Why Counterfactual Thinking Is Important

I recently made the case that many observers are not thinking properly about the Fed's Quantitative Easing (QE) programs. Using the analogy of George Bailey's life in the film It's a Wonderful Life, I argued that the critics who question the efficacy of the QE programs are doing the wrong counterfactual. Today, Barry Ritholtz makes the same point:

One of the analytical errors I seem to constantly come across is what I call the non-result result. It goes something like this: If you do X, and there is no measurable change, X is therefore ineffective.

The problem with this analysis is the lack of a control group, If you are testing a new medication to reduce tumors, you want to see what happened to the group that did not get the tested therapy. Perhaps their tumors grew and metastasized. Hence, no increase in tumor mass or spreading is considered a very positive outcome.

This seems to get loss in the debate over QE. The debate — either ignorantly or disingenuously — makes claims such as “Look how few jobs have been created, and look how high unemployment is.”

Understanding this logic, and lacking a control group, we must employ a counter-factual. The question one should be asking is “How many less jobs would have been created?;  How much higher would unemployment be?”

This idea is nicely summarized by the QE counterfactual produced by Poltiical Calculations. It shows what would have happened to nominal GDP had there been no QE3. It is not a pretty sight:

I too ran a QE counterfactual in my George Bailey post. There I considered what would have happened to employment, the stock market, PCE core inflation, and the repo interest rate conditional on (1) the Fed not increasing its share of marketable treasuries starting in late 2010 and (2) the Eurozone crisis, China slowdown, and fiscal policy shocks still occurring as they did. The implications for the economy were the same as in the figure above. 

These counterfactual exercises serve as a nice complement to the on-going quasi-natural experiments that indicate monetary policy at the ZLB can still pack a punch. Yes, monetary policy is far from perfect. But it is also far from impotent as claimed by some QE skeptics.

Thoughts for the Weekend

First, don't be derpy. In my previous post I presented evidence that small business are very concerned about Obamacare and, as a result, are cutting back on employee hours. The evidence came from three surveys--NFIB, Gallup, and Foundation of Employee Benefit Plans--and a list of firms who have cut hours because of the ACA. Despite this evidence, some observers were feeling derpy and rationalized it away. As a follow up, Jed Graham sent me an article that further supports the claim that Obamacare is reducing hours worked. Specifically, it shows hours worked for low income workers starting to fall. Casey Mulligan makes a similar point here. The real question, in my view, is how big of an effect this will have on the labor market. Here I think reasonable people can disagree. Meanwhile James Pethokoukis summarizes we know so far about the implementation of the ACA.

Second, can you see what I see? Paul Krugman reminds us that we are coming up on the five year year anniversary of being at the ZLB. He tells us that in this strange world everything changes:

[Hitting] the zero lower bound changes everything. It’s not just that the rules change for monetary policy, although they do: some people have been warning for the whole five-year period that the surge in the monetary base will cause runaway inflation, and it keeps not happening. It’s also true that we enter the territory of paradoxes; the paradox of flexibility, but also, and more crucially, the paradox of thrift, in which attempts by some players in the economy to save more end up leading to less, not more, investment.

Okay I see a strange world too, but instead of seeing a 'paradox of flexibility' I see central bank incompetence.  And instead of seeing a 'paradox of thrift', I see an ongoing excess money demand problem that is summed up nicely in the following figure:

This figure shows that the BIS finance neutral output gap is negatively related to the share of liquid assets being held by households. This liquidity demand measure also tracks the inverted real SP500 closely as seen below. This figure illustrates the portfolio rebalancing channel: as households decrease their liquid assets holdings and shift to riskier assets, those riskier asset prices soar. The Fed job, then, is to conduct monetary policy so that the right amount of rebalancing (i.e. the amount the closes the output gap) take place.

One area where I concede that ZLB in and of itself matters is that it leads to market segmentation between transaction assets and other assets.  This is a big deal and explains why we still have a safe asset shortage problem.

Third, model this wise guy. Cardiff Garcia points us to a new paper by Ricardo Caballero and Emmanuel Farhi. They develop a model that, among other things, implies the Fed's treasury purchases are harming the economy because they are removing safe assets from the financial system. This is an increasingly popular critique that I believe is wrong and said so here and here. I now have a model that bears this out. Josh Hendrickson and I coauthored a paper where a monetary search model is used to motivate transaction assets. With this model, we show that by properly managing expectations, the Fed can increase the supply of private label safe assets and at the same time reduce the demand for public safe assets. In other words, the Fed can solve the safe asset problem by improving the economic outlook. This really should not be shocking, but it is implicit in many of the arguments against QE. We also do some cool counterfactual forecasts that show what would happen to the stock of safe assets if the Fed did a better job managing NGDP expectations. Check it out.

My Supply-Side Senses Are Starting to Tingle

One of the most contentious debates over the past few years has been whether the U.S. slump is the result of  an aggregate demand shortfall or a spate of negative supply shocks. Depending on which view you take, the policy implications are very different. In the former case, monetary policy could help by closing the output gap. In the latter case, it would only be 'pushing on string' and might even create more problems. 

I have argued that the slump has largely been the result of an aggregate demand shortfall. One piece of evidence for this view I have repeatedly pointed to is a question on the NFIB Small Business Economic Trends survey. This survey question asks firms what specific developments they see as the "single most important problem" they face. Answers to question include lack of sales, regulation, taxes, inflation, financing costs, quality of labor, insurance availability, competition from large business, and other. This question, therefore, allows us to see from a small firm's perspective how important supply shocks are--as measured by regulation, taxes, financing, and quality of labor--compared to demand shocks--as measured by a lack of sales. Since the crisis the started, a lack of sales has been the number one problem. This series has led changes in the unemployment rate in a remarkably consistent manner since the data starts, as seen below. The easiest and most straightforward interpretation of this relationship is that firms cut back on production and employment as a result of the expected weak sales. From the firm's perspective, this suggests an aggregate demand shortfall is the key driver behind the slump.

Now I still buy this story, but recent developments in the survey question suggest supply-side concerns are becoming a bigger deal. In fact, a lack of sales is no longer the number one problem. Rather, it is concerns about regulation, as seen below.

What is even more alarming about these regulation concerns is that they have consistently risen since 2009 and are now at their highest level since the previous peak in 1994. See the figure below:

 

So what possibly could be driving the rise in regulation concerns? I think the answer is obvious: Obamacare. And I think observers like Brad DeLong, Paul Krugman, and myself who have been so quick to use this NFIB data when it fits our views need to be honest and acknowledge what this data is saying now.

So how do we interpret the rising small business concerns about government regulation? One manifestation of these concerns might be the claim that some firms are cutting back employee hours to under 30 so that they do not have to offer them health insurance. This claims resonates with me since I know people in my community who have had their hours cut back for this reason. Brad DeLong, Jared Bernstein, Max Sawicky, and others say no way, there is no evidence of this in the employment data. They also ask why would firms start doing this now if this requirement does not become law for another year.

I acknowledge their point on the employment data, but would direct them to two polls that suggest firms are cutting back on hours in one form or another. The first one was a Gallup poll and reported on CNBC:

Small business owners' fear of the effect of the new health-care reform law on their bottom line is prompting many to hold off on hiring and even to shed jobs in some cases, a recent poll found.

"We were startled because we know that employers were concerned about the Affordable Care Act and the effects it would have on their business, but we didn't realize the extent they were concerned, or that the businesses were being proactive to make sure the effects of the ACA actually were minimized," said attorney Steven Friedman of Littler Mendelson. His firm, which specializes in employment law, commissioned the Gallup poll...

Forty-one percent of the businesses surveyed have frozen hiring because of the health-care law known as Obamacare. And almost one-fifth—19 percent— answered "yes" when asked if they had "reduced the number of employees you have in your business as a specific result of the Affordable Care Act." 

A more telling survey was done by the Foundation of Employee Benefit Plans. They found that 15% of firm with 50 or more employees and 20% of firms with less than 50 employees had plans to adjust hours so that fewer employees qualify for full-time medical insurance under the ACA. The smaller firms also planned to make other changes in seen the figure below from their survey:

So according to these surveys and the data from the NFIB, Obamacare is now having an effect on labor markets. Now this evidence does not speak to the size or magnitude of this effect. The figure above suggests it is limited to smaller firms. Maybe that is why folks like Brad DeLong, Jared Bernstein, Max Sawicky, and others who are associated with big institutions have not met anyone adversely affected by the ACA. I too work at a large institution, but live in rural Tennessee where I have met people whose hours have been cut because of Obamacare. Further examples of people affected by the ACA can be found here in this Guardian article.

So yes, firms appear to be cutting back on employees and hours. And they are at a high point for concerns over regulations. This has my supply-side senses tingling. Your supply side senses should be tingling too. The only question is how big is this effect.

P.S. To be clear, I still view most of the lingering labor market weakness as a result of the ongoing shortfall in aggregate demand. However, the evidence above suggests supply-side concerns are increasingly becoming important.

Update: Jed Graham has compiled a list of firms who have cut hours or jobs because of the ACA.

Note to Fed: Get Out in Front of the Potential Debt Crisis.

The consequences of the U.S. government defaulting, even if for a few days, could be significant. The long-term effect is that the U.S. economy could lose some or all of its 'exorbitant privilege' which has allowed Americans for many years to live beyond their means. The short-term effect could be a major financial crisis as noted by Dick Bove over at FT Alphaville. Here is how he believe it could play out:

Money Market Mutual Funds
Should the United States government default virtually every money market mutual fund (MMMF) in the country would “break-the-buck” – i.e., be unable to pay investors 100 cents on every dollar invested. At present, MMMFs that do not actually earn enough money to pay back 100 cents on the dollar are subsidized by the fund management company. A Treasury default would make this virtually impossible and millions of Americans would lose billions of dollars.

Mutual Funds

The indentures of most bond and balanced mutual funds require that they immediately divest their holdings of defaulted securities. This could cause hundreds of billions in Treasuries to be sold immediately when the default was announced.

Depository Institutions
Let’s shift to another government document. This would be the FDIC’s aggregate balance sheet of the American banking industry... A reasonable estimate would be that the U.S. banking industry owns $1.85 trillion in government backed securities. It has $1.63 trillion in equity. If the Treasury and related securities were in default, one does not know what they would be worth. Assume a Latin American valuation of 10 to 20 cents on the dollar and an estimated $1.28 trillion in U.S. banking equity would be wiped out...In addition to the U.S. backed securities the banks own, they own an additional $1.27 trillion in other securities that would plunge in value... They have $7.73 trillion in loans which would also fall in value.

These developments would probably spill over into other markets. Such dire outcomes sound a lot like the concerns Fed Chairman Ben Bernanke had about the Eurozone Crisis back in 2011. Like now, I argued back then that Bernanke should get out in front of the Eurozone Crisis. Here is what I said:

So what can the Fed do? Here is a suggestion: the Fed could say if total current dollar spending begins to plummet because concerns about the financial system are causing investors to rapidly buy up safe money-like assets (time and saving accounts, money market accounts, treasuries, etc.) then the Fed would begin buying up less-safe and less-liquid assets until the investors' demand for money-like assets is satiated such that they return total current dollar spending to its previous level. The Fed would need to stress the "until" part means it would purchase as many trillions of dollars of assets as necessary to restore total current dollar spending. Since this  process would take place over time, the Fed would also want to set a target growth rate for where it wanted the level of total current dollar spending to go.    

If the above sounds reasonable to you, then you should be a fan of nominal GDP level targeting.  It is exactly what the U.S. economy needed in early 2008 when inflation expectations and velocity started falling.  And it is exactly what the  U.S. economy needs now.   

So if the Fed wants to get out in front of the potential debt crisis, one important step it could take is to announce a NGDP level target. The Fed's QE programs have been able to able to offset much of the fiscal drag over the past three years. A NGDP level target would step up the Fed's game and make it better able to deal with the fallout of debt default. It would not solve all problems--for example, the U.S. might face permanently higher risk premiums--but it at least could offset any drag it might create on aggregate nominal spending. Now would be a great time to act.

P.S. An interesting question raised by Jeffrey Frankel is if the default happens, will it create a financial crisis before or after October 17. 

More on George Bailey, QE, and the Shortage of Safe Assets

In my previous post I made the case that the Fed is not exacerbating the safe asset shortage problem. Rather, the Fed has actually prevented it from becoming far worse. Michael Darda, chief economist of MKM Partners, makes a great follow-up point in an email:

One thing that jumped into my mind when reading [your post] was the consistent tendency for the yields on safe assets to rise during “QE on” periods. We’ve now had three episodes of QE since 2009, and in each and every case, the 10-year yield has risen on net, only to fall when the Fed stopped too soon (after QE1 and QE2 came to a premature end, that is). This, I think, is a pretty airtight refutation that the Fed is somehow exacerbating the safe asset shortage; indeed, this alone suggests that QE is actually relieving it. And even with the recent pullback in rates, the 10-year yield is more than 100 bps above where it was before QE3 began (with the S&P 500 up 25% or so over the same timeframe). This, in my view, means it is working.

Ironically, one implication of this analysis is that the safe asset shortage problem exists, in part, because the Fed was not doing enough. Or at least, not doing it right. This a point I made in my last National Review article.

What George Bailey Can Teach Us About QE

Quantitative easing (QE) is a lot like George Bailey in the classic film It's a Wonderful Life. George was a man who began questioning his value to society. A number of cascading events--unfulfilled life dreams, a run on his bank, lost bank deposits, bank fraud charges--made it appear to hm that he was on balance a drag on his community. George decided it would be better for all if he 'tapered' or ended his life. Fortunately, an angel appeared at the last minute and revealed that despite his immediate problems, George's overall contributions to society were immense. Many lives were saved and changed because of his efforts. All that was needed to see this fact was a broader, longer perspective. George Bailey, in other words, was not doing the right counterfactual.

The same is true for skeptics of QE. Many point to the apparent flaws of the Fed's large scale asset purchases (LSAPs), but fail to step back and consider the counterfactual of what would have happened in their absence. This point is particularly poignant to those observers who claim the Fed's LSAPs of treasuries are particularly bad because they drain the financial system of safe assets. These critics note that these safe assets serve as collateral in the shadow banking system and thus facilitate transactions. Therefore, when the Fed increases its balance sheet it is actually restricting the funding capabilities of the shadow banking system and creating a drag on the economy. A growing number of smart people are making this point, including Izabella Kaminski, Tyler Cowen, Peter Stella, Arvind Krishnamurthy and Annette Vissing-Jorgensen, and Michael Woodford. I contend that while correct on the immediate effect of LSAPs, these critics like George Bailey are doing the wrong counterfactual. The right counterfactual is what would have happened had there been no QE2 and QE3 at all.

Before delving deeper into this counterfactual, I want to mention two other points about this critique. First, most of the safe asset shortage was created by factors other than the Fed's LSAPs. On the supply side, there was the destruction and subsequent anemic recovery of private-label safe assets as seen in this figure. On the demand side, the financial crisis and then a spate of subsequent bad economic news--Eurozone crisis, China slowdown concerns, debt limit talks, fiscal cliff talks--has kept the demand for safe asset elevated. Also, new regulatory requirements requiring banks to hold more safe assets is keeping safe asset demand elevated. 

Consequently, forces other than QE2 and QE3 have been driving much of the safe asset shortage. This can be seen in the figure below which shows the treasury general collateral repo rate and the Fed's share of marketable treasury securities. Note that the largest drop in the repo rate (reflecting the reduced supply and increased demand for treasuries collateral) occurred between late 2007 and early 2009, the very time the Fed was releasing treasury securities back into the market.

Second, even though the LSAPs do cause an immediate drain of safe assets, they potentially lead to more private safe asset creation. As I noted before:

The LSAPs, if done right, should raise expected economic growth going forward and cause asset prices to soar. This, in turn, would increase the current demand for and supply of financial intermediation. For example, AAA-rated corporations may issue more bonds to build up productive capacity in expectation of higher future sales growth. Financial firms, likewise, may start providing more loans as the improved economic outlook makes households and firms appear as better credit risks.

There is some evidence this is happening with QE2 and QE3. Even if these private safe assets are not used as collateral in the repo markets they will be used elsewhere to satiate liquidity demand. That, in turn, should free up more treasuries for use in the repo market. Both of these points are often overlooked by QE critics.

Now back to the counterfactual point. Are QE critics really making the same mistake as George Bailey? To answer that question, let's think through the counterfactual of no QE2 and QE3. First, assume the Fed's share of marketable treasuries was constant over this period. Also, assume that the shocks from the Eurozone crisis, China slowdown, debt limit and fiscal cliff talks still buffeted the U.S. economy during this time. What would have happened to the U.S. economy? Could the U.S. economy have been as resilient to these shocks had the Fed not been supporting it? If not, then imagine the mess in the financial system and what that would have done to the demand for safe assets. Repo markets, for example, would probably be facing an even larger collateral shortage. A reasonable counterfactual, then, is that the safe asset problem would be greater were it not for the Fed's QE programs. 

I took this idea to the data. I estimated a vector autoregression (VAR) over the 2003:1 to 2013:8 period and used it to create a counterfactual path for the 2010:10 to 2013:8 period.¹ This corresponds to the QE2 and Q3 periods. VARs are great for this type of exercise because one, they allow the variables in the model to interact and two, you can do dynamic forecast with them. The only twist here is that I did a conditional dynamic forecast. Specifically, I dynamically forecasted what would happen to employment, the stock market, PCE core inflation, and the repo rate conditional on (1) the Fed not increasing its share of marketable treasuries and (2) allowing the Eurozone crisis, China slowdown, and fiscal policy shocks to still affect the economy. For the latter, I used the actual, realized values of the economic policy uncertainty index over the forecasted period as a way to summarize and include these shocks in the VAR. The results can be seen in the figures below.

The first figure shows counterfactual path for the Fed's share of marketable treasuries used in the forecast.

The next figure shows what happens to the stock market. It declines over much of the period.

The following figure reports the counterfactual path for employment. Again, not a pretty picture.

This next figure shows that core PCE inflation stays around 1% and never recovers. So QE is inflationary, at least relative to where inflation would be in its absence.

Finally, this figure shows that the repo rates would have gone negative. Now this would not happen in practice because of the ZLB. But the VAR does not know the ZLB and simply forecasts based on estimated relationships. The fact, though, that it goes significantly negative is instructive. It indicates the repo markets would have faced even greater collateral shortages had the Fed not done QE2 and QE3.

Now we do not want to read too much into these results. They come from a forecasting model that is far from perfect. Still, they indicate that a proper counterfactual of the QE programs requires more than just narrowly looking at the immediate impact of LSAPs on the collateral asset supply. We do not want to be like George Bailey and do the wrong counterfactual. Neither should the Fed. Otherwise, it too may be tempted to pull the trigger and taper too soon.

My assessment is that QE2 and QE3 has done more to shore up the U.S. economy than many observers realize. We do not know how much worse the economy would be in their absence, but the analysis above suggests it could have been ugly. With that said, the QE programs have been flawed because of their ad-hoc, make-it-up-as-we-go-along approach that until recently was not tied to any explicit target. Tying the LSAP more firmly to conditional outcomes would do much to improve them. The more rule-like and predictable the better.  I think George Bailey would agree.

Update: Michael Darda provides some follow-up comments here.

¹The VAR was estimated with the Fed share of market treasuries, the log of SP500, the log of employment, the core PCE inflation rate, the repo rate, and the economic policy uncertainty index. Six lags were used on the monthly data. The results were generally robust to longer lag lengths, but due to limited data six lags were chosen. The conditional forecast was created by imposing fixed values for the Fed share and the uncertainty index as noted above.