From Chapter 1 of the IMF’s recent World Economic Outlook (Box 1.2), a set of findings by Jörg Decressin and Marco Terrones:
The econometric results confirm that net capital inflows, financial sector reform, and total factor productivity are good predictors of a credit boom. Net capital inflows appear to have an important predictive edge over the other two factors.
The Econometric Results
These results are reported in a table of estimates obtained from logit regressions over a sample up to 2010:
Table 1.2.1 from WEO Decressin and Terrones in Box 1.2.
Statistical significance is one issue. Another question is how much improvement in getting right signals does one obtain by adding financial reform and productivity to a basic specification using only net capital inflows? That is addressed in this depiction of the ROC curves for each model (although only for sample up to 2007):
Figure from WEO Decressin and Terrones in Box 1.2.
Receiver operating characteristics (ROC) curves were discussed in this post.
There is a slight increase in predictive power, but the graph validates the conclusions made by the authors — most of the predictive power is in net capital flows.
Foreign Factors
How much do international factors matter? The authors include additional variables to account for foreign effects.
To explore the possibility that net capital inflows are capturing the effects of easy international financial conditions on domestic credit booms, we include in the regression analysis proxies for return (the real interest rate) and volatility (Chicago Board Options Exchange Market Volatility Index) in the United States. Although these variables have the expected signs, they are not statistically significant (Table 1.2.1, columns 7 and 8). Moreover, when included with net capital inflows, the predictive power of the volatility variable remains broadly unchanged (Table 1.2.1, columns 9 and 10).
[Addendum: Slightly different findings are obtained by Jorda, Schularick and Taylor (IMF Economic Review, 2011), who find a primal role for credit growth, with greater importance for external imbalances in the pre-War era relative to the post-War.]
I think this is a particularly interesting set of findings; credit boom/busts are closely associated with financial crises. As an international finance economist, my view had been that net capital inflows were particularly dangerous for the United States in that they resulted in increasing net indebtedness to foreigners that would eventually lead to foreigners dumping dollar denominated assets (see Chinn (Council on Foreign Relations, 2005)). A competing view was that the net capital inflows signaled financial excess (of course, the two are not mutually exclusive — both forces could be in effect, but with different strength). The findings reported here are more consistent with the second view. (There is of course yet another view, that all that borrowing was justified by the high productivity and entrepreneurial energies unleashed by the tax cuts of 2001 and 2003); ‘nuff said of that.)
While net capital flows show up, it’s important in my view not to interpret the inflow as necessarily being “pushed” in from abroad (China, oil exporters in the years up to 2008), but also “pulled in” by policies, perhaps involving loosening of constraints on the financial sector. So, I don’t adhere to the Bush Administration’s “Blame it on Beijing” view as propounded in the 2009 Economic Report of the President.[1]
Several policy implications flow from these findings:
… Given the high costs of credit
boom-bust cycles, policymakers should closely monitor
the joint behavior of capital inflows and domestic lending.
6 There is also evidence that financial sector reforms
are predictors of credit boom-busts. Policymakers
must ensure that financial liberalization programs
are designed to strengthen financial stability frameworks.
Last, there is evidence that large productivity
gains increase the risk of a credit boom, particularly
in advanced economies, driven perhaps by exuberant
optimism in new sectors. Thus, even during particularly
good periods for the economy, policymakers must
be on the lookout for emerging threats to financial
stability stemming from credit booms.
I think this warning should be kept in mind when three years after the largest financial crisis in history, we see attempts to gut effective regulation of the U.S. financial system. From Moneyline.com:
Republicans will are likely to up attacks on the Dodd-Frank legislation that increases regulation on the financial sector as campaign season heats up, the New York Times reports.
Dodd-Frank aims to curb abusive lending practices, stop high-risk bets on complex derivative securities and protect consumers from financial fraud, among other goals.
Many of the bill’s provisions haven’t gone into effect, which will give Republicans new material to point out how President Barack Obama’s efforts to increase regulation will continue to hamper economic recovery.
“It created such uncertainty that the bankers, instead of making loans, pulled back,” says Republican presidential hopeful Mitt Romney, speaking at a South Carolina rally over Labor Day weekend where he again called for the law’s repeal, the New York Times reports.
The bill does have its supporters, some of whom in turn admit the law may hamper job demand.
“Dodd-Frank is adding safety margins to the banking system,” says Douglas J. Elliott, an economic studies fellow at the Brookings Institution.
“That may mean somewhat fewer jobs in normal years, in exchange for the benefit of avoiding something like what we just went through in the financial crisis, which was an immense job killer.”
A replay of financial boom and crash is what we risk if we let the agents the financiers have bought off to sabotage regulating the financial system (as discussed in Lost Decades).
Senator Dodd is mercifully retired. So who now is paying off Representative Frank, please? And Mr. Johnson can’t raid the taxpayers via Fannie Mae twice. Are there others we should watch out for?
Anyone else remember the red-lining who-ha? That’s where it started, guys. Good old liberal snivel. Always the best and always the most expensive down the road.
C Thomson: Thanks for the erudite analysis of the empirics.
Menzie,
How many jobs do you expect Dodd-Frank to create versus how many will it kill? (I expect those of us who are in finance, accounting, or the legal profession will see an increase in jobs.)
Lagged net capital inflows = easy-money-fueled trade imbalance.
Lagged financial sector reform: repeal of Glass-Steagall last time, Dodd-Frank making Too Big to Fail permanent this time.
Lagged U.S. real interest rate = Fed easy money policy.
So there we have it, a perfect indictment of those who caused the financial crisis: the Dirty Fed and the crooks in Congress.
And while we now know what predicts a financial crisis, let’s not forget who doesn’t predict a financial crisis. That would be the idiots at the Fed.
W.C. Varones: By convention, in reporting statistical results, we usually use the 10%, 5%, 1% marginal significance levels, denoted by some symbol. You will notice that the real interest rate coefficients do not exhibit any statistical significance, at the conventional levels.
Since the specification is a reduced form one, we don’t know what causes the net capital inflow. But if you want to blame the folks who pushed the Commodity Futures Modernization Act (like Phil Gramm), and the policy-level folks put in place by the Bush Administration at the Office of Thrift Supervision [a], and the policy level appointees at the SEC who in 2004 allowed the investment banks to leverage up well, please do.
Menzie,
Could this capital inflow/credit cycle relationship be explained by the relationship between net exports and the business cycle and credit cycles? Near the peak of an expansion, net exports are higher, so foreigners have more dollars to invest in U.S. treasuries. The extra captial inflow keeps the risk free rate lower than it ‘should’ be, causing lenders to make more bad loans, marking the peak of the credit cycle. Or if you are working with 5 yr trends, the trends will be a bit higher during expansion. The opposite is true during downturns.
Dodd Frank goes too far in some areas and not far enough in others. Menzie, you know full well that derivatives reduce risk when used as a hedge. One of the problems with Dodd Frank is that it does not differentiate between hedgers and speculators and speculators in some cases. The result is that some business owners, like farmers, will face more risk because of Dodd Frank. I applaud the effort to regulate systemic risk, but Dodd Frank needs an overhaul.
Menzie
I’m not sure that I quite buy into the binary way in which credit booms have to be defined in a logit or probit model. Not all credit booms are created equal and many might well be unobservable or censored. I’m wondering if a Tobit approach that assumed certain low levels of credit booms were essentially present but unobservable (i.e., = 0) might not be better because it would allow for different gradations of credit-boomness (new word). It’s just that the 0/1 formulation seems too abritrary. Not everyone agrees on what was or was not a credit boom episode.
“I think this is a particularly interesting set of findings; credit boom/busts are closely associated with financial crises. As an international finance economist, my view had been that net capital inflows were particularly dangerous for the United States in that they resulted in increasing net indebtedness to foreigners that would eventually lead to foreigners dumping dollar denominated assets (see Chinn (Council on Foreign Relations, 2005)). A competing view was that the net capital inflows signaled financial excess (of course, the two are not mutually exclusive — both forces could be in effect, but with different strength). The findings reported here are more consistent with the second view. (There is of course yet another view, that all that borrowing was justified by the high productivity and entrepreneurial energies unleashed by the tax cuts of 2001 and 2003); ‘nuff said of that.)”
Too much macroeconomics and not enough microeconomics and budgeting. Next, replace financial crises with too much debt crises. Lastly, it needs to be explained why more and more debt (whether private or gov’t) that does not produce price inflation means there are actually problems in an economy instead of being hailed as some kind of grand achievement.
I suggest watching the video from W.C. Varones. Why would anybody hire such a fool or someone who is only worried about his spoiled, rich banking buddies? I wouldn’t hire anyone that wrong and that can’t do personal finance at any wage rate.
“You will notice that the real interest rate coefficients do not exhibit any statistical significance, at the conventional levels.”
How about looking at the real interest rate in terms of prices of assets instead of prices of goods/services?
The captioned model is capturing the essence of an international financial crisis,capital inflows when capital outflows are equally devastating.The former nurtures the future crisis when the later testifies for it.The IMF falls short of questioning implicit variables.
Axiomatic no questions are asked on the manufacturing at little cost of volatility (Econbrowser ,the last Changing behaviour of crude oil futures prices),the same goes for interest rates yield curves,the same goes for the placebos of financial alert such as Value At Risks.
Caetaris paribus is no longer valid,supply and demand curves belong to the nitendo economies.Should the financial sector be properly regulated above few variables and more may be relevant in models.Until then models will be the protocol of an autopsy.
If I understood what the authors wrote, it was that credit inflows were ostrngly predictive for emerging economies and were the least predictive for advanced countries. Significant productivity gains and large financial sector changes were the strongest predictors for advanced countries, with productivity gains being the most prominent predictor.
This is clearly stated at the end of Box 1 and shown in the second bar graph. The conclusions that Prof. Chinn draws about the United States appear to be from taking the yellow bar, which is the sum of emerging and advanced countries, not the blue bar for advanced countries.
When I analyse trading methodology systems the one thing I look for as a huge flaw is the author “curve-fitting” his/her Results. Though not a trading methodology, Jorg and Marco have most certainly “curve-fitted” their results.
I enjoyed reading this right up until the quote from the New York Times, which set off both my weasel-word detector and partisan-spin alarm.
“Vague promises about popular reforms, also other things best not mentioned” is not a promising format for an article.
Yet again, the occasional snide partisan remarks make it difficult for me to take your analysis at face value, which means I am going to waste a lot of time googling responses to Jörg Decressin and Marco Terrones.
Menzie,
Why do you always hold the Bush administration up as justification for doing things wrong?
Secular credit cycles peak and collapse at the point at which the cumulative differential rate of growth of debt-money to wages and production reach an order of exponential magnitude (“Jubilee” threshold) from an earlier secular low, typically culminating in a faster-than-exponential blow-off phase lasting 3-5 years historically.
In effect, returns from labor and production can no longer grow sufficiently to service existing debt (today at effectively infinite term), let alone service additional debt in perpetuity.
At the secular debt-money peak in ’06-’08, US total credit market debt owed to GDP reached an unprecedented $6 of debt to $1 of GDP. China’s ratio recently surpassed $7 to $1, implying a crash in China (and by extension Asia, Russia, Brazil, Canada, and Australia) is imminent; only the precise timing is in question.
Peak Oil, falling net energy and oil exports, population overshoot, and the global Baby Boomer demographic drag effects make it “different this time” and will exacerbate the global debt-deflationary depression hereafter.
http://www.hussmanfunds.com/wmc/wmc110926.htm
Moreover, the U rate and total unemployed to private payrolls historically has risen 50-100%+ during recessions, which has likely already begun (as ECRI now “forecasts”), implying that the U rate could reach 13-14% to 18-20%, whereas total unemployed to private payrolls could rise from 15-16% today to 25-30% in the next 18-24+ months. Gov’t payroll losses and pension and benefit cuts will worsen the labor market conditions and associated spending.
Also, Boomers turning age 62-65 and leaving the labor force en masse (voluntarily and otherwise) over the next 7-10 to 12-15 years will exert an unprecedented drag effect on consumer spending, stock and real estate prices, and local, state, and federal receipts and fiscal conditions.
Real estate prices will fall another 15-20% (albeit at a slower rate) and stock and commodities prices will crash. The historical self-similar secular bear market pattern for stock prices (1830s-40s, 1890s, 1930s-40s, and Japan since ’00) implies the S&P 500 going back to the Mar. ’09 lows to as low as the 400s-500s by summer-fall ’13. Intervention by the Fed in the stock market will only increase volatility and yield the same result in the end.
The criticisms of Dodd-Frank are justified. I do not attack its stated purpose, but the language used is vague and ambiguous. It was a cut and paste act, which will have huge unitended consequences. We will not actually know what it says until the various regulators write the actual regulations. However, some of the negative effects are showing up now.
Community and small regional banks are being consolidated into the larger national banks. Even the smaller banks that survived the recent bust are giving up and being acquired. CFO Magazine reports that the costs of complying with D-F make it very expensive for local banks to stand on their own. “No matter what size a bank is, it believes it needs to bigger to digest the cost of the increased regulatory burden…” (CFO, April 2011 p.47 quoting Michael Clark president of Access National Bank). Charles Maddy CEO of Summit Community Bank testified to Congress that D-F “…will have an enormous and negative effect on my bank.” He complained that there are already over 1,000 pages of new rules and that the carve out for the reduced swipe fee for smaller banks will be ineffective to protect them from the bigger banks. He also testified that bank regulators are encouraging banks with less than $500M in assets to merge. (Id.)
Whether due to failure in the bust or due to acquisitions, the FDIC has 1600 fewer reporting member banks (down from 10,204 in 2000). Every report I have found expects increasing contraction.
I am just a lawyer, but my econ professors in Madison taught me that fewer providers in a sector is generally bad. In this case it does not take a Ph.d to figure out that fewer small banks will cause less credit, increased costs of credit and less capital available for small businesses.
As a lawyer I am sure that my self interest is served by D-F. However, as a citizen I cringe when the legislature writes in sweeping generalities–establishing goals and letting the regulators write whatever rules they think appropriate to achieve those goals. D-F may be well meaning, but it is lousy legislation.
We have seen this before. Sorbannes-Oxley was another well meaning act. It was written in response to activity which was already criminalized in other laws. The SEC itself reported that a small cap issuer would have to spend 72% of all its assets in the first year on compliance costs alone. After 4 years of compliance the company would spend only 44% of its assets on compliance. Since Sorb-Ox there have been very few small companies to go public. There have been a few public offerings, but they had already hit large cap status since only the huge can survive the compliance hit.
Some of Sorb-Ox is good (i.e. independence of auditing committee). However, most of it was written to regulate huge companies and it is an unreasonable barrier to capital markets for small businesses.
I am not against regulating the financial sector. It is in my personal financial interest to get paid for assisting businesses through the regulatory morass. However, tailoring the laws to prevent abuses is preferable to writing sweeping changes which handicap small businesses from competing with the large businesses.
The legislature did do one thing right. The Small Business Jobs Act was fairly clearly written. It streamlined the SBA’s process and allows more companies to qualify (CFO Aprill 2011 p.49).
I doubt that this will wholly offset the negative effects of D-F on small businesses, but it certainly helps. I think everyone who regularly reads this blog will agree: We need more small businesses.
IMO, models should be able to show a credit boom (I’d rather call it a debt boom) without involving the foreign sector.
Ricardo said: “Menzie,
Why do you always hold the Bush administration up as justification for doing things wrong?
It seems to me there are way too many people (including economists) caught up in what I consider to be this stupid left/right, liberal/conservative, democrat/republican argument(s).
Anybody ever consider that both sides are wrong?
colonelmoore: The bar graphs refer to a sample that ends at 2007 (hence the reference to the Mendoza-Terrones (2008) paper). I preferred to rely upon the updated results that extend to 2010 and hence include the 2008 crises in the advanced economies — i.e., the table and the ROC. Clearly, the number of advanced countries experiencing crises was much smaller in the sample extending up to 2007.
Prof. Chinn,
The bar graph is part of the box. It says, “Sources: Mendoza and Terrones (2008); and IMF staff calculations.” The staff calculations could very well include the period after 2007.
The box says “Given the strong association between such credit boom-bust cycles and financial crises, it is important to understand what drives them. This box studies credit booms in 47 economies—19 advanced and 28 emerging market economies—during 1960–2010.
You wrote: “I preferred to rely upon the updated results that extend to 2010 and hence include the 2008 crises in the advanced economies — i.e., the table and the ROC. Clearly, the number of advanced countries experiencing crises was much smaller in the sample extending up to 2007.”
The paper is about predicting credit booms using data. Where are the data in the table showing that capital inflows in advanced countries predicted credit booms after 2007?
2slugbaits, as a demographer I would add the demographic imbalance of a glut of mature households producing capital in the presence of a dearth of new households demanding capital. Too much capital is a tautology for bad investments.
JLR, thanks to Bernie and “Sir” Allen my industry has just had to deal with a legislative “cut and paste” response with rules to be worked out by the regulators. The regulators were nice enough to sit down with the industry and write the rules together, and this is how D-F will be filled out. As to the “small” banks, the housing boom was also a small bank boom. The market had turned into an evergreen money machine for entrepenuer bankers, with no chance of failure. The population of small banks is part of what has to be right-sized.
Menzie, how about a refresher course in the history of alt-a and subprime debt. As I recall, there was no subprime debt in 2001, only alt-a, and that the subprime market was an invention of our investment banks’ never ending search for yield. Also, I’m under the impression that the GSE’s were prohibited in dealing in subprime debt. Given the central role of subprime debt, it is hard to see anyone but the investment banks as the driver of the subprime mortgage crisis.
colonelmoore: Well, yes, the bar charts could refer to data up to 2010. But I consulted one of the coauthors of the box, and was conveyed to me is what I have written: table and ROC refer to data up to 2010; bar charts to 2007.
Prof. Chinn,
Very good. Thank you for doing that. Now to the meat of the matter.
In your main article you use Table 1.2.1 to claim that net capital inflows are a significant factor in predicting credit booms. When I questioned the claimed effect for advanced economies, pointing out that you may have been looking at the composite yellow bar in Figure 1.2.1 rather than the blue bar for advanced countries, you said: I preferred to rely upon the updated results that extend to 2010 and hence include the 2008 crises in the advanced economies — i.e., the table and the ROC. Clearly, the number of advanced countries experiencing crises was much smaller in the sample extending up to 2007.
The authors state that the study is about credit booms from 1960 to 2010 and that there were 19 advanced countries in the study. They did not omit any credit booms based on whether crises had taken place in those countries.
Any credit boom in advanced countries surely ended with the collapse of Lehman. So at most your statement above would have to refer to advanced country credit booms that began in the period January – August 2008. I am unaware of any such country booms.
Is there something in Table 1.2.1 that backs up the notion that net capital inflows became significant triggers for advanced-country credit booms after 2007?
That be me.
colonelmoore: Whether a credit boom occurs depends on the detrending technique which depends on the sample…
While the bar charts indicate greater import for financial dereg and productivity, according to table 6 of Mendoza and Terrones, it’s not clear whether there is a statistical significance to the difference; certainly the difference between 0.27 and 0.33 doesn’t seem big economically.
So I relied upon the ROC which gives the statistics for all countries, and is up to date. While the productivity trigger is more important in industrial countries, that doesn’t deny the characterization that overall, in terms of right prediction, capital flows are key.
Prof. Chinn,
I’ll admit that possibly I misunderstand the authors when they say “In particular, credit booms in emerging market economies seem to be associated mostly with large capital inflows, whereas those in advanced economies often coincide with productivity gains (Figure 1.2.1, middle panel).”
However I received additional information in an email from one of the coauthors that his analysis of the most recent data up through 2010 does not indicate any material changes in the bar graphs. So it seems that we need only consider the blue bars, which clearly show far less predictive power for capital inflows to advanced countries.
As for the author’s statements that you quoted about regulatory watchfulness, I could not agree more. Greenspan’s notion that the rules about inflation had changed so that he could keep interest rates low, and that risk modeling would allow banks to hold less capital for off balance sheet RMBS and no capital if they lent to sovereigns definitely caused a credit boom. Then too late he raised interest rates, hoping to land softly. The increased cost of lending turned the leveraged investments into money losers.
colonelmoore: If the coauthor states that the bar charts don’t change, that must be true. My point is (i) whether the differences are statistically significant cannot be determined from the charts (or the table in the paper) and (ii) the bar charts notwithstanding, for the total sample, the ROC chart indicates that the incremental gain from adding the other variables to capital flows is small. For those of us who have had experience in the prediction/early warning system business, I put more weight on that finding.
Prof. Chinn,
Somehow this got turned around into you challenging me on the importance of productivity increases as a predictor of credit booms. So let’s agree that, due to statistical insignificance, one cannot make any assertions as to the relative importance of productivity in predicting credit booms.
The main issue is whether one can deduce from the aggregated data for all countries that large capital inflows are a signifant predictor of credit booms in advanced countries.
Let’s suppose that Decressin and Terrones wrote two different papers using the same exact data and analytical methods. Decressin’s paper only dealt with the data from advanced economies and Terrones’ paper only dealt with emerging economies. Would you be able to examine the data in Decressin’s paper alone and draw the same conclusions as you stated above?
Prof. Chinn,
Here is an excerpt from page 3 of the paper you referred to above, “AN ANATOMY OF CREDIT BOOMS: EVIDENCE FROM MACRO AGGREGATES AND MICRO DATA, Enrique G. Mendoza and Marco E. Terrones”
http://www.federalreserve.gov/PUBS/ifdp/2008/936/ifdp936.pdf
Third, the frequency of credit booms in emerging markets is higher when preceded by periods of large capital inflows but not when preceded by domestic financial reforms or gains in total factor productivity (TFP), while industrial countries show the opposite pattern (credit booms are more frequent after periods of high TFP or financial reforms, and less frequent after large capital inflows).
In your main article you stated, “overall, in terms of right prediction, capital flows are key.” I was saying that the overall prediction could not be applied to the specific case of the United States, which is a developed economy. Mendoza and Terrones would seem to agree.
Further support for this would seem to be found i Table 6 of Mendoza-Terrones. I have reproduced the part of the table for industrial countries.
Table 6. Credit Booms: Potential Triggering Factors
(Frequency distribution)
0.27 Large Capital Inflows
0.40 Significant Productivity Gains
0.33 Large Financial Sector Changes
0.07 Large Capital Inflows & Significant Productivity Gains
0.00 Large Capital Inflows & Large Financial Sector Changes
0.07 Significant Productivity Gains & Large Financial Sector Changes
0.07 Large Capital Inflows & Significant Productivity Gains & Large Financial Sector Changes
You stated: the bar charts notwithstanding, for the total sample, the ROC chart indicates that the incremental gain from adding the other variables to capital flows is small.
This would actually apply to any of the three variables. But I don’t know why one would need to use ROC to show that the incremental gain from adding variables together when Table 6 shows the frequency distribution of any two factors or all three factors together. This ranges from 0.0 to 0.7. For advanced economies, adding Large Capital Inflows as a predictive factor to Significant Productivity Gains & Large Financial Sector Changes adds little predictive power. So there seems no way to say, “ most of the predictive power is in net capital flows. ” when referring to the United States and other advanced economies. Therefore whatever inferences that you drew on this score in the main article and the comments would appear overdrawn.
I have a small issue with another statement you made. You stated: While the bar charts indicate greater import for financial dereg and productivity, according to table 6 of Mendoza and Terrones, it’s not clear whether there is a statistical significance to the difference;
I am unconcerned with the differences between productivity and deregulation, but I wanted to point something out. You used the numbers 0.27 for productivity and 0.33 for deregulation. These are the numbers in Table 6 for all countries. The left side of the same table is for industrial countries and it lists the numbers 0.40 for Significant Productivity Gains and 0.33 Large Financial Sector Changes. I assume that these are statistically significant differences.
(Note: I see some issues with this paper that might not make it the best reference for this discussion. But the coauthor of the IMF paper said that its conclusions would not change after 2010, so that is good enough for me.)
colonelmoore: From the box:
Prof. Chinn,
I wanted to take a break from this and do some business. But getting back to things, I see that you are particularly adamant about referring to the aggregated data.
I asked a question above that is particularly relevant now that you have posted the answer above. Do you want to answer it?
Let’s suppose that Decressin and Terrones wrote two different papers using the same exact data and analytical methods. Decressin’s paper only dealt with the data from advanced economies and Terrones’ paper only dealt with emerging economies. Would you be able to examine the data in Decressin’s paper alone and draw the same conclusions as you stated above?