by John H. Cochrane¹
For nearly a decade in the US, UK, and Europe, and three decades in Japan, short-term interest rates have been stuck at zero, known as the “zero bound,” because central banks can’t lower interest rates substantially below zero. Central banks also embarked on immense open market operations. The US Federal Reserve bought nearly $3 trillion of bonds and mortgage-backed securities, in return for newly created money. Bank reserves, essentially checking accounts that banks hold at the Fed, rose from $10 billion on the eve of the crisis in Aug 2008 to $2,759 billion in Aug 2014. Figure 1 summarizes the US experience.
The response to this important experiment in monetary policy has been surprising silence. Inflation is stable, and if anything less volatile than before. Similar plots of GDP growth and unemployment show that no large difference in the behavior of the economy during the time that interest rates are stuck near zero and not responding to economic conditions. The zero bound is not an obvious “state variable” for economic dymanics. Unemployment came down if anything a little more quickly than in previous recessions. GDP growth, while too low in many opinions, has been if anything a little less volatile than before.
Existing theories of inflation make sharp predictions about the zero bound. Old-Keynesian models, in use throughout the policy world, predict that inflation is unstable at the zero bound, and more generally when interest rates cannot or do not move in response to inflation. These models predict a deflation spiral: Inflation goes down, so the real cost of borrowing money rises. That depresses the economy, inflation goes down more, the real cost of borrowing money rises more, and so on ad infinitum. Thing of the Fed as a seal, balancing a ball (inflation) on its nose. If the seal does not or cannot quickly move its nose (interest rates), then the ball will fall over. It did not happen.
Monetarist thought, that inflation fundamentally comes from increases in the quantity of money, predicts that a massive increase in reserves must lead to galloping inflation. It did not happen.
Since the late 1980s, most academic work on monetary policy has been conducted in the framework of “new-Keynesian” models. These models recognize that people make their decisions about what to do today thinking about the future, not the past. In economicspeak, they are “intertemporal” and feature “rational expectations.” They are also fully-specified economic models, obeying all the rules that well-posed models should obey. For example, they impose that people’s plans to work, save, and spend are consistent. They impose budget constraints and market-clearning conditions.
Simple new-Keynesian models predict that inflation is stable when interest rates do not move, and they predict that quantitative easing operations are irrelevant. The intuition is fairly straightforward. If a driver looks only in the rearview mirror, forming his expectations of the road ahead by what lies behind, he will soon veer off the road. He needs a prescient Federal Reserve in the back seat to induce stability. If the driver looks forward, the car will be seen returning to the center of the road on its own, even if the Federal Reserve does no back-seat driving. Likewise, from the perspective of mdoern finance, reserves at the Fed are indistinguishable from government debt. An exchange of short-term debt for reserves is like an exchange of $20 bills for two $10 bills. Without lots of extra “frictions” such an operation does not change overall spending.
Thus, the observed stabiliy of inflation and apparent ineffectiveness of quantitative easing are big feathers in the new-Keynesian cap. But they fail on quiet. Standard new-Keynsian models predict that at the zero bound, or when interest rates do not move, inflation jumps around randomly. The models have “multiple self-confirmng equilibria” or “sunspots” when interest rates do not move. If people expect inflation, inflation happens. These models tie down expected inflation, but not actual inflation. Think of a coin being flipped. Interest rate targets tie down the fact that on average half the flips will be heads. But the actual flips are a volatile mixture of heads and tails. When interest rates can move, in these models, the Fed can guarantee all heads or all tails, and eliminate the random volatility.
The prediction that inflation is more volatile at the zero bound is a central component of the new-Keynesian paradigm. The central empirical success of these models was to explain the greater volatility of inflation in the 1970s relative to the 1980s by such “sunspots,” resulting from interest rates that did not move enough in response to inflation in the 1970s, but did so in the 1980s. Two decades of new-Keynesian research starting in the 190s was devoted to devising means to escape the “zero bound” or “liquidity trap,” of zero interest rates, precisely and only to avoid the reemergence of such “sunspots.” Well, here we are, and the long-feared volatility did not happen. As Figure 1 emphasizes, instead of extra “sunspot” volatility, inflation is if anything quieter than before.
New-Keynesian models also predict a menagerie of policy paradoxes when interest rates are stuck at zero: Productivity improvements are bad, promises further in the future have larger effects today, and reducing price stickiness makes matters worse, without limit.
One last theory remains. The fiscal theory of the price level states that inflation is fundamentally anchored by fiscal policy. In the end, the value of money comes from the government’s commitment to accept its money, and only its money, for tax payments. If there is more government debt overall outstanding than people expect to be soaked up by tax payments, the value of that debt falls, and inflation breaks out.
More deeply, the fiscal theory proceeds from the observation that the real value of government debt must equal the present value of the primary budget surpluses that will eventually pay down that debt. If people think that surpluses will not be sufficient to pay off the debt, they will try to get rid of that debt by buying goods and services. This will drive up the price level, until the now-lower real value of the debt is equal to that lower value of expected surpluses. Nominal debt is, formally, just like stock in the government, with the price level as the stock price and the discounted value of surpluses as the discounted value of dividends.
This theory can be merged easily with the new-Keynesian description of the rest of the economy, including its interest rate targets and sticky prices. The Fed, by setting interest rates, still determines expected inflation. But now fiscal policy determines the actual outcome-whether the coin comes up heads or tails. Each new-Keyenesian sunspot corresponds to a change in expectations about fiscal policy. With no big changes in fiscal policy (the present discounted value of future primary surpluses), there will be no sunspot volatility.
The resulting theory is consistent with stable and quiet inflation at the zero bound. It also resolves the policy paradoxes of the new-Keynesian model. This small change in ingredients has a large effect on the models’ prediction for what we see and for the effects of policy.
Telling these theories apart was difficult before interest rates hit zero. Each offered a plausible account of the data up to that point. If a seal does a good job of balancing the ball, it’s hard to tell if the ball is unstable and the seal is doing a great job, or if the ball is glued to the seal’s nose. If someone holds the seal still, it’s easier to tell. The zero bound period starting in 2008 offers a genuine and important experiment.
Theories fail sometimes in a dramatic manner. In the 1970s, prevailing Keynesian theory predicted little inflation, and the emergence of stagflation dramatically disproved that theory. In the 1980s, the same theory predicted inflation would remain intractably high. The sudden disappearance of inflation in 1982084 again proved it wrong. Theories fail no less when they unambiguously predict a large inflation, deflation, or volatile inflation, and nothing happens. It’s just a lot less public.
Do higher interest rates raise or lower inflation?
What does this experience, and theorietical interpretation, imply about monetary policy going forward?
First, if inflation is stable when interest rates are stuck at zero, then it follows that if the central bank were to raise interest rates permanently, then inflation must eventually rise to meet the higher interest rates. This reversal of the usual sign of monetary policy has become known as the “neo-Fisherian” hypothesis.
However, higher interest rates might still temporarily lower inflation before eventually raising it. The traditional belief that raising rates lowers inflation could still be right in the short run, and most evidence is about short run correlations anyway. Is that possible? What do the models say?
It turns out the standard simple new-Keynesian model, with or without fiscal theory, robustly predicts that a rise in interest rates produces a steady rise in inflation, with no temporary decline. It does produce an output decline – our central bankers are half right. Figure 2 illustrates.
This model produces a temporary inflation decline only if we pair the interest rate rise with a fiscal contraction – the fiscal contraction produces the temporary negative inflation, then the higher interest rates kick in to produce higher inflation.
That mixture may describe historical events – fiscal and monetary policy react to the same events – and therefore account for experience and econometric estimates. But if we define “monetary policy” as an increase in interest rates that does not come with a fiscal contraction, then our model still predicts that a future pure monetary policy interest-rate rise will lead only to inflation.
I investigate what minimal set of ingredients it takes to produce a negative short-run impact of interest rates on inflation. The obvious candidates do not work: pricing frictions, adding money and monetary frictions to the model, and even adapting classic backward-looking Phillips curves. With any forward-looking behavior, higher interest rates mean higher inflation. It is simply not true to say that “sure, in a frictionless model higher rates mean higher inflation, but since prices are sticky/ the real world has money in it/ price setting seems to look backward, higher rates temporarily lower inflation.” They don’t.
One ingredient can robustly and simply produce the desired temporary negative sign. If we add long-term debt, then a rise in interst rates can produce a temporary decline in inflation. In brief, when the fed raises interest rates, and communicates that interests rates will be higher for some time in the future, then long-term bond prices fall. Now, the total market value of the debt falls. But if the treasury does not make any change in fiscal policy, then the real value of debt has not changed. We have an imbalance. Treasuries are worth more than their market price. People try to buy more treasuries, and buy less goods and services to get them. But with the supply of Treauries fixed and their price (interest rate) fixed, the lower aggregate demand for goods and services pushes the price level down. Once the price level has fallen, the higher inflation corresponding to higher interest rates can take over.
Figure 3 illustrates this mechanism. Here I plot the response to a permanent one percentage point increase in interest rates, using the same economic moel as in Figure 2, but with long-term debt, calibrated to the US maturity structure. The main “inflation” line shoes the temporary decline, and later rise. The output gap line shows that this temporary tightening still produces a substantial recession.
This theory works even in a completey frictionless model – no price stickiness, no money, no frictions at all. It allows the analysis of monetary policy to (finally) start with simple supply and demand, like the rest of economics, and then add frictions to better match the economic dynamics in the data, rather than requiring monetary, financial, pricing, or other frictions just to get the basic determination of the price level and basic signs and stability properties of monetary policy right. And it describes interest rate policy, quantitative easing, and forward guidance in one breath. The interest rate rise involves bond sales that look just like quantitative easing. Forward guidance of future interest rate declines lowers bond prices – in fact the expectation of future high interest rates is the key mechanism.
However, this mechanism does not restore classic beliefs. First, it only works for unexpected interest rate increases. If people know the interest rate increase will happen, hty cannot be surprised by lower bond prices, and the inflation happens immediately without a temporary dip. Second, in part for this reason, it does not rescue policy advice that relies on expected interest rates lowering inflation. Central banks cannot plan to systematically raise and lower rates in response, say, to inflation, by this mechanism. Third, the mechanism works entirely via fiscal policy. If this is “monetary policy” it has nothing at all to do with money, credit, lending, price stickiness or anything else. In turn, wheter it works or not depends entirely on the Treasury. When the Fed raises rates, and thus future inflation, the Treausry could say “great, now we don’t have to raise as much taxes to pay off the debt. The Fed is inflating it away for us.” If it does so, the inflation dip disappears.
We are left with a logical conundrum: Either 1) The world really is Fisherian, higher interest rates raise inflation in both short and long run; 2) more complex ingredients, including frictions or irrationalities, are necessary as well as sufficient to deliver the negative sign, so this hallowed belief relies on those complex ingredients; 3) the negative sign ultimately relies on the fiscal theory story involving long-term debt – and has nothing to do with any of the mechanisms commonly alluded for it.
The first view is not as crazy as it seems. The empirical evidence for the traditional sign is weak. Estimates for years confronted the “price puzzle” that the data indicated higher interest rates led to higher inflation. This finding was only tempered with lots and lots of effort. Perhaps the price puzzle was trying to tell us something for all these decads.
Proof is rare in economics, and one can imagine many patches might rescue existing theories. Perhaps inflation really is unstable at the zero bound, but clever central bankers around the world just offset a pending deflationary sprial with just enough hyperinflationary quantitative easing, helped by fiscal stimulus, that all we see is quiet. Even in Japan. Perhaps. Or perhaps the stability we observe is just what it seems, stability. Occam’s razor – accept the simplest explanation – suggests the latter.
Similarly, one might rescue the long-standing prediction that interest rates at the zero bound would result in additional sunspot volatility by supposing that sunspots just didn’t happen. But taking that path, one would have to throw out the theory’s central empirical sucess, and ask why the literature made such clear and loud predictions. And we’re not here for cocktail party ex-post explanations, we’re here for theories, with predictive content. Just why were there no sunspots this time, and there were lots of them in the 1970s in this theory’s reading? Are our central banksers that much better at making speeches? In any case, choosing what’s on the menu, this possibility remains in the realm of future possibilities, as no new-Keynesian research has offered a serious explanation. And again, Occam’s razor speaks loudly. Yes, perhaps there are no sunspots now, or in the 1970s. Perhaps the whole sunspot theory is wrong. Perhaps the very simple fiscal theory, which has no sunspots at any time, describes now and the 1970s.
Perhaps the long zero bound represents the proverbial seven years of bad luck, and twenty-five years in Japan, not a true zero bound. Perhaps people, like many professional forecasters, expected a swift recovery and interest rates to rise above zero within a year the whold time, allowing conventional “active” (moving quickly with inflation) interest policy to emerge. Perhaps this wasn’t really a period of passive (interest rates do not move enough with inflation) like the 1970s. That would explain the absence of sunspots.
This story is also offered for the 1980s. The 1980s pose a similar challenge to new-Keynesian models, because they predict that persistent interest rate rises eventually raise inflation. They can only generate a decline in inflation from a quite temporary rise in interest rates. But the conventional view of the 1980s is that persistent, indeed dogged, high interest rates were required to squeeze out inflation. Wel, maybe the 1980-2000 experience was 20 yearsof good luck. Maybe people continually expected inflation to return, and were surprised that it did not. I call these the “springtime in Chicago” expectations, as it seems every week the weather forecast reads, “snow and ice this week, returning to the 70s next week.”
Well, perhaps. We should not be religious about rational expectations. Perhaps the 1980s and 2008 were unique events, that people had no way of preparing for or knowing what would happen. Perhaps the time-series we observe are a fundamentally misleading measure of the structural response functions, the former stable but the latter really unstable.
Or, perhaps not. At some point, after some amount of decades, perhaps we should take the very simple model sitting on the plate before us, that describes these episodes with simple supply and demand economics, not requiring people to be fundamentally wrong in how they perceive the world – and to ignore the ample historical precedents of financial crises, inflations, deflations, and near-zero interest rates. Perhaps every day is not a new stochastic process, but just a day like the last.
Furthermore, a stable quiet zero bound does not require extreme rational expectations. Small amounts of forward-looking behavior will do. The stable quiet zero bound still obtains if one of the consumption or pricing decisions is irrationally blackward looking. To rescue classic beliefs, one needs all expectations to be mechanically adaptive.
To generate the longstanding belief that higher interest rates generate at least temporarily lower inflation, one might naturally start adding complications to the very simple models I outline here, such as extensive borrowing or colalteral constraints, hand-to-mouth consumers, a lending chanel or other financial frictions, habits, durable goods, housing, mutliple goods and other nonseparabilities, novel preferences, labor/leisure choices, production, capital, variable capital utilization, adjustment costs, alternative models of price stickiness, informational frictions, market frictions, payments frictions, more complext monetary frictions, timing lags, individual or firm heterogeneity, and so forth. Going further, perhaps we can add fundamentally different views of expectations formation and equilibrium concepts.
Even this is not so easy. One must face the twin challenges of producing a negative temporary effect of interest rates on inflation, together with the observed long-run stability of inflation at the zero bound. “Let’s just go back to adaptive expectations” will not do. That course produces a negative sign, but it also produces instability, and the prediction of a deflation spiral, which we did not see.
One can, and many papers do, add complex ingredients to the new-Keynesian frame-work, which is consistent with stability. If we must go down this path, however, we then accept that there is no simple economic model that produces the hallowed belief that higher interest rates reduce inflation. The extra complexities become necessary not just sufficient. Imagine a Fed chair trying to explain to Congress that monetary policy necessarily relies on such ingredients for the basic sign of its effect. In the absence of the Fed’s technocratic understanding of such ingredients, the Fed would steer the ship the other way, raising interest rates to raise inflation, not the other way around.
If so, that circumstance radically changes the nature of monetary policy. And one must admit that the scientific basis on which we analyze policy, and offer advice to public officials and the public at large, becomes more tenuous.
I do not mean to disdain frictions including the above list of ingredients. Such frictions surely are important to understand the details of real-world dynamics. Ideally, we add such frictions to simple models that get the basic sign and stability right. The trouble comes when frictions are necessary to the basic sign and stability.
Again, proof is rare in economics. But ex-post patches, in the face of clear predictive failures, are always suspect. Sometimes it is right to patch a theory. Planetary orbits are elliptical, not circular. More often, ex-post patches are epicycles, and the Occam’s razor advice is right.
That advice is not easy. The theoretical interpretation of the long quiet zero bound I have offered is indded strikingly simple. But it asserts that longstanding classic doctrines of monetary economics – that interest rate pegs must be unstable, or that “money” creation must inevitably lead to inflation – were simply wrong. That pill should be hard to swallow.
What are the implications of this experience, and its theoretical interpretation, for policy going forward?
First, we should not unuly fear the zero bound. Much current policy dicussion regards the past zero bound as a narrow scrape with the deflation spiral, and argues for a higher inflation target, or dry powder in the arsenal of unconventional monetary policy and large fiscal stimulus to prevent the spiral from breaking out should we return to the zero bound in the next recession or crisis.
Second, we should not unduly fear large interest-paying reserves. We have discovered that abundant, safe, government-provided, interest-paying electronic money will not cause inflation, any more than government-provided banknotes necessarily did so in the 19th century. (That proposition, the inflationary consequences of paper money, was also hugely contentious.) Much current policy discussion by contrast sees large reserves as permanently stimulative, in urgent need of reduction.
Third, we can live with permanently low and steady interest rates, if we wish so, so long as people trust fiscal policy. If the real interest rates need to rise and fall, inflation will eventually fall and rise, respectively, to accommodate that change.
However, the Fed may wish to vary nominal interest rates according to its best guess of needed real interest rates. Such policy can further reduce inflation volatility. So actual day-to-day polcy need not change radically. The Fed will still raise rates when the economy is doing well, and lower them when it is doing poorly.
Shoals remain ahead. This fiscal foundations that theory needs to understand the stable quiet zero bound could easily fall apart.
It would be easy to misinterpret these results to say that all a country like Brazil or Turkey that wishes to lower its inflation rate needs to do is lower its interest rate.
First, such an interest rate move must be persistent and credible. You can’t just try the waters. Second, it must wait out a potential move in the other direction, via the long-term debt effect, or the many real-world complications discussed above. Most importantly, the fiscal backing and fiscal coordination must be there especially for dis-inflation. Lowering nominal rates cannot cure a fundamentally fiscal inflation.
Successful stabilizations, such as the 1980s in the US and Europe, involved joing monetary and fiscal reform. Conversely, many countries have seen all sorts of monetary stabilization plans fall apart when fiscal cooperation was lacking. Just lowering interest rates will not work with fiscal trouble brewing.
LIkewise, it does not follow from the analysis here that the US, Europe and Japan can just peg low interest rates and sleep soundly. The fiscal foundations of our quiet inflation could evaporate quickly as well.
The fiscal theory says that inflation is determined by demand for government bonds, which in turn comes from the expected discounted value of future surpluses. This is an identity – the only question is which one is in investors’ minds at the moment. Are investors holding lots of government bonds, and not trying to buy goods and services or real assets instead – because they think surpluses will be strong, or because they are willing to hold government debt at very low rates of return? The answer seems pretty clear – the value of government debt is high now because discount rates – expected real returns on government bonds – are very low right now.
But low discount rates can evaporate quickly, especially when government debt is largely short term and frequently rolled over. A change in discount rate provokes exactly the same sort of unexpected inflation that a change in fiscal surpluses provokes. And like such a ghange, there is nothing a central bank can do about it.
Concretely, if in the next moment of economic trouble, when our governments try to borrow another several percent of GDP to bail out trouble financial institutions, or fight a war or a recession, or all at the same time, while simultaneously rolling over a large stock of debt, bond market investors may decide our governments are not serious about long-run fiscal solvency. Investors will demand higher real interest rates to hold government debt, putting more strain on budgets. Investors abandon government debt, driving up infltaion. Such an event feels like a “speculative attack,” a “bubble,” or a “run” to central bankers.
Inflation’s resurgence can happen without Phillips curve tightness. It can surprise central bankers of the 2020s just as it did in the 1970s, just as inflation’s decline surprised them inthe 1980s, and just as its stability surprised them in the 2010s.
1. Hoover Institution, Stanford University and NBER. Also University of Chicago Becker Friedman Institute, Booth School of Business, SIEPR, Stanford GSB, and Cato Institute. This not summarizes “Michelson-Morely, Fisher, and Occam: The Radical Implications of Stable Quiet Inflation at the Zero Bound,” forthcoming in NBER Macroeconomics and available at http://faculty.chicagobooth.edu/john.cochrane. I thank Lars Peter Hansen for helpful comments.