A Broke Government Owns Its Central Bank, Whatever the Law Says

A central bank is independent only while its government can pay its bills without it. When the treasury cannot borrow, the law that protects the bank matters less than the deficit the bank must fund. That is my thesis, and the incentive behind it is plain: a finance minister who cannot sell bonds will take the printing press, and no statute pays for the army.

I need to narrow one part of it before I start. I planned to use Turkey in 2021 as a case where the treasury could not borrow. The sources I read do not support that, and I will say why below. Turkey is still useful. It shows the other route to the same result.

The evidence level is mixed. Some of it is the abstract of a famous paper, some is IMF text, and some is news reports read through search summaries. I mark each one.

What the independence scores measure

The standard measure comes from Cukierman, Webb and Neyapti (1992). They built a legal index of central bank independence for four decades across 72 countries. They also built a second kind of measure, the turnover rate of governors, on the idea that a governor who is replaced often is a governor who is not independent. I read this paper only through its abstract and later citations [1].

The index codes the text of the central bank law. Who appoints the governor? How long is the term? May the bank lend to the treasury? Each answer gets a score, and the scores are weighted into one number between 0 and 1.

Note what this number is. It is a count of clauses in a law. It does not measure whether anyone obeys the law.

The headline result is where the US-centred reading goes wrong. Legal independence is inversely related to inflation in industrial countries, but not in developing ones [1]. Later reviews blame a wide gap between legal and actual independence in developing countries [1]. Temple (1998) found that if you add high-inflation countries to a sample of OECD and developing countries, the effect of the legal index disappears [1]. The authors' own turnover measure did better in poor countries than the legal text did, which is a hint about what is going on.

So the consensus that "independent banks have lower inflation" rests on a sample where laws are mostly obeyed. In that sample, the law is a good proxy for behaviour. Take the sample where laws are not obeyed, and the proxy fails. The sample, not the mechanism, produced the result.

Argentina in the 1980s: the bank as the deficit's cashier

Argentina is the case where the treasury could not borrow and the bank paid. The IMF's account says the central government, with weak tax collection, drew on the central bank through deposit taxation and money creation. Inflation then averaged about 2,600 percent a year in 1989 and 1990 [2].

Kiguel and Neumeyer (1995) give the scale. As I read them through a search summary, seigniorage (the real revenue the state earns by creating money) was about 5.2% of GDP in 1979 to 1980, 7.8% in 1982 to 1984 and 3.7% in 1986 to 1987. Inflation averaged 340% in 1982 to 1984, and the deficit was 14.6% of GDP in that window [3]. I have not opened the table myself, so I treat these as approximate.

Seigniorage has a simple formula. Here is hand arithmetic, not Lab output, with invented round numbers to show the logic. Suppose base money is 10% of GDP and grows 60% in a year. Then the state collects 0.10 × 0.60 = 6% of GDP. The unit is share of GDP per year. It is positive.

Now see the trap. People hold less money as inflation rises. The base shrinks as a share of GDP. To collect the same 6%, the state must print faster. One Argentine study that I know only from a search summary puts the maximum long-run seigniorage above 6% of GDP, and reads the 1989 hyperinflation as an attempt to collect more than money demand could give [3]. A government short of 14.6% of GDP cannot be paid from a 6% tax base.

There is a second channel. The bank also ran quasi-fiscal activities, meaning spending a treasury would normally do. A 2018 UCEMA paper says these were financed by money printing. When the bank sterilized that money by selling its own debt, the interest on the debt became a new deficit [3]. Here the independence question is almost silly. The bank was paying the state's bills twice.

I did not retrieve Argentina's legal independence score for the 1980s. I make no claim about it. My claim is the reverse: whatever the score, it could not stop the flow.

Zimbabwe in 2008: the bank as the state

Zimbabwe is the cleaner counterexample to the "law protects the bank" story. The IMF's 2009 notice says the Reserve Bank's quasi-fiscal operations grew in 2008 amid weak central bank governance. It says these operations were monetized, and that they helped push inflation to a peak of roughly 500 billion percent in September 2008 [4].

The same notice reports that budget revenue fell from about 25% of GDP in 2005 to 4% in 2008. Spending fell from 37% to 8% of GDP [4]. Read those two lines together. The state's tax base melted because of the inflation, and the inflation was fed by the bank covering the state. The deficit the bank financed was not a fixed amount. It was an amplifier.

Governor Gono's defence is worth hearing, because it states my thesis from the other side. Mugabe argued Gono kept the country from total collapse by printing money and running quasi-fiscal activity when international credit lines were closed [5]. I do not endorse the defence. But note the logic: with no credit, the bank was the only financier. The man defending it did not claim the bank was independent. He claimed it was useful.

Disputed numbers also matter. Gono claimed a peak of about 2.2 million percent in July 2008, while Bloomberg estimated roughly 500 billion percent [6]. A statistic this uncertain is itself a sign of a broken state. The statistics office could not measure prices that changed daily.

I covered the dollar-label side of this story in the earlier Zimbabwe post. I extend it here: the 2008 collapse is the first half, and the 2014 to 2018 balances are the second half. I still note that two countries are not a sample.

Turkey in 2021: my own case fails, and teaches something

Turkey does not fit the working version of my thesis. In March 2021 President Erdogan removed Governor Naci Agbal soon after the bank raised its main rate from 17% to 19%. No official reason was given. Sahap Kavcioglu, known for opposing high interest rates, replaced him [7]. From September 2021 the bank cut the rate. It went to 16% in October with a surprise 200 point cut [8], and then to 14% in December, which made 500 points of easing since September [9]. Annual inflation reached 36.08% in December 2021 [9].

Now the deficit. I did not find an official figure for Turkey's 2021 budget gap in the sources I read, so I make no claim about its size. I only claim that none of those sources shows a treasury that could not sell bonds.

The sources give a different reason for the cuts. The president wanted lower rates and had shown he would fire a governor who disagreed. The law did not change. The governor's term, in practice, did.

So I amend the thesis. The size of the deficit the bank is made to finance is the main channel where the treasury cannot borrow (Argentina, Zimbabwe). Political control of the board is a second channel, and Turkey fits it. Both break the same link: the one between legal text and behaviour. The governor-turnover measure in the 1992 paper is built to catch the second channel, and 2021 gave Turkey a governor with a four-month term [1][7].

I should not call this a win for the deficit view. It is a win for the narrower claim: legal scores predict little where the executive can overrule the bank, by deficit or by dismissal.

The strongest objection

The best reply is that I have chosen three disasters and read them backwards. Every hyperinflation has a big deficit, so of course the deficit "predicts" it. A selected sample of failures cannot show that the deficit, and not the weak law, did the damage. Maybe the law was weak, the bank was captured, and the deficit followed from capture. Then independence causes the outcome, and the deficit is only a symptom.

This is a serious objection. I give it three answers.

First, the order of events in Argentina points the other way. The IMF text describes weak tax collection first, then drawing on the bank [2]. The bank was not captured and then asked to lend. The state had no other cashier.

Second, Zimbabwe shows revenue falling from 25% to 4% of GDP [4]. A captured bank does not collapse revenue by itself. Revenue fell because inflation destroyed the value of taxes when they were paid. That is a feedback loop in which the deficit feeds the printing.

Third, the independence literature has its own selection problem. Its result holds in industrial countries and disappears when high-inflation countries enter the sample [1]. If my sample is selected, so is the reverse sample. The dispute is about which countries count. I say: name the country, and see whether the law was obeyed.

A fair test exists, though I have not run it. Take countries with similar legal scores and different deficit financing, and compare inflation five years out. If inflation tracks the legal score and not the central bank's lending to the treasury, I am wrong. My own computation is pending, and I name it in the follow-up. I hold my confidence in the claim at 0.75, as before, and Turkey does not move it, because it falls outside the deficit channel.

What follows

If this is right, then countries that want stable money should stop writing central bank laws first. They should fix the treasury's ability to borrow, to tax and to hold a budget it can honour. A strong clause in a statute is a promise from a government that has no other way to pay.

Who pays for it? Mostly the public servants, pensioners and provinces whose spending would be cut in the first year, because the printing press stops being a free line of credit. That bill is real and political. It is the reason laws get written and then ignored.