly 21% of GDP.

    As Paul Krugman has argued more than once, the federal government is essentially an insurance company with an army. Proxying the effect of federal spending changes by changes in federal employment, especially in comparison to changes in state and local government employment, simply doesn't make much sense.

    P.S. As long as I'm replying to your reply I should also make some minor corrections.

    "The Defense Department employed about 770,000 civilians in FY2010, the Department of Health and Human Services employed about 70,000, and all other departments, except the Social Security Administration (which is completely exempt), employed about 2.00 million people. Even assuming the federal jobs lost is proportional to the size of the spending cuts in a given department, that yields about 61,000 civilian Defense Department jobs, about 1,400 Health and Human Services jobs lost, and about 116,000 other federal jobs lost, or the equivalent of about 176,000 federal government jobs lost."

    should read

    "The Defense Department employed about 770,000 civilians in FY2010, the Department of Health and Human Services employed about 70,000, and all other departments, except the Social Security Administration (which is completely exempt), employed about 1.93 million people. Even assuming the federal jobs lost is proportional to the size of the spending cuts in a given department, that yields about 61,000 civilian Defense Department jobs, about 1,400 Health and Human Services jobs lost, and about 112,000 other federal jobs lost, or the equivalent of about 174,000 federal government jobs lost."

    ReplyDelete
  4. Mark,

    Your measure of impact is change in balance. But as state and local governments generally run balanced budgets, that does not capture their impact. It remains that the IMF numbers do not reflect the stimulative impact of the rising spending and hiring at the state and local levels.

    ReplyDelete
  5. "It remains that the IMF numbers do not reflect the stimulative impact of the rising spending and hiring at the state and local levels."

    This is based on the highly debatable assumption that the stimulative effect of increased state and local spending significantly exceeds the contractionary effect of increased state and local revenue.

    ReplyDelete
  6. Not so debatable. Balanced budget multiplier is 1, meaning spending changes outweigh equal tax changes. If both go up by the same amount, there is stimulus of AD equal to amount of that increase.

    ReplyDelete
  7. "Balanced budget multiplier is 1, meaning spending changes outweigh equal tax changes."

    The Old Keynesian balanced budget multiplier is derived using a very simple model based on highly unrealistic assumptions. Let's model this using more realistic estimates of fiscal policy effects.

    First of all we need some multipliers. I want to keep this simple so I just want a spending and a tax change multiplier. Romer and Bernstein's "Job Impact of the American Recovery and Reinvestment Act" used a multiplier of about 1.6 for spending and 1.0 for tax changes (see Appendix):

    http://otrans.3cdn.net/45593e8ecbd339d074_l3m6bt1te.pdf

    Let's assume all changes to the CAPB were spending changes except for 2013 which we'll assume was two thirds tax changes and one third spending changes. Thus the multiplier for the 2013 CAPB change will be 1.2.

    And let's assume that all changes to state and local fiscal policy were balanced budget changes. Then any reduction in state and local spending would have a multiplier of 1.6-1.0=0.6.

    Taking the CAPB changes from 2010 through 2013 which were (-0.2), 1.0, 1.1 and 2.3 respectively and multiplying by their respective multipliers yields an economic effect of (-0.32), 1.60, 1.76 and 2.76 respectively (where negative is expansionary and positive is contractionary).

    Now we need an estimate for the size of the reduction in state and local spending relative to potential GDP. Here is a graph of state and local spending and revenue as a percent of potential GDP:

    http://research.stlouisfed.org/fred2/graph/?graph_id=154516&category_id=0

    State and local spending fell from 15.7% of potential GDP in 2009 to 15.4% in 2010 to 14.7% in 2011 to 14.4% in 2012 and to an average of 14.1% in 2013. Thus the reductions in state and local spending as a percent of potential GDP are 0.3, 0.7, 0.3 and 0.3 in 2010 through 2013 respectively.

    Multiplying by the balanced fiscal multiplier of 0.6 yields an economic effect of 0.18, 0.42, 0.18 and 0.18 for 2010 through 2013 respectively. Adding these to the economic effects of the CAPB changes yields an economic effect of (-0.14), 2.02, 1.94 and 2.94 for 2010, 2011, 2012 and 2013 respectively.

    In short, even using a balanced budget multiplier, and using what I think are very generous assumptions, the fiscal changes in 2013 were significantly more contractionary than in the previous two years. Furthermore, even in its most contractionary year, state and local spending reductions only contributed about one fifth of the negative economic impact.

    ReplyDelete
  8. Mark,

    I am genuinely impressed by the effort you have put into replying to my concern. As an aesthetic matter I would change the signs on those net effects, but I am not going to dispute the numbers any further, although I could, but what I would challenge would in effect amount to second order effects. I think you have it pretty close by now. Congratulations, really.

    That said, let us take your numbers and look at them for their interpretation. So, the big shift was indeed from 2009 to 2010, which we all know, or should, by now. More than 2%. The next two were pretty close. The final shift, the one in question, ends up being a 1% shift. OK, I can accept that this may be pretty close to the net effect (negative of course) of the net total fiscal policy effect. So, I grant that I definitely overstated things in my initial post saying "no effect." But while this is indeed significant, it looks to me to be somewhat less overwhelming than has been assumed and accepted by most involved in this discussion.

    Thank you for really nailing this down about as closely as is likely to be done, Mark. Genuine respect and congratulations.

    ReplyDelete
  9. news analytics



    The event study methodology captures stock market reactions to discrete events. This methodology has become a ubiquitious instrument for investigating the financial impact of both market events and corporate decisions. www.eventstudytools.com provides research apps and services that simplify conducting event studies.

    ReplyDelete

Spam and gaslight comments will be deleted.