Monday, May 21, 2018

Inflation, Interest Rates, and Neo-Fisherism In Turkey

In most respects, President Erdogan of Turkey is not known for his progressive instincts, but in economic policy he may be the only convinced Neo-Fisherite on the planet who potentially has any power over monetary policy decisions. Erdogan has been at odds with the Central Bank of Turkey, and seems intent on changing the Bank's approach to inflation policy. In a recent interview in the UK, and recent speeches, Erdogan has made clear that he thinks that high nominal interest rates are the cause of high inflation in Turkey, and that disinflation can be achieved if the Central Bank reduces its policy rate.

Erdogan's views have been derided by market participants, and some panic ensued, manifested in a depreciation in the Turkish currency. But, apparently Erdogan's ideas aren't coming out of thin air, as his economic advisor Cemil Ertem has written a defense of Erdogan's views, which does a good job of finding the relevant supporting evidence. He cites John Cochrane's work, and speeches by central bankers, including Jim Bullard, that support neo-Fisherite ideas.

So, what's been going on in Turkey? Here's the time series of CPI inflation rates:
Turkey had very high inflation from the early 1980s until 2002, with the inflation rate sometimes exceeding 100% per annum. But, there was a large disinflation, with the inflation rate falling from a peak of 73% in 2002 to 7% in 2004. Since 2004, inflation has been much lower than over the previous 20 years, and much more stable. However, if we compare the path of recent inflation in Turkey with the United States, the picture looks like this:
So, inflation in Turkey is still considerably higher than in the United Stages, averaging about 8% since 2004, and it has recently been creeping up above 10%. Further, the Turkish central bank has had inflation targets that it has been persistently overshooting for more than 10 years:

What has the Turkish central bank been doing, and what does it propose to do in an attempt to hit its inflation target? From the central bank's most recent Inflation Report:
Given a tight policy stance that focuses on bringing inflation down, inflation is projected to converge gradually to the 5-percent target...
And:
...the disinflation process will continue in 2018 due to the decisive implementation of the tight monetary policy and convergence of economic activity and loan growth to a milder growth path.
So, that seems like boilerplate central banking. "Tight" monetary policy, i.e. a high nominal interest rate target, will lower inflation, and this disinflationary process will get some help from the Phillips curve, i.e. "a milder growth path" will reduce inflation, according to the central bank.

This is the Turkish central bank's inflation forecast:
So, in a little over two years' time, the central bank thinks it will have inflation down to its 5% target. Of course, that's what it thought in 2016:
In contrast to that optimistic forecast, inflation went up, not down, in the intervening period, and is currently well outside what was denoted the "forecast range" in the 2016 forecast.

So, you might conclude that the Turkish central bank is not capable of reducing inflation. But that's a puzzler. A central bank that could reduce inflation by about 66 percentage points in a two-year period from 2002 to 2004 can't get inflation down from 7% to 5% from 2016 to 2018? What's that about? Well, what did the central bank do to produce the 2002-2004 disinflation? Let's look at the path for short-term interest rates and inflation over the same period as in the first chart. Here I'm using an overnight interbank nominal interest rate (the only short-term interest rate I could find for Turkey - if you know where to find other interest rate data, please let me know):
So, in that chart we're seeing a typical Fisher effect - higher inflation is associated with higher nominal interest rates. The spikes in the overnight rate occurred during 1994 and 2001 financial crises in Turkey. Next, let's focus on the disinflationary period and after:
As we all know, if central banks can do nothing else, they are at least capable of pegging short-term interest rates. So, the path we see in the above chart for the overnight rate was determined by the central bank of Turkey. Did the central bank engineer a disinflation by keeping the nominal interest rate high? No. Monetary policy acted to reduce short-term nominal interest rates, and the inflation rate fell.

I'd say President Erdogan isn't as nutty as people are making him out to be - at least in the inflation policy realm. And Turkey is a very interesting example, as it appears to be following the orthodox central banking rulebook: Phillips curve, Taylor rule. There are plenty of countries - Japan being the most extreme - where inflation has been chronically below central bankers' inflation targets, so if Turkey is following the standard rulebook, why is inflation chronically above the inflation target there? Well, that's exactly what mainstream theory predicts. A central banker who blindly follows a Taylor rule (with the Taylor principle in place - more than one-for-one response of the nominal interest rate to changes in inflation) reduces the nominal interest rate target when inflation is low, believing that this will increase inflation, but inflation falls, and the central banker gets stuck in a low-inflation policy trap. Similarly, a Taylor rule central banker who sees high inflation increases the nominal interest rate target, believing that this reduces inflation, but inflation goes up. If the central banker followed the Taylor rule blindly, then inflation would increase indefinitely. But in Turkey's case that may not happen, even without President Erdogan in the mix, due to public resistance to higher interest rates.

There's a lesson here for countries like Canada and the United States, where central bankers are currently hitting their inflation targets. The Bank of Canada and the Fed avoided becoming Japan - falling into the low-inflation policy trap - because they either kept nominal interest rates off zero (Canada), or lifted off from zero (US). But interest rate hikes can be overdone - the risk as that you become Turkey.

12 comments:

  1. I found an English version of the same article by Cemil Ertem on https://www.dailysabah.com/columns/cemil-ertem/2018/05/18/a-theoretical-overview-of-erdogans-remarks-the-issue-of-interest-inflation, in case anyone wants to see exactly what he is saying.

    I don't know if this was pointed out elsewhere, but central banks following Taylor rules have a tendency to make temporary shocks to inflation permanent. Since for various reasons (measurement issues, movements in real discount factors, the fact that a nominal interest rate peg only controls the first moment of the nominal discount factors in asset markets) there can be unexpected shocks to inflation, when these shocks persist long enough to convince the central bank that there is a problem, they begin applying the Taylor rule and they are set on going down either the explosive path upwards or hitting the zero lower bound. If they only held their target fixed, the temporary shocks would disappear and they would be back to hitting their target again.

    Because of this, I think most successes that are traditionally attributed to independent central banks following Taylor rules are actually caused by sufficiently many negative shocks to inflation, which the central bank makes permanent by lowering policy rates in response. In the fiscal theory of the price level, these shocks would be fiscal shocks - a fiscal reform could increase the PV of primary budget surpluses and push inflation down temporarily, and the central bank would respond to the shock by lowering rates, thus achieving a permanent disinflation. This, incidentally, is a story which matches Turkey's experience very well. I find it funny how in this story, central banks which pride themselves on being independent can end up completely dependent on the fiscal authority to achieve disinflation because they don't understand how to use the tools they have to do so independently.

    ReplyDelete
  2. Hi Stephen,
    Again coming back with the same question (not sure why my comment on earlier post didn't appear), possibly in clearer way: How would short-run Fisher effect obtained under RE work out if agents have a (misspecified) Phillips curve model of inflation and a Taylor rule like monetary policy rule as part of their expectations? Btw, another story which may end up being related, the recent Argentine macro dynamics:
    https://www.reuters.com/article/us-argentina-economy/argentines-brace-for-crisis-as-nation-again-seeks-imf-help-idUSKCN1IP0KO
    .

    ReplyDelete
    Replies
    1. As I like to tell my students, we would be bad scientists if we proceeded by reverse-engineering our beliefs, i.e. starting with beliefs about how policy works, and conducting a search to find a model that fits those beliefs. I think neo-Fisherian ideas are useful because they appear to be consistent with the data, and they solve a policy problem that seems to be vexing central bankers. What's curious is that the standard models these central bankers claim to embrace are in fact neo-Fisherian. That's pretty interesting. It's also interesting that when you point these things out to people, they start constructing bizarre models to support orthodox beliefs about how policy works.

      Delete
    2. Again, the question is what happens though if agents have non-RE expectations, because they've been conditioned by the standard central bank view to believe in a Phillips curve and stabilising Taylor rule.

      Delete
    3. But then the question is, where do these orthodox beliefs come from? And, why have they persisted for 20 years or so if they are based on such shaky foundations? I mean, these are smart people well-educated in economics with many of them having significant contributions to the science.

      Delete
    4. They come from elementary undergrad economics textbooks, as far as I can tell. Sometimes this has served as a convenient fiction that seems to sound right, and allows for consensus in central bank policymaking. That doesn't make it right, though.

      Delete
  3. Professor Williamson,

    we know that when the central bank raises the interest rate, (i) Fisher effect sets in which has a positive effect on inflation, and (ii) the liquidity premium has a negative effect on the inflation. Swapping Turkish lira for foreign currency (as a result of higher interest rates) relaxes the collateral constraints as the Turkish private sector highly and increasingly depends on external debt. Turkey's gross external debt stock was $453.2 billion, 53.3 percent of GDP at the end of December 2017. In Turkey, it seems that (ii) dominates (i). The situation reminds me of the Asian crisis in 1997. These countries highly depended on external funds to support their domestic projects. When the foreign lenders started to withdraw their funds, the result was a strong depreciation in domestic currencies. My hypothesis is that it is not the case that Neo-Fisherism fails in Turkey, but there are stronger financial frictions associated with the private sector and the ineffectiveness of the public sector than the Fisher effect. Lastly, Turkey's short-term external debt statistics are given by
    http://www.tcmb.gov.tr/wps/wcm/connect/b19b5048-dd64-4075-ae01-47b2fd3ede50/st.pdf?MOD=AJPERES&CACHEID=ROOTWORKSPACE-b19b5048-dd64-4075-ae01-47b2fd3ede50-mdO36yX

    ReplyDelete
    Replies
    1. I think there's little to no question that in the short run raising nominal rates seems to cause disinflation, looking at asset market data, for instance. John Cochrane described this as the "mild neo-Fisherian view" - it tosses out the old-Keynesian doctrine of unstable nominal interest rate pegs, but it maintains a short run negative relationship due to the liquidity effect. The three-equation NK model, closed using the FTPL, is consistent with this in the presence of some long term government debt (since the central bank targets ON rates, anything longer than 1 day is long term in this setting). The problem is that since the underlying mechanism is so different, raising nominal interest rates to cause a currency appreciation or cause lower inflation is only a temporary measure; you get a little bit of appreciation today at the expense of committing to a long future path of higher (nominal) depreciation. All of this comes out of the standard FTPL model, with or without pricing or wage setting frictions, so we don't need weird "gross external debt frictions" to make it work. The fiscal theory story is also consistent with the Asian financial crisis period,

      You don't need external financing to get a big fall in the value of your currency. There are some connections, like how temporary "stimulus" measures tend to push up the current account deficit while also worsening the fiscal situation of the government, but if you look at the data you will find that no country has experienced a currency crisis while there was confidence in the fiscal authority's ability and/or willingness to repay its debts, regardless of the current account situation. (The US has been running reasonably high current account deficits for two decades now without a hint of currency crisis, for example.)

      The focus on external debt matters insofar as the debt is being guaranteed by the government in some way - if when things go south bondholders of companies which took a lot of foreign exchange risk are going to get bailed out, you get a situation of perverse incentives which usually ends in some kind of crisis just due to the sheer volatility of exchange rates. If, however, the private sector is taking on the downside risk as well, then trying to sabotage them from doing that in the name of "correcting imbalances" can be quite counterproductive. Krugman-style Keynesian "multiplier" models of "currency crisis" are just unrealistic.

      I don't know why Turkish bonds would actually be better collateral than US treasuries - perhaps if you really need nominal insurance it can be helpful, but even in that case it seems much safer to me to buy foreign government bonds and hedge the exchange rate risk using forward contracts or cross currency swaps. The Turkish banking sector does this, so despite having a substantial amount of foreign exchange liabilities, their net position in the market is close to zero. The only explanation that I think is sensible is that financial regulation and "macroprudential policy" may force financial intermediaries to hold Turkish government bonds specifically, but I'm not sure about the status of reverse repos held at the central bank in that case, and surely companies won't sit on cash at double-digit inflation. Unless the central bank can actually create government bonds (Turkey does not have IOR), I don't see how to ease this kind of regulatory collateral constraint using monetary policy at high inflation.

      It's not clear to me how "financial frictions" matter - surely with market segmentation you can get some liquidity effect of monetary policy, but the market for Turkish bonds is hardly segmented. If you're more explicit about the kinds of frictions we have to think about it would be more helpful.



      Delete
    2. "I think there's little to no question that in the short run raising nominal rates seems to cause disinflation, looking at asset market data, for instance."

      The way you put that makes it clear that there are plenty of reasons to question this. "Seems to.." and "looking at...data" isn't very convincing.

      Delete
    3. I looked at some time series data after making that comment and you're right - it's quite difficult to find this "conventional" behavior in the data past a daily trading horizon. For instance, if you detrend both dollar exchange rate and fed funds rate data by subtracting a moving average to just look at monthly fluctuations, there's in fact a small negative correlation in the data between these two since 1999 - the fed funds rate falling relative to "trend" (12 month MA) seems to be associated with the exchange rate appreciating relative to "trend". A correlation of -0.15 isn't much evidence of anything, of course, but it's fairly surprising that despite all of the talk about how a central bank should raise interest rates to support its currency, there's no evidence to support this even in a monthly horizon.

      It looks like I still have too much faith in the orthodox beliefs about how monetary policy works. I was going to edit out that part of the comment if I could after doing this, but Blogspot wouldn't let me...

      Delete
    4. Thanks for the information. In general, it's extremely difficult trace the effects of policy in the data, and it doesn't help that everyone - central bankers, financial market participants, you and me - have a murky understanding of what is actually going on. Over time, people tend to coordinate on ideas that sound right, on various levels. You can sell the idea to your colleagues (maybe), and to the general public. It can be disturbing to find that there are cases where people endorse a particular theory, and the theory itself doesn't actually say what its promoters claim.

      Delete
    5. I think it does not help that ideas often get deformed in the process of trying to sell them to the public. When the public hears about the central bank controlling interest rates, they don't think about how the Fisher effect means a higher interest rate target also has to bring higher inflation with it, or monetary policy can't move real interest rates around forever. They assume that the central bank can set the rate they pay on their mortgage while holding "everything else" constant. This is also strongly suggested by the language a lot of economists use when talking about monetary policy in public - the transmission mechanism is often explained in terms of how low interest rates are going to lead to more "demand", which "pushes up prices". Even economists who have worked on more modern NK models use this language, which is confusing since that's not what actually happens in their models.

      I had the opportunity to observe some of the reaction to the neo-Fisherian idea in Turkey; and market participants, journalists, economists, etc. all seem convinced that conventional beliefs about policy are supported by volumes of data which cannot be found when you actually look for it. The combination of no data and no explanation for how a policy is supposed to work should be a warning sign that something might be wrong, but people seem to find a way to convince themselves that everything is okay. It's very weird, given that if I did not know better, I would think that the mild form of neo-Fisherism that raising nominal interest rates and holding them there eventually produces high inflation would be one of the most uncontroversial statements a person could make about macroeconomics. Alas, it is not so.

      Delete