Wednesday, September 12, 2012

Monetary confusions abound

Macroeconomics, especially monetary economics, is hard. It is not hard like stochastic calculus is hard. It is hard at the level of intuition itself because of its connect-ALL-the-dots nature. The outmoded and narrow pedagogy that most of us are exposed to doesn't help. The charged politics of macro policy-making further blinkers analysis.

Of late, the economic discourse in India  has been increasingly critical of the incumbent UPA-2 government's macroeoconomic management. One of my frustrations with this discourse has been its failure to focus on what is truly important and truly wrong, while shooting critical arrows at random targets. The increase in the analytical content of public discourse, as well its right-ward shift is surely to be welcomed. However, the quality of the analysis hasn't necessarily been strong. Part of this is because of the inevitably charged politics. But I believe a more significant role is played by the fact that macro is, simply put, hard.

A good illustration is this article by R Jagannathan in FirstPost. He also links to an earlier piece by Latha Venkatesh, also in FirstPost. In what follows, I will basically critique these articles, using them to hopefully illustrate a few key monetary concepts. I wish to make it clear that I don't intend this as a personal criticism of either of them Mr. Jagannathan's writing performs the indispensable function of speaking truth to power - witness his dogged pursuit of the shady dealings behind K Abraham's removal from SEBI, for example. It's just that his article is a veritable goldmine for my purpose here.

I'll tackle only one of his arguments in this post here, leaving the rest for later.

Ms. Venkatesh's article, which discusses the possible impact of a 1% cut (from 5% to 4%) in CRR in India, asserts that the money multiplier is between 4 and 5 in India. But it's all a little confusing because she's not talking about an injection of monetary base, which is what the money multiplier is supposed to work on. 'Liquidity' is being 'freed up' by a cut in CRR, and the monetary base is being held constant. Later she is more precise as she talks about the money multiplier as that which when multiplied by the liquidity freed up as a result of the CRR cut gives us the corresponding increase in deposits. It is tempting to assume that this is the same as the textbook money multiplier but this is not so. 

Let's call this the modified money multiplier (mmm). I'll skip the algebra but it's easy to show show that mmm = 1/(K+CRR) where K = currency/deposits is the currency drain ratio. [the traditional money multiplier mm = (1+K)/(K+CRR)]

We know that Ms Venkatesh is discussing a cut in CRR from 5% to 4%. The other variable is K - so what is K for India? Bank deposits currently stand at Rs. 70 trillion, currency is Rs. 11 trillion. K = C/D = 11/70 = 16%. So, mmm = 1/(0.16+0.04) = 5. Ms. Venkatesh seems correct about the size of the modified money multiplier.  

But if you back out from all these numbers a bit, what is driving the size of these multipliers? Even at 5% CRR is an order of magnitude lower than K. Multipliers are being driven almost entirely by K, the currency drain ratio. What does that behavioural quantity have to do with actual banking operations or CRRs or anything? Can it not change if the quantity of deposit grows a lot? A simple illustration through a banking system balance sheet with numbers will show my point.

What happens when CRR is cut?

Consider the following stylized balance sheet of the Indian banking system.


CRR is, as now, 5%. Consider the extreme case - what happens if the CRR legal requirements is completely removed? This frees up reserves of 5 units. Now the actual reserves will not go down to 0, because banks will want to have some for inventory-theoretic reasons. Let's say this 'desired' CRR (as opposed to a legally mandated CRR) is 1% of deposits. I have chosen 1% because that is about the % that banks keep innon-crisis times in countries where reserves are miniscule or not legally mandated - UK, New Zealand, Canada - actually keep. (Yes, there are such countries, and no they don't suffer from inflation on this count. We will soon see why).

So now we have free reserves of  5 units and a desired CRR of 1%. What happens to loans and deposits? They can in  theory jump five fold! After all, with reserves = 5, a desired reserve ratio of 1% will give deposits of 500. 

Fig. 2

A five-fold increase in deposits sounds crazy, right? Let's try another scenario. What if banks just purchase safe assets with the freshly freed up reserves ?

Fig. 3

So will deposits go to 500 or remain 100 or be something in between? The multiplier story thinks it may have the right answer. Deposits will go to 120, as the freed up reserves of 4 units will lead to increased deposits of 20 via a modified money multiplier of 5. If you keep the proportions of safe assets and loans constant, this is how the banking system's balance sheet may look like. 

Fig. 4

Now what if the economy with the banking system of Fig. 3 was to grow by 20% organically? It would end up with a banking system like this

Fig. 5

     Fig.5 and Fig.4 are very similar. Indeed, from a monetary perspective they're nearly identical. Shouldn't they represent the banking systems of two very similar economies? They do, and that's the problem with the argument as presented by Mr. Jagannthan and Ms. Venkatesh.

What I've gone through several hoops is to try and demonstrate that in the (modified) money multiplier story, if the CRR were to be completely removed, we should expect that the aggregate nominal economy- not just the quantity of bank deposits - is larger by about 20%. 

So why is growing the nominal economy  by 20% a problem? Even if you fear inflation and think that the entire 20% extra spending goes into prices (and nothing into quantities), sales and excise taxes collected by the government on the extra spending would by themselves close the fiscal deficit by 2% odd. Mr. Jagannathan doesn't realise this, but he seems to have uncovered a miracle way to close the fiscal gap that he's so worried about - government expenditures are largely salaries, so keep the path of government salaries as previously expected (government salaries are only revised once in 10 odd years and the yearly inlation adjustment is not built for a 20% increase) and remove CRR. If some part of the 20% goes into real GDP rather than prices, even better.

The miracle way sounds absurd, and it is. And that's because the modified money multiplier is absurd. Even at a seemingly text-bookish value of 5, it is absurd for the same reason that the situation in Fig. 2 is absurd. You cannot make an economy 'jump', not five-fold nor 20%. 

Fig. 3 is the most likely scenario of a post-CRR cut banking system. Dropping CRR increases the return banks make on their safe assets and thus increases bank profits and re-capitalizes them. That part of Mr. Jagannthan's story is correct. But it doesn't do too much for bank deposits, lending or nominal spending. CRR is used by RBI to manage liquidity stresses in atypical situations - ahead of an advance tax payment date, for example. Or, it's used as fiscal mechanisms to extract (or give back) profits from (to) the banking sector. But it's not a mechanism to influence bank lending or aggregate spending. For that the RBI uses, in line with central banks across the world, policy rates.

So if the modified money multiplier is incorrect, doesn't that cast doubt on the traditional money multiplier as well? It does, and that's why I referred to bad pedagogy as one of the reasons why deciphering macro can be hard. While the money multiplier is trivially true as an arithmetic identity, the direction of causality between deposits and reserves is far from obvious, and things depend more on price mechanisms and overall central bank targets than on any quantity mechanisms. While central bank guidance and policy can be forward looking and pro-active, its operations are mostly defensive or reactive. The CRR is an operational tool to manage near-term liquidity, not a policy tool for macroeconomic management. The reserves 'freed up' by cutting the CRR will somewhat increase lending on account of higher profitability for banks, but will mostly just have to be mopped up by the RBI, through open market or reverse repo operations.

 (Reverse repos are operations where banks 'lend' to RBI at interest - they essentially have a deposit at RBI that they promise to draw down at a certain date. Of course they can simply roll choose to roll the deposit over, thus in effect creating reserves that earn interest.

Indeed, if not for operational continuity and legacy reasons, the RBI could choose to drop the CRR to 0 tomorrow without changing the path and effect of monetary policy by much.

The RBI is monopolist on reserves. A monopolist can either set prices or quantities. As a setter of price (reverse repo rate), the RBI forgoes control of the quantity of reserves. The CRR is not a quantity control, it's simply a bank tax. Unless you believe that a tax rebate of ~0.3% of bank assets (7%  interest rate * 4% CRR) will encourage so much lending and spending that it will prop up the economy by 20%, you shouldn't talk about money multipliers for CRR cuts. Not if it's a 1% cut as discussed by Ms. Venkatesh, nor if it is removed completely as in my thought experiments.