Thursday, June 25, 2015

The Long Unwind of Excess Money Demand

Two years ago I was part of a panel discussion on Fed policy at the American Enterprise Institute. I talked about why money still matters as way to make the case that the economy was still being plagued by excess money demand. This problem occurs when desired money holdings exceed actual money holdings. This imbalance causes a rebalancing of portfolios toward safe assets away from riskier ones and in the process causes a decline in aggregate demand. This is one way to view the long slump.

My argument back then was that even though the excess money demand problem peaked in 2009 it still was being unwound in 2013 and therefore still a drag on the economy. Among other things, I showed as evidence a chart portraying the sharp rise in the share of household assets that were liquid and noted it was still elevated in 2013. You can see it in the video or slides from the panel, but I thought I would update the chart in this post to see how much progress we made on the excess money demand problem.

The figure  below shows the liquid share of household assets and plots it against the U6 unemployment rate. In both the 2001 and 2007-2007 recessions this measure leads unemployment. It appears there may still be some residual excess money demand, but a lot of progress has been made. The most striking observation for me, though, is how long it has taken for household portfolios to return to more normal levels of liquidity. It has taken six years and appears not completely done yet! Remember this the next time someone tells you that the aggregate demand shocks were all were worked out years ago.

This next figure plots the household liquid share against small businesses concerns about sales. The latter measure comes from the following question in the NFIB's Small Business Economic Trends: "What is the single most important problem facing your firm?"  There are nine answers firms can chose, but below I plot just the one about concerns over lack of demand. Here too you see a good fit, a non-surprising result.

The fact that a large portion of the liquid share of household assets has declined over the past six years shows that its elevated rise was not a structural development but a cyclical one. It also suggests that more could have been done to hasten its return to normal levels. That is, more could have been done to satiate the excess demand for money. It should not take this long for a business cycle to unwind.

P.S. This is not to say there were no structural problems over the past six years. There were plenty. But what the above analysis speaks to is that many observers grossly underestimated the size and duration of the aggregate demand shock.

Update: I count cash, checking, saving, time, money market mutual funds, treasuries, and agencies as the liquid portion of household assets. Here is the link to the data in FRED.


  1. It's the denominator that's the problem, not the numerator. We destroyed the housing market. Mortgage markets still aren't growing. So, there is $10 trillion or so in home equity that is missing from household balance sheets. Liquid assets/Total assets would be recovered if we let home values recover. In fact, I think that's the only way for it to recover, because it is home values that are out of whack.

    1. Kevin, the numerator is very much a problem. Below is a link that shows the numerator--cash, checking, savings, time deposits, mmmfs, treasuries, and agencies--held by households.

      Note from this figure that the liquid assets continue to grow even though overall household assets have declined. In normal times, we think of the demand for money growing as a function of nominal income. But we know nominal income fell and portfolios fell so we would expect, all else equal, the level of liquid asset to decline as well. But it didn't, it continued to grow (even spiked in 2009). This growth of liquid assets in the absence of income or other asset growth is the problem. It's both a sign and cause of increased risk aversion. Theoretically, we would expect such growth to lead to a decline in asset prices and income. In short, most of the causality is running from increased liquidity demand to broken household balance sheets.

      See my review of Mian and Sufi's book for more on this point:

      Also, I have an older post that looks more closely at the growth of the liquid assets:

    2. Thanks for the links.

      I think there is still something to think about here, though. As a start, rising home values would both increase the denominator and decrease the numerator, as households increased their access to home equity lines of credit.

      Also, regarding this sentence: "Had the interest rates been allowed to reach their negative market clearing (or 'natural') interest rate level, then the present value of housing would not have fallen as much"

      I think if you look at home prices, relative to long term real bond yields, there was a disconnect. The home market collapsed because catastrophic monetary policy destroyed the mortgage market. Relative yields, in terms of present values of expected future rents, implied a much higher yield than bonds, relative to any previous period. Even now, home price/rent ratios imply a yield near past ranges, even though long term bond real yields are well below any previous experience. The ZLB was not a factor in home prices. Home prices had basically bottomed out by the time the zero lower bound was an issue.

      Now, I do think there are interesting things to think about regarding an inverted yield curve. I wonder if in low inflation contexts, the yield curve can't invert enough, because if a lack of liquidity is the cause of the correction, falling rates would correlate with rising home prices. But, to bid up home prices, buyers need cash. I wonder if this dynamic keeps the yield curve flat during monetary corrections.

      I think this might have been a factor from 2006 to 2007, but once housing started free falling, arbitrage between homes and long term real bonds had completely failed.

    3. Kevin, yes, there is something interesting here and I don't want to claim the denominator played no role. What I am saying is that if you believe monetary policy played the key role in causing the bust--and I think we both do here--then it had to work initially through changes in the numerator. No doubt the subsequent declines in the denominator reinforced changes in the numerator.

  2. Here's the same graph with real estate market values instead of equity.

    Home balance sheets would be doing just fine if we allowed housing markets to function efficiently.

  3. It does take this long to unwind when the financial seizure is that bad. Instead of jogging at a good speed, it ends up being gingerly moving through the forest.

  4. I think the implication of this post is that financial seizures need not be that bad if fiscal/monetary policy response is appropriate.

  5. "more normal levels of liquidity"

    How do you define normal? A look further back shows liquidity levels collapsed between 1990 and 2006. The 2013 level matched that of the late 90's, a more prosperous time than normal, I would argue.

    Some of this was due to sustained innovation (credit cards, automatic overdrafts). The latter part, however, was arguably due to failed innovation (HELOC's) and a mistaken belief that the period of low financial volatility would persist.

    Lastly, as others point out, there is a strong denominator effect. It is not clear why households should view their liquidity needs as a percent of total assets. The traditional view is to hold precautionary balances as a minimum percentage of the household budget (i.e. months of income or expense). Since these have barely budged in real terms since 2008 for the median household, this might explain why your numerator might be relatively stable.

    1. Diego,

      The 1990s change was a structural one that occurred as a result of innovation allowing household to more easily tap into equity. If you take deviations around the trend of the liquidity ratio series since the 1950s you will see similar deviations (though not as big) for other recessions.

      But here is the bigger point. Your question is moot because this ratio has been falling for six years back toward some pre-crisis value. Where exactly it is, no one knows. But no one can now say the high liquidity ratio of the past few years a new normal (i.e. structural) given the track record. That is the point of my graphs, as they show similar declines in unemployment and small business concerns about demand which would most would see as cyclical changes.

      Regarding the denominator, I am arguing that it is largely endogenous to the the stance of monetary policy. And the stance of monetary policy is shaped in part by money demand. As I pointed out to Kevin, the numerator better reflects these changes in monetary policy. Another way of saying this, your viewing the crisis from an impaired balance sheet perspective while I am see the impaired balance sheets being endogenous to the stance of monetary policy (See the link I gave Kevin to the my review of Mian and Sufi's book for more on this point.)

      Finally, the numerator has not been relatively stable. It has been growing over this period. Given the sizable collapse in household portolios and incomes that occurred, this growth can only be happening if risk aversion is high.

  6. David,
    I think we could debate several points. On innovation, I'm hard pressed to think of liquidity enhancing products that were not already widespread by 1997. A list: credit cards, money market and brokerage a/c's with checking, automatic overdraft lines, 401k loans... The new innovation post-'00 was essentially the 100% LTV HELOC. There was nothing "technological" about this development. Rather, it reflected a new set of expectations about mortgage delinquency rates. These expectations, by virtue of being "new", are not necessarily "permanent", and can be found to be, in hindsight, wrong. This is what happened. I think your response is that they were not so much "wrong" as dispelled due to a policy mistake.

    Which gets me to your endogeneity argument. It seems to bake in any positive expectation (i.e. '00 internet stock prices; '04 beliefs about home prices) while tagging moves away from those as evidence of policy failure. This seems highly asymmetric, and overall excessively dependent on the idea that progress follows a steady trend that need only be reinforced by the central bank.

    Perhaps more important, though, is the circularity of the endogeneity argument. To prove that liquidity balances were caused by a policy failure, it seems necessary for you to argue that policy determines liquidity balances. Relax this assumption (as you are trying to prove it), and it is much harder to argue that '00 liquidity balances were "normal".