Deleveraging implies slow growth in total credit, and according to the usual reasoning, slow growth in GDP. Several of Deutsche Bank’s economists, however, focus on what they call the credit impulse. They provide the following provocative graph, which suggests a rapid recovery:
Figure 8 from Deutsche Bank, Global Economic Perspectives, Sep. 16, 2009 (not online).
From the same document:
“…The pace of deleveraging reached a post-WW2 record in Q1 and, as we have argued in a number of notes on the credit impulse, all that is required for demand to rebound is for the pace of de-leveraging to slow. Even if the pace of de-leveraging
remains constant, the credit impulse will improve dramatically
through the remainder of the year (Figure 6). Private sector domestic demand should follow suit.”
The gray diamonds are forecasts for the “credit impulse” variable. The trajectory of this variable, and the obvious correlation with C+I suggests that there will be a substantial rebound in the two private components of aggregate demand. This quarterly credit impulse variable is the first difference of the change in credit normalized by GDP. Michael Biggs, in “The Impact of Credit on Growth,” Global Macro Issues 19 Nov 2008, describes the variables thus:
“The conventional measure used when associating developments in credit with developments in domestic demand is credit growth. This is growth in the stock of credit. In our view, domestic demand depends on the amount borrowed in a particular period, or the flow of credit. If this is correct, then growth in domestic demand is a function of growth in new credit issued, and not growth in credit. The impact of credit on demand will also depend on the amount of credit extended relative to the size of the economy, and consequently our preferred credit measure is the change in new credit issued as a % of GDP. For the remainder of this piece we refer to this measure as the ‘credit impulse'”.
Real credit growth and the Deutsche Bank “credit impulse” series are compared in this Figure 1:
Figure 1 from Michael Biggs, “The Impact of Credit on Growth,” Global Macro Issues 19 Nov 2008 (not online).
The theoretical motivation and empirical backing for this approach are contained in this working paper by Michael Biggs, Thomas Mayer and Andreas Pick. I’ll observe that this perspective differs from that I’ve stressed in my posts: [1], [2], [3]
To a certain extent, the focus on the new amount of lending hearkens back to the controversy spurred by Chari, Christiano and Kehoe (2008) (hereafter CCK). Readers will recall that the Minnesota Fed economists asserted, after looking at the aggregate stock of lending, “… the first claim, that banks have essentially stopped lending to nonbank entities and individuals, is false, at least in the aggregate as of October 15.”
The best rejoinder is by Boston Fed economists Ethan Cohen-Cole (a UW Ph.D!), Burcu Duygan-Bump, Jose Fillat, and Judit Montoriol-Garriga “Looking Behind the Aggregates: A reply to ‘Facts and Myths about the Financial Crisis of 2008′”.
In their rejoinder to the first CCK “myth”: “Bank lending to nonfinancial corporations and individuals has declined sharply,” they write:
“CCK show a number of graphs using Federal Reserve data that plot total loans from bank balance sheets over time. They conclude from these charts, which show an increasing trend over time, that there has not been an impact on lending to non-financial firms or consumers during the crisis. We agree that the aggregate patterns show no evidence of a decline. However, these aggregates hide the underlying dynamics and belie the fact that the components of this data show a sharp weakening of credit conditions.”
During the current credit crisis, banks balance sheets have expanded for a number of reasons. First, loan “securitization,” the business of packaging various loans (home, business, auto and other) into assets for investors, has become more difficult. Accordingly, banks have had to keep the loans they make on their balance sheets. Figures 1A–1C present strong evidence on the declining issuance in securitization markets (ABS, agency MBS and non-agency MBS). Figure 1D shows the impact of the securitization break-down on three selected banks’ balance sheets. In Panel A (Panel B), we plot bank specific data on ABS (MBS) issuance and bank balance sheet ABS (MBS) holdings classified as Available-For-Sale and Held-to-Maturity. Focusing on JPMC, for instance, it can clearly be observed that, starting in the fourth quarter of 2007, the value of ABS available for sale grew exponentially reaching about $30 bil. at the end of September 2008. At the same time, the total value of the deals underwritten by JPMC declined sharply. The same pattern can be observed for private-label MBS. Call Reports classify MBS in banks balance sheets into two categories: Pass-through securities (PT MBS) and Other MBS (O MBS) that include CMOs, REMICs, and stripped MBS. The former class is the most straight forward class of MBS, while the latter corresponds to more structured products, for which valuation models are more complex. Looking at Panel B, and again focusing on JPMC, we observe that in the third quarter of 2007 the value of Other MBS available-for-sale started to pick-up. Three quarters later, around mid-2008, the value of PT MBS available-for-sale on JPMC’s balance sheet also started to increase. At the same time, the number of non-agency MBS underwritten by JPMC declined markedly, being below $4 bil. at the end of 2007. Overall, these figures suggest that the break-down of the securitization market contributed to the ballooning of bank’s balance sheets.
Second, during this and other times of financial weakening, companies increasingly rely on their existing loan commitments and lines of credit. This is because general liquidity dries up and commercial paper markets become strained. As a result, the aggregate figures in CCK do not reveal the weakening in new lending. In fact, CCK acknowledge this point in the revised version of their working paper and highlight the fact that they would like to see data supporting this argument. In Figure 1E, we plot the ratio of total outstanding to total commitments in syndicated loans and show a significant increase in draw-downs from lines of credit, comparable to the levels during the Savings and Loan crisis in the early 1990s. Similarly, in Figures 1F and 1G, we plot data based on publicly available call reports and show that unused lending commitments at commercial banks, especially for commercial and industrial loans, have contracted since the last quarter of 2007.”
Here is Figure 1F from the paper:
So, the case seems to be made for lending contracting at the end of 2008 (not sure too many people dispute this, but at least CCK did). The question remains open whether the credit impulse captures the amount of new lending, and then whether the new lending will manifest itself in rapid growth in consumption and investment.
Current economic indicators do indeed suggest that an upswing in GDP is in progress — for now.
Updated (6pm) Figure 1: Real GDP in billions Ch.2005$, SAAR (blue bars), Macroeconomic Advisers 9/17 release (green), and e-forecasting 9/16 release (red). NBER defined recession dates shaded gray, assuming end occurs at 2009M06. Source: BEA 2009Q2 2nd release, Macroeconomic Advisers, e-forecasting, NBER.
What seems to be left out of the argument entirely are the capacity and demand for new loans.
Among households, most of the demand is from those without the capacity to take on new debt. The rest are still deleveraging. Here is a story from Bloomberg on consumer intentions:
“Only 8 percent of U.S. adults plan to increase household spending, almost one-third will spend less, and 58 percent expect to stay the course, a Bloomberg News poll showed. More than 3 in 4 said they reduced spending in the past year. ”
http://www.bloomberg.com/apps/news?pid=20601087&sid=aDhCOltOR1RY
Among businesses, how much new investment is likely when capacity utilization is still below 70%?
There are still about 1 million vacant residential units and commercial real estate investment is just beginning what is likely to be a multi-year downturn.
The jump in GDP called for is a result of expected changes in inventories and temporary stimulus effects, like the cash-for-clunkers nonsense. But at extremely low capacity levels, why would this spur investment?
Sounds true as their model focuses on dynamics.
Seems like “credit impulse” is a useful concept, but it is “the change in new credit growth” and the graph shows it rebounding from record lows, but it is still a very negative value. So wouldn’t that mean new credit is still contracting?
I see they project it to go up like the right side of a V. But I think we need to wait a while for the data to come in, and then we can safely say that this is a leading indicator for GDP growth, provided they can collect their data faster than the GDP people collect theirs.
But I have some reservations that this really happens. Deleveraging is occurring two ways, paying down debt as supported by recent consumer saving rate data, and the other ways…default, foreclosures and bankruptcies. Banks would need a different borrowing pool to go to (I hope) for new customers.
Money markets have been guaranteed and loosened up quite a bit, so large business may have better access there again and not need (expensive)bank credit facilities as much. Corp bond issuance has been going strong post green shoots as well.
So that leaves the other traditional banking customers, small business and commercial real estate. These do not yet sound like attractive areas for banks to make new loans to.
But if corporations are flushed with newly-US-minted, speculative cash from the stock run-up, then they will, after cashing out some of their stock options, entertain some risky, fundamentally unsound investments.
Private credit is contracting, but TOTAL credit is still growing at a decent clip, especially when normalized by GDP.
Its not clear which one (private or total) Deutsche Bank is using for its credit impulse. It should be using total credit, in my view. The latest Fed Flow of Funds report has total credit growing by 4.9% at a seasonally adjusted annualized rate for 2q. This is not de-levering. Debt to gdp grew at a stronger rate in q2 given the decline in gdp.
So what is Deutsche Bank’s point? That a decline in the rate of decline of private credit is positive? Only if public credit growth stays constant, and that is certainly not the plan.
The other thing the projection for new credit doesn’t address is where do banks really get their money. The way it did work for the majority of banks is loans go straight thru to securitization markets like poop thru a goose. The track record on all these financial products is so poor we should just nickname them all “youscrewedization”. Many think investors have satisfied their appetite for these products for a very, very long time.
So banks may need to sell a lot more stock than they have already. Good luck with that.
David Pearson
Total credit is growing because USG debt growth is more than making up for the decline in private credit.
The DB chart shows credit shrinking(negative still) so that has to be private credit.
Whether total debt should be used depends on whether you believe the banks should buy treasuries or not.
I think DB’s point is it’s a 2nd derivitive type of number with powerful predictive qualities.
But I stick by my prior reservations.
Let me see if I understand this:
“Credit” is the total amount that someone owes someone else at a given moment in time. “Growth of credit” is its first derivative. “Credit impulse” is its second. The Deutsche Bank claim is that the first derivative of demand (thus, C+I contribution to growth of GDP) is proportional to the second derivative of credit (credit impulse). Sounds reasonable. We need an integration constant to get back a relationship between credit growth and GDP level, which is presumably inflation (insofar as credit acts like money, and expanding credit/money supply could coincide with flat GDP via inflation).
They forecast this “credit impulse” variable to return to zero within a few quarters. That’s better than the alternative, but it’s a forecast for an end to the decline in GDP, rather than for actual growth.
[Side question: Is normalization to GDP done at the first stage, total credit, or at one of the derivatives? The english sentence is ambiguous, e.g.:
“This quarterly credit impulse variable is the first difference of the change in (credit normalized by GDP).”
“This quarterly credit impulse variable is the first difference of the ((change in credit) normalized by GDP).”
“This quarterly credit impulse variable is the (first difference of the change in credit) normalized by GDP.”
]
Meanwhile, CCK are claiming that “credit growth” has remained positive based on bank balance sheets. The Boston Fed authors argue that bank balance sheets are a flawed measure of total credit, for reasons that I agree with (they exclude securitized credit and credit commitments; I would add that previously off-balance-sheet SIVs have been brought on-balance-sheet, further distorting the relationship between bank balance sheets and “total credit”).
So, if bank balance sheets are not a good measure of “total credit”, then they’re not a good source for estimating the derivative (credit growth), and it becomes increasingly foolish to try to compute higher derivatives (credit impulse) from them.
So I can’t evaluate whether the Deutsche Bank forecast is credible without knowing what raw data they started from — bank balance sheets or something else? — to estimate total credit and/or credit growth.
lilnev:
“They forecast this ‘credit impulse’ variable to return to zero within a few quarters. That’s better than the alternative, but it’s a forecast for an end to the decline in GDP, rather than for actual growth.”
Menzie: “They provide the following provocative graph, which suggests a rapid recovery:”
Agree with lilnev – to me it looks like the DB economists are just predicting that consumption plus investment will stop shrinking, not that they will start growing. This is certainly not surprising. Am I misreading the post?
Reminds me of a cash-in-advance constraint.
Cannot comment on the precision of measuring rates of change in credit flows but the general idea has some intuitive appeal.
don: Note that the two vertical axes have different scales. If C+I growth follows the credit impulse, then C+I growth will be going positive.
General Clarification. The credit impulse is based on the first difference of flow series (normalized by GDP) from the Fed Flow of Funds (i.e., F tables). The change in credit stock is based on the level of the credit stocks (i.e., L tables) (divided by GDP deflator).
Some responses to these comments
David Pearson – We only look at private credit. We separate private and public because we think the private credit impulse can turn positive in 2010, whereas the fiscal impulse will only turn negative in 2011. If this is correct, 2010 could provide a very favourable environment for a rebound in growth.
Cedric Regula – you are absolutely correct in your first comment (08.54am). The credit impulse can rebound even when new credit is contracting, so long as it is contracting at a slower pace. In our view this is the power of the argument – we can see a rebound in demand and a paying down of debt at the same time. For the world to stay this bad, the private sector has to keep deleveraging at an accelerating pace. This is possible, so your concens are valid. But deleveraging at a slower pace, which is all that is required for a recovery, is a lot easier to do than re-leveraging
lilnev – we use the flow of funds statement to estimate the credit impulse, so it should capture all types of lending (and not just bank lending). And for the credit impulse to return to zero (which as Menzie correctly points out is consistent with positive growth) all that is required is for the pace of deleveraging to stabilise. That is all we assume to to get those “diamonds”. New borrowing does not have to return to zero. With regards your ambiguity question, option 2 is correct (credit impulse = d(d(C)/GDP))
Many thanks for all these very interesting comments
Mike Biggs: Am glancing at the working paper and am wondering how you derived r = A – delta. What’s the implicit assumption? What am I missing?
Thanks.
My other thought is that I am completely unconvinced by post-WWII relationships used to forecast from our present position – the post-war experience does not include a like case of over-indebted consumers. In particular, unlike the recent experience, we may be more credit-demand constrained and less credit-supply constrained.
Don – I think your concerns about post-WWII relationships are appropriate. Unfortunately, the pre-WWII experience won’t allay your fears either. The credit impulse versus C+I growth chart holds wonderfully well all the way back to 1928 (the extent of the data we could find). However, the build up in debt before the great depression appears to have been more of a corporate debt than a household debt problem.
As I have stated before recovery will not come from graphs and equations. Recovery will come from US producers beginning once again to produce. Right now the business climate in the US is very suspects. Here is something that was sent to me by a friend giving his perspective.
By way of personal example, I just finished a budget process in a manufacturing company and we project that there will be no opportunity to raise prices in 2010 and we have niche products. I suspect more commoditized manufacturers will be facing price roll backs in 2010. We expect that sales will increae only 5% over 2009 levels and remain 20% lower than 2008 levels. We are looking for all increase to come from export and foreign activity. We do not project any increase in domestic wages except that we have decided to hire up the skill curve and pay a higher wage for a more skilled employee (more productive and lower training management cost)…becuase they are available in the current market where they have not been available before. We do not plan to increase domestic employees on net basis (domestic employee count is down 20% from 2008) and we believe that the step up in pay for skills is still a disount to what those skills would have cost last year or the year before. We do plan to achieve slightly better gross magins on lower volume of sales than we achieved in 2008 but we will produce less gross profit because of the lower sales volumes. If the economy picks up faster than we expect we are likely to add employee and capital investment capacity abroad before we do so domestically.
My friend’s last point is critical to discussions of unemployment. While some firms may indicate that they will increase some hiring next year they do not differentiate whether that hiring will be foreign or domestic. I believe that my friend is typical in that he sees growth and profit prospects greater overseas than in the US. This is a sad state of affairs for a nation that has been the world’s economic engine for so long.
“Note that the two vertical axes have different scales. If C+I growth follows the credit impulse, then C+I growth will be going positive.”
The chart shows credit impulse today with much wider distribution curve (higher standard deviation) than in the 70’s relative to C+I. So even if credit impulse moves back up to top diamond (+4ish), it looks more likely C+I would not swing far enough to turn positive (above the zero on right hand scale).
I can agree with understanding GDP growth as a function of the change in new debt (or the second derivative of total debt).
But in what exactly is the author’s expectation founded that the credit impulse will become positive in 2010? Why won’t we see an acceleration in negative debt change instead? Private debt in US has reached 41 trillion US-dollars, or almost 300% of US GDP. According to my coarse estimates, this mountain of private debt requires an amount of total interest payments alone so that no GDP growth to be reasonably expected, can generate enough additional income from which the interest can be paid. That is, net debtors won’t be able to just comfortably pay off debt at these debt levels and increase demand at the same time.
And if the concept is right, the total amount of stimulus by the USG doesn’t matter as much as the change in stimulus from one to the next period. The total projected increase in stimulus from 2009 to 2010 is only 48 billion US-dollars. In 2011 it will decrease by 164 billion dollars.[1]
[1] http://www.whitehouse.gov/assets/documents/CEA_ARRA_Report_Final.pdf, Page 1, footnote 1
rc
As for the first figure. The projections shown as diamonds are for Q2 2009 to Q4 2009, aren’t they? Well, looking at the Fed’s Flow of Funds Account, which has just been released on Sep 17, it doesn’t look like that credit impuls from private credit improved from Q1 to Q2 2009. It seems to have worsened instead, although the Flow of Funds Account shows seasonally adjusted annual rates, whereas the first figure shows y-o-y changes. Anyway, I really doubt that the projections are well founded.
rc
rootless cosmopolitan: You’ll notice in the graph the impulse is defined using y-o-y growth rates.
For a good summary of the latest Flow of Funds just released go here and click on the link “U.S. Flow of Funds Shows Improved Financial Condition of Households, But Dramatic Volumes of Treasury Borrowing”. Says who and who isn’t doing borrowing.
http://www.haver.com/
For a good summary of the latest Flow of Funds just released go here and click on the link “U.S. Flow of Funds Shows Improved Financial Condition of Households, But Dramatic Volumes of Treasury Borrowing”. Says who and who isn’t doing borrowing.
http://www.haver.com/
Rootless cosmopolitan – Our big point is that GDP growth should be viewed as functions of the change in new debt. Given that you accept that, we are largely in agreement here. This is very different to the consensus view captured in Menzie’s first line “Deleveraging implies slow growth in total credit, and according to the usual reasoning, slow growth in GDP.”
This brings us to the much tougher issues of whether deleveraging will actually accelerate of decelerate. As a point, though, note that the q2 flow of funds statement is consistent with an improvement in the yoy credit impulse. The amount of new borrowing in Q2 2009 was less than in Q2 2008, so the credit impulse is negative. However, the amount of new borrowing fell by USD500bn in Q2 2008 from Q1 2008, whereas it “only” fell by USD250bn in Q2 2009. The amount of new borrowing is falling, so the impulse is still negative. However, it is falling at a slower pace, so the impulse is rising. We made the chart before the Q2 flow of funds statement came out, but it doesn’t change much. If the pace of deleveraging stabilises, those diamonds keep rising.
This deleveraging includes all interest payments, so households were managing to deleverage from Q4 – Q2 despite the interest burden (although they might struggle if the Fed were to hike rates). So, if the pace of deleveraging stabilises through 2009 and then starts to slow through 2010, debt levels continue to fall. And, if the credit impulse gives rise to positive nominal GDP growth, the debt levels fall more sharply as a function of GDP. This is perhaps the power of the argument – deleveraging and balance sheet repair (via falling debt levels) can happen even as the credit impulse turns positive and domestic demand rebounds.
As you point out, the pace of deleveraging could accelerate from here. In this case, the credit impulse would stay negative and growth would be negative or at best very weak. We think people might be less pessimistic now than they were six months ago and consequently be willing to risk a stable or slower pace of deleveraging, and banks might also be under less pressure to deleverage in the panicked fashing seen in Q4 and Q1. The credit conditions indices in the loan officers surveys have improved a lot. However, we cannot rule out your more negative scenario.
Finally, I agree with your reading on fiscal policy – a narrowing in the budget deficit in 2011 means a negative fiscal impulse, which is bad for growth. Our hope is that there is a sweetspot in 2010 where the credit impulse turns positive, but the fiscal impulse is yet to turn negative.
To this layman, these kind of data are why economics fails so bad on macro.
In the first graph, the blue line tracks the credit impulse nearly without delay on any large movements, and seems to filter out the high frequency stuff.
Right now, the blue line is substantially delayed.
The blue line always seems to meet the local minima of credit impulse, but a reversal tomorrow would be at nearly 6% from the minima.
To me, this tells me that something is seriously broken. They did track, and now they don’t.
You might find out why. It could lead to a clue in the future when looking for signs of trouble.
Does anyone else see these two data as decoupled right now?