While reading an excellent new book by Peter Conti-Brown, I came across this interesting observation:

Liberals like Tobin, Galbraith, and Harris weren’t the only ones who thought a policy separation problematic. Milton Friedman, the great monetary theorist on the right, thought central bank independence much less desirable than a straightforward monetary policy rule that increased the money supply at an agreed-on rate.

I’d never quite thought of things that way.  He suggests that back in the 1960s, many liberal economists favored a supportive Fed.  Let’s think about three ways the Fed could have interacted with the fiscal authorities during the Kennedy/Johnson fiscal expansion:

1.  The Fed could have supported the fiscal authorities by holding interest rates at a low level, despite the fiscal stimulus.

2.  The Fed could have maintained a neutral policy, with a 4% money supply growth rule.

3.  The Fed could have sabotaged fiscal stimulus with a 2% inflation target.

Policy #1 would be the most expansionary.  The second option would also be somewhat expansionary, as fiscal stimulus would push up nominal interest rates, and also velocity.

Under option #3, however, monetary policy would be not at all stimulative, and would essentially neutralize the impact of fiscal stimulus.  And I find that to be a rather odd way of looking at things.  In this framework, Milton Friedman’s monetarism is the “moderate” position.  (Quite a contrast to the rules vs. discretion debate, where Friedman took an extreme position.)  Even more surprisingly, you find Keynesians at each extreme.  The older 1960s Keynesians favored a supportive monetary policy, and the new Keynesians of the 1990s wanted monetary policy to sabotage fiscal policy, in order to keep inflation at 2%.

Once I started looking at things this way, I noticed an uncanny similarity to the three ways that one could categorize monetary policy interrelationships under an international gold standard.  In my book on the Great Depression, I did not look at the interaction of monetary and fiscal policy, but rather the way various central banks reacted to moves at the dominant central bank.  For instance, in the late 1800s the Bank of England was dominant.  Other central banks could be supportive, neutral, or sabotage discretionary actions by the BoE.  In 1930, Keynes argued that they tended to be supportive:

During the latter half of the nineteenth century the influence of London on credit conditions throughout the world was so predominant that the Bank of England could almost have claimed to be the conductor of the international orchestra. By modifying the terms on which she was prepared to lend, aided by her own readiness to vary the volume of her gold reserves and the unreadiness of other central banks to vary the volumes of theirs, she could to a large extent determine the credit conditions prevailing elsewhere. (1930 [1953], II, pp. 306–07, emphasis added)

Here’s how I interpreted this issue in my book on the Depression:

If Keynes were correct then the Bank of England would have had enormous leverage over the world’s money supply.[1] For example, assume the Bank of England doubled its gold reserve ratio. This would represent a contractionary policy, which would then lead to a gold inflow to Britain. If other countries wished to avoid an outflow of gold they would have had to adopt equally contractionary monetary policies. In that case the change in the Bank of England’s gold reserve ratio would have generated a proportional shift in the world gold reserve ratio.

Although it is unreasonable to assume that foreign gold stocks were not allowed to change at all, during the interwar period some countries did not allow significant fluctuations in their monetary gold stocks.[1] Other central banks may have varied their gold holdings, but not in response to discretionary policy decisions taken by the Bank of England. Thus it is quite possible that the estimates in Table 13.5 significantly understate the ability of central banks to engage in discretionary monetary policies.

An alternative form of interdependence occurs when countries refuse to allow gold flows to influence their domestic money supplies. If one country reduces its gold reserve ratio, other countries can offset or “sterilize” this policy by raising their own gold reserve ratios. Complete sterilization would occur if the other countries raised their gold reserve ratios enough to prevent any change in the world money supply.

If central bank policies are interdependent, then there are a variety of ways in which a change in one country’s gold reserve ratio might impact the world gold reserve ratio:

1. d(ln r)/d(ln r_i) = 0 (the complete sterilization case)
2. d(ln r)/d(ln r_i) = G_i/G (the policy independence case)
3. d(ln r)/d(ln r_i) = 1 (the extreme Keynesian case)

In Table 13.5 the estimated impact of central bank policy changes was computed under the “policy independence” assumption (i.e., that a change in one country’s gold reserve ratio had no impact on gold reserve ratios in other countries).

[1] There also may have been an asymmetrical response, with central banks being more reluctant to allow gold outflows than gold inflows. Note that the unwillingness of central banks to vary their gold holdings does not necessarily imply an unwillingness to freely exchange currency for gold. They could set their discount rate at a level to prevent gold flows.

[1] McCloskey and Zecher (1976) criticized Keynes’s assertion by noting that the Bank of England held a gold stock equal to only 0.5 percent of total world reserves. This would seem too small to allow the Bank of England any significant influence over the world money supply (or price level). There are several problems with their criticism. They have assumed that central banks were indifferent between holding gold or Bank of England notes as reserves. Yet many countries placed great importance on their gold stocks, and in some cases laws specified a minimum gold reserve ratio. Thus the relevant size variable is the ratio of England’s monetary gold stock to the world’s monetary gold stock, not the ratio of England’s gold stock to total world reserves. More importantly, McCloskey and Zecher ignored Keynes’s assumption that other central banks were unwilling to vary their reserves of gold (as the rules of the game required).

These three cases offer an interesting parallel to the three types of fiscal/monetary coordination discussed above.  The “extreme Keynesian case” (relying on his 1930 hypothesis) is equivalent to the supportive central bank preferred by 1960s Keynesians.  The policy independence case occurs if countries follow the “rules of the game”, i.e. they must allow their money supplies to move in proportion to their monetary gold stocks.  This case is obviously similar to Friedman’s money supply rule.  And the complete sterilization case is equivalent to a central bank that sabotages fiscal actions. Indeed another term I could have used for “monetary offset” is “monetary sterilization of fiscal policy”.

PS.  This post is rather tangential to the main thrust of the Conti-Brown book.  I also have a new post on his book at Econlog, which gives a better sense of the book’s focus.

Each day my time management spirals more and more out of control.  If only there were 73 hours in a day.  And now there are new “must listen to” podcasts by Tyler Cowen and David Beckworth.  I finally caught up with David’s two most recent offerings. After starting off his podcast series interviewing me, he dug up a couple of obscure economists named John Taylor and John Cochrane.

Not surprisingly, the Taylor interview focused on the Taylor Rule.  I have mixed feelings about this rule.  I do think that something close to the Taylor Rule (not exactly the same) largely explains the Great Moderation.  However, after 2008 I became increasingly disillusioned with this approach, for a couple reasons.  First, it’s hard to know what to do when rates hit zero.  And second, it looks like the equilibrium real interest rate might be more unstable than we had assumed.

I still favor the concept of policy rules, but would prefer an instrument with no zero bound issues, such as Singapore’s exchange rate instrument, or the quantity of base money, or even better, the price of NGDP futures contracts.

For me, the most interesting part of the interview was the discussion of David’s “monetary superpower” theory of the Fed, which Taylor thought had a lot of merit.

The Cochrane interview was focused on the Fiscal Theory of the Price Level, and not surprisingly I have a lot of reservations about this idea, except for countries like Zimbabwe or Venezuela, where it’s clear the fiscal authorities are the dog and the central bank is the tail.  In the US, however, I believe it’s exactly the opposite. Let me respond to a few of Cochrane’s points, and apologize ahead of time if I’ve mischaracterized his views.

1. Cochrane contrasts the FTPL with traditional monetary theories, which imply that a swap of base money for government bonds has important macroeconomic effects. He suggests it’s more like exchanging two fives for a ten dollar bill, just swapping one government liability for another.  But if so, why is it that open market operations (OMOs) clearly do have important macro effects?  The Fed used OMOs to keep inflation close to 2% during the Great Moderation.  Asset markets respond to changes in monetary policy (even before IOR in 2008) in a way that suggests it has powerful effects on the economy.  Global stock indices can move 2%, 3% or even 5% in just minutes after an unexpected quarter point change in the fed funds target.  And until 2008, those changes in the fed funds target merely reflected the impact of OMOs.  So I don’t quite get the FTPL.

2. Cochrane says that in the FTPL the entire future path of government debt matters, not just the current level of debt, and contrasts that with what he calls the M*V = P*Y approach, which (he says) focuses on the current money supply.  But in modern monetary models (market monetarist or New Keynesian) the entire future path of the money supply matters for current prices.

3.  He doesn’t think 30-year T-bonds are currently a good investment, because he’s pessimistic about our future fiscal policies.  And he doesn’t think the Fed would be able to prevent the future inflation that would result from a debt situation that deteriorated sharply.  I’m also pessimistic about our fiscal situation, but believe the Fed can offset the effects of deficits.  Hence I don’t expect much inflation.  My views are consistent with current market expectations, whereas Cochrane seems to have doubts about market efficiency, at least in the case of 30-year T-bonds.

4.  I also read the historical record differently than Cochrane.  I don’t believe the FTPL can explain either the onset of the Great Inflation, or its end.  I don’t believe the common myth that LBJ ran big budget deficits during Vietnam.  I would note that LBJ raised taxes in 1968, and yet inflation continued to rise, because it’s monetary policy that drives inflation.  I also don’t agree with Cochrane’s view that Reaganomics made the deficit situation look better after 1981, because it increased expected future growth. Reaganomics led to much bigger budget deficits, and so if I had believed in the FTPL, I would have forecast higher inflation in 1981, not the much lower inflation we actually got.  Unlike Cochrane, I don’t believe that Reaganomics unleashed a surge in real economic growth, even though I am a fan of Reaganomics, and believe it did lead to higher growth that we would have had under alternative policy regimes.

To clarify, I believe growth rates in many other developed countries fell below US levels after 1982, precisely because their policies were less neoliberal.  Thus in criticizing Cochrane’s claims about Reaganomics, I’m actually starting from a much more sympathetic position that someone like Paul Krugman.  To summarize, I do not believe that the FTPL offers any explanation for the onset of the Great Inflation, or its end.  Monetary policy clearly explains the onset (base growth rates soared), and can also explain the end with a bit more difficulty (money growth was strong in the early 1980s, as lower inflation led to a one time rise in velocity.)

John Cochrane favors the same sort of market approach to monetary targeting as we MMs favor.  His specific suggestion was something like Robert Hetzel’s 1989 proposal to target TIPS spreads, although you could also use CPI futures prices.  He basically views this as a more sophisticated version of the gold standard, which has always been my view as well.  Unlike Cochrane, however, I believe the CPI is essentially meaningless; it doesn’t track anything meaningful in any economic model.  (No, it’s not the average price at which stuff sells.  The CPI is based on hedonics, which have no role to play in macro.)  Thus I think the most useful definition of the “value of money” is 1/NGDP, that is, the share of nominal output that can be purchased with a dollar bill.  That’s the variable I want stabilized, along a 3% or 4% per capita growth path.  To his credit, Cochrane favors level targeting, and so do I.

Oddly, Cochrane and I end up with fairly similar views on the optimal monetary policy, despite having radically different views on the theory of money.  Indeed I favor having the Fed use OMOs with ordinary T-bonds as a way of stabilizing NGDP futures prices, and as far as I can tell Cochrane doesn’t think OMOs do anything. Unless I’m mistaken, Cochrane thinks the gold standard only worked because central banks promised to buy and sell unlimited amounts of gold at the legal price, or at least something other than T-bonds.  (To head off objections from George Selgin, you don’t need a central bank; private banknote issuers can also assure redeemability into gold.)

In contrast, I believe the central bank could have pegged the price of gold merely using OMOs with Treasuries.  Alternatively, in the 1990s the Fed could have used the Taylor rule without buying any bonds at all.  It was the liquidity effect from monetary injections that caused fed funds rates to change, and hence the Fed could have bought and sold gold or zinc to target fed funds prices, and used that instrument to keep inflation close to 2% via the Taylor Rule. In other words, it’s all about the Fed’s liability (base money); the asset side of the balance sheet hardly matters, at least when not at the zero bound, and hence when OMOs are small in size.

PS.  Over at Econlog, I have a new post discussing a NBER study that is critical of the New Keynesian claim that artificial attempts to raise wages can be expansionary.  Their result supports my claims in The Midas Paradox.

PPS. George Selgin is putting together a very nice series on monetary economics.  The second entry discusses the demand for money, and why the price of money and the price of credit are two entirely different things.

Ben Bernanke has done a series of posts on what central banks can do at the zero bound. His first post looked at negative IOR, and the second examined targeting long-term interest rates.  Of course Bernanke has also advocated the use of QE. Now he looks at the helicopter drop option. Bernanke agrees with my view that helicopter money should be used as a last resort.  Where we may differ slightly is how many options need to be tried before reaching that point.

In my view, it’s too soon to jump to helicopter money, just because the techniques mentioned by Bernanke have been exhausted.  I recall Bernanke once arguing that the inflation target might have to be raised if there was a danger of hitting the zero bound, but he doesn’t mention that here. In my view there are many alternatives that we’d need to run through before considering helicopter drops, such as a higher inflation target, or price level targeting, or better yet NGDPLT.  I’d also want to go beyond T-bond purchases, to the purchase of other assets.  Thus creation of a sovereign wealth fund would be far superior to helicopter drops.

For some reason Bernanke doesn’t consider those alternatives, perhaps because he doesn’t think they would be needed:

In this post, I consider the merits of helicopter money as a (presumably last-resort) strategy for policymakers. I make two points. First, in theory at least, helicopter money could prove a valuable tool. In particular, it has the attractive feature that it should work even when more conventional monetary policies are ineffective and the initial level of government debt is high. However, second, as a practical matter, the use of helicopter money would involve some difficult issues of implementation. These include (1) the need to integrate the approach with standard monetary policy frameworks and (2) the challenge of achieving the necessary coordination between fiscal and monetary policymakers, without compromising central bank independence or long-run fiscal discipline. I propose some tentative solutions for these problems.

To be clear, the probability of so-called helicopter money being used in the United States in the foreseeable future seems extremely low.  (emphasis added)

Almost everyone agrees that the US is likely to hit the zero bound in the next recession.  So obviously Bernanke doesn’t think being at the zero bound, in and of itself, calls for helicopter drops.  But then what would be the trigger?  On that issue he’s a bit vague.

Suppose we assume that “extremely low” means “1% chance”.  My response would be that we could lower than number to something closer to one in a million by adopting a different policy target, and giving the Fed the responsibility to “buy whatever it takes” to keep as close to the target as possible.  Then helicopter drops could be used as a fallback to make the “Chuck Norris effect” credible, without ever actually having to use the policy.

To his credit, Bernanke sees the problems with the helicopter drop theory—it doesn’t really solve any fundamental problem.  Even a helicopter drop may not be effective if the money supply increase is not viewed as permanent.  This causes Bernanke to suggest exotic extensions of the traditional helicopter drop:

As I’ve stressed, MFFPs [money financed fiscal policies] differ from ordinary fiscal programs by being money-financed rather than debt-financed. In my illustrative fiscal program, government spending and tax cuts are paid for by the creation of $100 billion in new money. To have its full effect, the increase in the money supply must be perceived as permanent by the public.

In practice, however, most central banks do not make monetary policy by choosing a fixed amount of money in circulation. Instead they set a target for a short-term interest rate (in the U.S., the federal funds rate) and allow the money supply to adjust as necessary to be consistent with that target. The rationale for this approach is that the relationship between interest rates and the economy appears more stable and predictable than that between the money supply and the economy. If central banks target interest rates rather than the money supply itself, than it’s not immediately obvious how the idea of a “permanent increase in the money supply” can be made operational.

One possible solution for that problem is that the central bank, rather than making an explicit promise about the money supply, could temporarily raise its target for inflation—equivalently, it could increase its target for the price level at each future date. Since the price level and the money supply tend to be proportional in the longer run, aiming for a higher price level could approximate the effects of committing to a higher money supply. A shortcoming of this approach is that it obscures the fact that the fiscal package is being financed by money creation rather than by new debt—a distinction that, again, the public must appreciate if the MFFP is to be fully effective.

This is not just a theoretical possibility.  We know that helicopter drops failed in Japan, because the monetary injections were viewed as temporary.  Bernanke correctly notes that shifting to a level target for prices can overcome this problem. But then level targeting also overcomes the main weakness of QE.  Thus if we do what Bernanke suggests, we don’t even need the fiscal component.

The methods that central banks use to meet their interest-rate targets pose further complications. Before the financial crisis, the Fed continuously varied the amount of money in the system (more precisely, the quantity of bank reserves) to keep the funds rate near the desired level. In the years since the crisis, however, several rounds of quantitative easing have resulted in very high levels of bank reserves, to the extent that the traditional method of making marginal changes to the supply of reserves is no longer effective in controlling the federal funds rate. Instead, following practices similar to those of other major central banks, the Fed currently influences the funds rate by varying the interest rate it pays on bank reserves and on other short-term investments at the Fed. [8]

As my former Fed colleague Narayana Kocherlakota has pointed out, the fact that the Fed (and other central banks) routinely pay interest on reserves has implications for the implementation and potential effectiveness of helicopter money. A key presumption of MFFPs is that the financing of fiscal programs through money creation implies lower future tax burdens than financing through debt issuance. In the longer run and in more-normal circumstances, this is certainly true: The cost to the Treasury of spending increases or tax cuts – and thus the future tax burden – will be lower if the Fed provides the financing. In particular, when the Fed’s balance sheet has shrunk and reserves are scarce again, the Fed will be able to manage short-term rates without paying interest on reserves (as it did traditionally), or in any event by paying a lower rate on reserves than the Treasury must pay on government debt. In the near term, however, money creation would not reduce the government’s financing costs appreciably, since the interest rate the Fed pays on bank reserves is close to the rate on Treasury bills.

Both interest-rate targeting and the payment of interest on reserves make it more difficult to achieve and communicate the cost savings associated with money financing. Here is a possible solution. Suppose, continuing our example, that the Fed creates $100 billion in new money to finance the Congress’s fiscal programs. As the Treasury spends the money, it flows into the banking system, resulting in $100 billion in new bank reserves. On current arrangements, the Fed would have to pay interest on those new reserves; the increase in the Fed’s payments would be $100 billion times the interest rate on bank reserves paid by the Fed (IOR). As Kocherlakota pointed out, if IOR is close to the rate on Treasury bills, there would be little or no immediate cost saving associated with money creation, relative to debt issuance.

However, let’s imagine that, when the MFFP is announced, the Fed also levies a new, permanent charge on banks—not based on reserves held, but on something else, like total liabilities—sufficient to reclaim the extra interest payments associated with the extra $100 billion in reserves. In other words, the increase in interest paid by the Fed, $100 billion * IOR, is just offset by the new levy, leaving net payments to banks unchanged. (The aggregate levy would remain at $100 billion * IOR in subsequent periods, adjusting with changes in IOR.) Although the net income of banks would be unchanged, this device would make explicit and immediate the cheaper financing of the fiscal program associated with money creation.

Or we could just raise the inflation target to 3%.

Or even keep it at 2% and do level targeting.  Or NGDPLT.  All these epicycles to make helicopter drops work make me dizzy.  The simple truth is that monetary policy is all we need if used intelligently, and if not used intelligently (as in Japan pre-2013), even helicopter drops won’t get the job done.  So let’s K.I.S.S., and work out fallbacks that don’t require wildly unrealistic assumptions about cooperation between the Fed and a GOP-controlled Congress.  Instead let’s simply shift the target slightly (4% NGDPLT anyone?), and perhaps add to the securities that the Fed is eligible to buy.

I am more sympathetic to Olivier Blanchard’s view of helicopter drops:

One thing he is not worried about is running out of monetary ammunition. “There is an argument that QE actually becomes more effective, the more you use it,” he said.

As a central bank buys more bonds, the more it has to pay to convince the last hold-outs to sell their holdings. “The effect on the price plausibly becomes stronger and stronger,” he said.

Prof Blanchard said the authorities should stick to plain vanilla QE rather than experimenting with “exotic stuff”.

He waved aside talk of ‘helicopter money’ with contempt, calling it nothing more than a fiscal expansion by other means. It makes little difference whether spending is paid for with money or bonds when interest rates are zero.

Blanchard favors a 4% inflation target, so that central banks would not hit the zero bound.  Again, K.I.S.S.

HT:  Benn Steil, James Alexander, et al.

Joshua Hausman has a very good paper on the 1936 veterans’ bonus, in the AER.  This was a payment of $1.8 billion to about 3.2 million WWI veterans.  The benefits were originally supposed to be in the form of a pension, but Congress bowed to public pressure and voted to make immediate lump sum payments to the veterans (in the form of bonds).  FDR viewed the bill as irresponsible, but it passed over his veto.

Hausman provides lots of pretty convincing evidence that the bonus payments boosted consumption, at least by those who received the payments.  He estimates a MPC in the .60 to .75 range.  One telling fact is that 80% of the bonds were cashed in during 1936, despite the fact that they paid an above market rate of interest (3%).  I imagine that after 7 years of depression, lots of the veterans were liquidity constrained and we were not in a Ricardian equivalence world.  Nonetheless, I’m skeptical that the bonus program was effective, not surprisingly for monetary offset reasons.

The bonus bill was passed in January 1936, and the payments went out to veterans in late June 1936.  As a bit of background, the recovery had temporarily stalled under the NIRA, indeed from July 1933 to May 1935 there was no increase in industrial production.  After the NIRA was ruled unconstitutional in May 1935, industrial production started rising rapidly, and continued doing so into early 1937, when growth slowed.  In late 1937 and early 1938, industrial production plummeted, as the US entered one of the most severe depressions of the 20th century.  So the bonus payments came part way through a period of strong growth.

I do believe that 1936 GDP was higher than it would have been without the bonus program, but I don’t think the overall effect was expansionary.  Less than a month after the payments were distributed, the Fed embarked on a major anti-inflation program.  On July 15, the Fed voted to raise reserve requirements sharply.  By itself, this action probably had only modest effects, not enough to offset the fiscal stimulus.  But then in December 1936, the Treasury began sterilizing gold inflows.  In March 1937, reserve requirements were raised again, and then again in May 1937.  By this time the reserve requirements had doubled from the levels of early 1936.

In my book on the Great Depression, I suggested that other factors largely explain the severity of the 1937-38 slump, especially higher labor costs in early 1937 combined with a bout of gold hoarding in late 1937.  This gold hoarding was a dramatic turnaround from the gold panic of late 1936 and early 1937 (which involved gold dishoarding).  Nonetheless, the various contractionary monetary policies were probably enough to offset the bonus payments, which were about 2.1% of 1936 GDP.

There are three possibilities, anyone of which seems plausible.  The monetary policy tightening might have offset, more than offset, or less than offset the bonus payments.  We don’t know, and given the “game theory” aspects of this problem, we will never be able to know for sure.  But it’s worth noting that 1936 was an almost ideal time to do fiscal stimulus:

1.  Generally speaking, monetary policy was quite passive during the period after 1934, with the monetary base tending to rise along with the monetary gold stock.  Under a gold standard regime, monetary offset is less likely to occur, and indeed the constraint of the gold standard is one reason why I was skeptical of the effectiveness of the various contractionary monetary policies in 1936-37.

2.  FDR had allies at the Board of Governors.

3.  Interest rates were near zero, which is generally assumed to be a condition where monetary offset is less likely to occur.

Thus I find it interesting that 1936-37 looks like an almost classic example of monetary offset.  There was a very specific, highly visible fiscal stimulus.  It occurred during a period of rising (wholesale) prices, and inflation seemed to accelerate further after the bonus was paid.  Then policymakers take not one but four steps to restrain AD and slow inflation, with the first occurring less than a month after the bonus payment.  And then the economy slows and eventually falls sharply. Admittedly that’s all circumstantial, but it’s almost a textbook example of what you’d expect monetary offset to look like.

Interestingly, Hausman doesn’t focus on monetary offset (his study uses cross sectional data, as well as survey data).  But he does allude to possible fiscal offset:

As is customary, I consider the multiplier for the bonus ignoring any political economy affects of the bonus’ passage on other spending and taxing decisions. The veterans’ bonus was itself deficit financed, but its passage led to political pressure for higher taxes. Thus the veterans bonus contributed to the enactment of the undistributed profits tax (the Revenue Act of 1936) in June 1936 (Romer and Romer 2012). This bill imposed taxes on undistributed corporate profits and also raised taxes on dividends. It did not affect other personal taxes. The political dynamic through which the veterans bonus contributed to the passage of this tax increase can be compared to the way the American Recovery and Reinvestment Act (the Obama stimulus) may have contributed to later spending cuts.

PS.  Just to be clear, I think the bonus program was expansionary during 1936.  My skepticism has to do with the effects of monetary offset on the economy in 1937 and 1938.  And my skepticism about the bonus as fiscal stimulus doesn’t mean I necessarily think it was a bad program.  I certainly have sympathy for WWI veterans, although I also understand the reasons why FDR vetoed the program.

I was stunned to see a recent paper on fiscal multipliers use a 90% confidence interval, which seemed far too lenient.  After all, economics and many other sciences suffer from problems such as data mining, publication bias, and inability to replicate findings.  I’d like to see the standard statistical significance cut-off point raised from 95% to something stronger, maybe 98%.  When I did this recent post I wondered if I was making some elementary error, as econometrics is not my strong suit.

It turns out the problem is even worse than I assumed.  Indeed Ryan Murphy recently published a study of fiscal multiplier research (in Econ Journal Watch), and found that many studies use 68%!!

In recent decades, vector autoregression, especially structural vector autoregression, has been used to study the size of the government spending multiplier (Blanchard and Perotti 2002; Fatás and Mihov 2001; Mountford and Uhlig 2009). Such methods are used in a significant proportion of empirical research designed to estimate the multiplier (see Ramey 2011a). Despite being published in respected journals and cited by prominent members of the profession, much of this literature does not use the conventional standard of statistical significance that economists are accustomed to in empirical research.

Results in the literature on the fiscal multiplier are typically communicated using a graph of the estimated impulse-response functions. For instance, the effect of government spending on output may be reported by reproducing a graph of an impulse-response function of a one-unit (generally, one percentage point or one standard error) change in government spending. The graph would show the percent change in output over time following the change in government spending. To report statistical significance, authors of these studies may then draw confidence bands around the impulse response function. Ostensibly, if zero lies outside the confidence band, it is statistically distinguishable from zero. But very frequently in this literature the confidence bands correspond to only one standard error. In other words, instead of representing what corresponds to rejecting the null hypothesis at a 90% level or 95% level, the confidence bands correspond to rejecting the null hypothesis at a 68% level. By conventional standards, this confidence band is insufficient for hypothesis testing. Not every useful empirical study must achieve significance at the 95% level to be considered meaningful, of course, but a pattern of studies which do not use and reach the conventional benchmark is a cause for attention and perhaps concern. Statistical significance is not the only standard by which we should judge empirical research (Ziliak and McCloskey 2008). It is, however, a useful standard, and still an important one. Here I examine papers in the fiscal multiplier literature which apply vector autoregression methods. Sixteen of the thirty-one papers identified use narrow, one-standard-error confidence bands to the exclusion of confidence bands corresponding to the conventional standard of 90% or 95% confidence. This practice will often not be clear to the reader of a paper unless its text is read rather carefully.

I can’t even fathom what people are thinking when they use 68%.  It seems like something you’d see in The Onion, and yet apparently this stuff gets published.  Can someone help me here, what am I missing?