Scott Sumner and Cantillon Effects–Part 2

The intense (heated?) debate around the Cantillon Effects after an injection of money has produced a new post by Scott Sumner. Scott Sumner argues that it doesn’t matter where money in injected first because all possible injection points produce the same initial result; buying T-Bonds. I don’t think that Sumner is wrong on this, I do think, however, that this is just a first step of analysis and that Cantillon Effects depend on what happens after the money gets into the market through the exchange of T Bonds.

First things first. Sumner is not saying that there are no effects at all. When discussing the first myth, he acknowledges that the bond seller earns a commission when selling T Bonds. He thinks, however, that the amount of this benefit is trivial at a macroeconomic level. As I said on my previous post, it is different to say that there are no Cantillon Effects than to to say that the empirical relevance is considered to be trivial. Are all these trivial effects summed together still trivial? Could there be a mechanism such that this benefit is not trivial anymore? Is it the decision to bail-out or to not bail-out Lehman Brothers trivial? Even if the Lehman Brothers example falls outside Sumner’s scenario, it does exemplify a non-trivial effect on the first steps of a policy that buys assets from a financial institution.

I think his treatment of the myth 4 helps to square what Sumner is trying to say. Sumner suggests that given the news of a monetary injection the prices of (financial) assets should rise instantaneously, so there’s no benefit in receiving the money first. This means, however, that there is a drawback if money is received later on (this is not mentioned by Sumner). Sumner says that a person that hears the news about the expansive monetary policy thinks the price of gold is going to raise and so he uses his cash or takes a loan to buy more units of gold. He, therefore, is not constrained to buy gold. Therefore, it doesn’t matter how the money gets into the economy. The bank can always lend to him the money he needs to buy more gold.

I don’t see, however, how this goes against Cantillon Effects. There are two ways I can buy more gold. With money already in my balance or by taking a credit. To argue that someone uses his savings to buy more gold means that his money demand is decreasing. We have two, rather than one, effect. Injection of money and a change in money demand. We start to move away from the ceteris paribus analysis. But this also implies that the price of savings is going to change. Therefore, the relative prices of savings with respect to other goods is going to change. That is what Cantillon Effects are about, changes in relative prices.

The second option is to get a loan a to buy the units of gold. If this person gets $100 in loan, means that those $100 cannot be given to someone else. Some buyers, and not others, can make use of the $100. What should happen for this to be immaterial with respect to Cantillon Effects? We need to distribute the $100 equivalently among all consumers. Namely, we need the helicopter.

Now, regardless of this, I think is inaccurate to stop the analysis where Sumner does. The Cantillon Effect is the result of n-steps, not 1-step. Let’s say that step-1 is the same for any potential way to inject money in the economy as Sumner describes. What happens in steps 2 to n is as important as step 1. Cantillon Effects are not just about asset prices, but about all prices in the economy. Person 1 may use the $100 to buy gold. But person 2 to buy a new house and person 3 to buy a new car… etc. Financial asset prices can react immediately because the new money is already in the financial system, but goods and services outside the financial cannot raise if the purchasing power of their clients have not increased yet. The focus on the financial market risks to do away with the problems involved in the rest of the economy.

Finally, Sumner says that “In aggregate, the total level of nominal purchases is constrained by the amount of currency in circulation. But not at the individual level. Hence being the first to get the new money doesn’t confer any advantage at all–as the new money has no more purchasing power than the existing money.  A dollar is a dollar—and a $100 bill is a $100 bill.”

It is true that the new money has no more purchasing power than the existing money. But is also true that there is more money in the economy and therefore the total purchasing power increases. How is the new purchasing power going to be distributed in the economy is what gives the initial kick of the Cantillon Effects. The final result depends on the complete sequences of kicks, not only in the first one.

It is also true that the Fed buys bonds at market prices, and that therefore the operation is a swap of assets and the size of the balance sheet may remain the same. But it also true that the composition of the asset (how much cash versus how much T Bonds) can affect how much can be bought today by however receives the swapped cash first (that is, after the bank). Eventually the banks need to decide who to lend the new amount of money, or if they are going to invest it themselves. Wouldn’t, for instance, a policy favoring the housing market potentially result in a housing bubble and a financial crisis? The institutional framework can have a bearing on the empirical extent of the Cantillon Effects.

Still, there are $100 extra that will land somewhere. Even if that somewhere immediately results on the same effect on the T Bonds, they will get into the economy afterwards with potential different paths. It may well be the case that it doesn’t matter who gets the money first, but it may well be the case that it matters who gets it second. The story in the broken window fallacy can help to illustrate the point. There are two ways the $100 value of the window can be spent in the economy. Give it to the window maker to repair the broken window or use it to buy a new suit by the shop owner (forget about the lost value of the window). In both cases the money flows differently through the economy even if the $100 were initially in the bank and got there through an exchange of T Bonds.

Still. I’m not sure of what Sumner is trying to say. I think his second post would been much more clear if it  had followed something like this: “Yes, there are Cantillon Effects. No, they do not depend on how money is initially introduced in the economy. No, I don’t think the empirical extent of the Cantillon Effect is important.” This is the best interpretation I can make of Sumner’s argument.

PS: Note that Sumner is not saying (as far as I can tell) that there are no effects by the Fed buying  Bonds. He’s saying that the effect is the same regardless of who he buys them from and that the difference between buying T Bonds through channel A or B is not economically relevant.

5 comentarios en “Scott Sumner and Cantillon Effects–Part 2

  1. I read both posts and think that yours is a great contributuion 🙂

    I also thought point 4 was the key to Sumner’s argument, and think that the problem is due to so many impossible assumptions. Will everyone go and buy gold? Are we all James Turk readers? And if so, as you say, how will this be possible? We need to abandon ceteris paribus and then we will have a change in relative prices…

    And then another thing, when he says a 100$ bill is 100$ bill. Well, that is true only in a vacuum, but if we consider time (or step 1 and step 2 in your words) then that is the most untrue statement ever and it’s precisely what we are talking about.

    Me gusta

  2. I think there’s a combination of 3 things:

    1-. Sumner not being very clear (at least for me).

    2-. A too narrow scenario, all seems to start and end at the financial market where there are no wealth constraints and a «perfect information» assumption is less debatable (Fed policy is publicly known) [Not that this overlooks the problem of different interpretations of the same policy, expectations are subjective as well].

    3-. The time problem. Are the Cantillon Effects «unimportant» in economic terms after 5 or 6 years of an expansionary monetary policy? What about the housing bubble (though I suspect Sumner will reject the idea that the monetary policy between 2002 and 2007 was a major driver of the financial crisis in 2008).

    In the meantime, I make a «fundamental mistake» by not assuming a (plausible) change in money on part of consumers:

    Me gusta

  3. Not necessary. The creation of T Bonds (Gov. Debt) is fiscal policy, not to exchange them. In the first 3 cases in Sumner’s first post there is no creation of new T Bonds. The Fed buys already existing securities.

    Number 4 is a little different. The government has a deficit, therefore it needs to issue T Bonds to pay his expenses (salaries in Sumner’s example). Therefore the Fed gives money to the Treasury in exchange of the new T Bonds. But this case can be divided into two effects (Nick’s Rowe’s post). The pure monetary policy of buying T Bonds at market prices and the issuance of new bonds by the Treasury. Therefore, it is not monetary policy, but fiscal policy, what produces Cantillon Effects.

    I agree (mostly) that things happen as Sumner argues as far as his argument goes, but I disagree on where he stops his analysis. In all 4 cases the Fed is creating *new* money and unless we expect all prices to go up together, then we have the Cantillon Effects. What happens with the prices of goods and services when the new cash goes from the financial market to the rest of the economy? This is the problem I don’t see clearly address. Maybe I missed it in the long list of comments.

    The problem is not whether or not the particular injection point matters, but it if relative prices change due to monetary shocks. I’m not sure I completely agree with Rowe’s definition of Cantillon Effects. The fact that the injection point doesn’t change the final effect doesn’t mean there are no Cantillon Effects. It means they are always the same. To isolate those effects into «fiscal» rather than «monetary» policy sounds to me as a way to isolate the problem and keep monetary policy free of these effects. If new money is injected into the economy, the size of the balance sheet of bond holders may not change, but the cash balance does change.

    If monetary expansion causes inflation, then one should expect relative prices to be affected. The extent of this effect is, of course, an empirical question to be addressed in each case. It is not the same a one time shock of money supply (as seems to be the implicit assumption in the posts) than to assume that the monetary expansion takes place for several years.

    Maybe Sumner is right that Cantillon Effects are less relevant today in USA given the present institutional conditions (with respect, for instance, to Cantillon times), but present institutional framework was not always in place and differs from country to country. Doesn’t Argentina suffer Cantillon Effects? (if there were no Cantillon Effects then there would be no need for price controls in the first place!). Probably the argument would be that the Cantillon Effects in Argentina are due to «fiscal» policy. But is still the case that is the central bank expanding money supply. The government may have deficit, but it can issue domestic debt, in this case there is no increase of money supply and no inflation. So, isn’t monetary policy still a driver of Cantillon

    Nick Rowe’s post:

    Me gusta

  4. Pingback: Scott Sumner and Cantillon Effects–Part 2 « ESEADE

Los comentarios están cerrados.