Reply to David Beckworth

David Beckworth has a new post which attempts to use a VAR (vector autoregression) to estimate how much of the decline in nominal GDP was due to monetary policy, and how much was due to other factors such as the financial crisis.  Because VAR is not my area of expertise, I hope people will take this reply as merely first stab at what I hope will be an interesting conversation.  I am pretty sure that some of my commenters are more knowledgeable about VARs than I am, and I know that David is more knowledgeable.David faced two disadvantages in trying to test my hypothesis with a VAR model:

1.  I have always been skeptical of VAR models because I worried that they tended to misidentify monetary shocks.  They often used the Fed funds rate, for instance, whereas I think monetary shocks are changes in the expected future path of nominal aggregates.  In the early studies this led to a “price puzzle” where contractionary monetary policy appeared to raise prices, which contradicts one of the most well-established theoretical propositions in all of macro.  This could have reflected the fact that higher interest rates might have been used as an indicator of tight money, and those high rates might have merely reflected expectations of more rapid nominal growth.  Later studies avoided the price puzzle, but that doesn’t mean they avoided the fundamental weaknesses of this approach.

2.  I have a very unconventional way of thinking about the monetary transmission mechanism.  A monetary shock is essentially a change in the future path of the money supply relative to velocity, in other words, a change in the expected NGDP growth.  If the NGDP target is 4%, for instance, than an expected NGDP growth rate of 3% is tight money, and an expected NGDP growth rate of 6% is easy money.  Two things prevent this from being a tautology.  One, under a fiat money regime I assume that policymakers can hit any NGDP target, if necessary with a futures targeting regime.  And second, that I am trying to explain movements in both nominal and real GDP.  Thus my hypothesis is roughly that when the Fed allowed NGDP expectations going one and two years forward to collapse, that caused a collapse in near term NGDP, which in turn (because of a relatively flat SRAS) caused a collapse in near term RGDP as well.

David Beckworth set up several specifications.  In the first he used unexpected changes in NGDP as the dependent variable, and unexpected changes in inflation, and also risk spreads, as the two independent variables.    I think that the more interesting dependent variable would be RGDP, as there are some economists (especially conservatives) who acknowledge that monetary policy could have prevented the collapse in NGDP, but who argued that the problem was partly or completely real, and that RGDP would have fallen sharply in any case.

However there are also many economists who think that if NGDP could have been supported at a roughly 5% growth rate the recession would have been far milder, indeed we might have avoided a recession entirely.  This group includes liberals who worry about the liquidity trap, as well as centrists (like James Hamilton, and presumably other new Keynesians) who worry that policy lags would have prevented monetary policy from having much effect on the sharp plunge in NGDP last fall.

As far as the independent variables are concerned, I have no idea which variable is the best measure of financial stress, but I have no objection to David using the Aaa-Baa yield spread.  For monetary policy, I would prefer unanticipated changes in NGDP growth expectations, not inflation expectations.  And where inflation is used, it might be interesting to look at TIPS spreads.  I suppose if monthly data is being used you could just average the first and last day, along with the mid-month figure for TIPS spreads.  I’m not sure about maturity.  For technical reasons the 5 year TIPS spread is more accurate, but the two year more closely corresponds to the time period we are looking at.   Real growth expectations would probably have to come from surveys of economists.  The inflation and real growth expectations could be added.

I should warn people that the place I end up will look dangerously tautological.  That is because most of us are taught to think of RGDP and NGDP fluctuations as being highly correlated in the short run.  And they are.  Furthermore, I assume that NGDP growth can be pegged just as surely as the price of gold or foreign exchange was pegged under Bretton Woods.  So why is it not tautological?  Three reasons:

1.  I favor targeting expected NGDP one or even two years forward.  It is very possible for near-term NGDP growth expectations to fluctuate even as long term expectations are stable.

2.  I favor targeting expectations, actual NGDP can fluctuate even when expected NGDP is very stable.

3.  I hope to use monetary policy to stabilize actual NGDP, but RGDP can fluctuate even when actual NGDP is stable.

So there are actually three slippages between targeting longer term NGDP growth expectations, and stabilizing near term RGDP.  So once again, my hypothesis is that if we stabilized NGDP expectations, over one or two years, the expected NGDP growth rate would be fairly stable in the near-term.  And I believe that if NGDP growth was expected to be fairly stable over the near-term, so would actual NGDP.  And I believe if actual NGDP was fairly stable over the near term, then RGDP would also be fairly stable over the near term.  But how to model all this?

In principle, if we had a NGDP futures market (and it is bizarre that we don’t, we have CPI futures for instance) the one or two year expected NGDP growth rate would be the independent variable, and the near term actual NGDP and RGDP growth rates could each be tried as the dependent variable.  The NGDP variable would answer the Keynesian question of whether NGDP futures targeting could have stabilized AD, thus making fiscal stimulus superfluous.  The RGDP dependent variable would answer the classical question of whether NGDP targeting is capable of smoothing the business cycle.  This is a two part question; can money stabilize aggregate demand, and can stable AD stabilize RGDP.

But I still puzzle over several issues.  In principle, the monetary shock variable could be constructed as follows:

1.  Expected one year NGDP growth rate minus 5%

In that case a bigger number would mean policy is more expansionary.  But this assumes the Fed is targeting growth rates, and I have argued that they should do level targeting:

2.  Expected one year NGDP growth rate minus [target NGDP minus actual NGDP]

Because we are now below trend, the figure in brackets would now be more that 5%, perhaps much more.  This is because we assume that policymakers will try to “catch up” to trend.

But here is the problem.  In the real world it is probably a little bit of both.  Policymakers may exhibit some trend reversion (although I see no evidence in this recession) but there is also a “let bygones be bygones” attitude, which makes nominal aggregates somewhat “difference stationary,” rather than trend stationary.  For non-economists, imagine a 5% trend line for NGDP going 200 years out into the future.  When we fall below trend by mistake do we try to return to the trend line, or start a new 5% trend line from where we are?   (A “memoryless” procedure.)  I think the Fed does a bit of both, but probably mostly starts a new trend line when they go off course.  On the other hand I think they should do “level targeting” which means trying to return to the old trend line.  Even worse, how much of each they do will vary over time.

Unfortunately, this means I’m not sure exactly how monetary policy should be modeled, but I am willing to accept that an estimate of expected one year NGDP growth minus 5% would be good enough.  For estimated NGDP growth I’d add private sector forecasters’ real growth estimate, and various inflation estimates (TIPS spreads and private forecasts.)  I notice that in the final VAR David models 2008:4 pretty well, but actual 2009:1 is worse than predicted.  Perhaps a severe shock causes a GDP decline that is spread out over several quarters.  I suppose the VAR already models this, but I wonder if some other VAR specification would do it better.  (This may also relate to the levels vs. rates of change issue.)

In a second regression, David uses the fed funds rate minus the expected NGDP growth rate.  Although I don’t like any measures of monetary policy using interest rates, I’d like to congratulate David for coming up with the best I have ever seen.  In an earlier post I argued that almost any time economists use “inflation” in macro models, they really ought to be using “NGDP growth.”  Nominal interest rates aren’t just affected by inflation expectations; they are also affected by changes in expectations of real growth.  Nevertheless, in my view this indicator would only work at turning points.  Consider the following example:

Suppose the Fed sets the Fed funds target at 4% when NGDP growth expectations have fallen to zero.  That would be tight money.  Now suppose NGDP growth expectations stay at zero, but the fed funds target is also lowered to zero.  Now monetary policy looks neutral.  And in a sense it is, but also note that NGDP is in a new steady state path of growing at 0% per year.  Eventually the public’s expectations will adjust, and in the very long run output will return to its natural rate.  But what worries me is the medium term, the period when nominal interest rates have adjusted, but not all wages and prices have adjusted.  So if we assume that in one quarter the fed funds target is 4%, and then in the next quarter it falls to 0%, then David’s definition the “tight money” policy has only lasted for one quarter.  But wages and prices will take many quarters to adjust to the new zero NGDP growth path; perhaps a few years.  In that case money is still tight by my definition, as expected NGDP growth (0%) is still below the rate of NGDP growth that was expected when wages were negotiated.

Once again, these comments are just my initial reaction.  I hope commenters can look at my response, and David’s VARs, and come up with some other useful insights.  I also want to thank David for taking the time to test out my hypothesis.


Tags:

 
 
 

27 Responses to “Reply to David Beckworth”

  1. Gravatar of TGGP TGGP
    3. October 2009 at 07:08

    You should link directly to his post:
    http://macromarketmusings.blogspot.com/2009/10/was-it-nominal-or-real.html

  2. Gravatar of ssumner ssumner
    3. October 2009 at 07:39

    Thanks TGGP, I made the change.

  3. Gravatar of The Arthurian The Arthurian
    3. October 2009 at 09:03

    A reply to Reply to David Beckworth

    Your post links to Beckworth’s. There is a reply to Beckworth’s post from ECB who says: “And the root of America’s national distress is the decline in America’s potential GDP growth path and feeble job creation over the last decade. The lax monetary policy that laid the ground for the crisis was in response to that real weakness.”

    ECB’s comment makes the topic of potential GDP significant. Therefore I am encouraged to ask the following:

    In your post, Scott, you write: “For non-economists, imagine a 5% trend line for NGDP going 200 years out into the future. When we fall below trend by mistake do we try to return to the trend line, or start a new 5% trend line from where we are? (A “memoryless” procedure.)”

    I know this is not the main point of your post. But maybe it is relevant. Which way is Potential GDP is calculated? And which way was it calculated back in the 1960s, when it was relevant to policy?

    I assume we now use the “memoryless” approach. The honest approach is to stick with the original trendline.

  4. Gravatar of 123 123
    3. October 2009 at 11:40

    VAR results imply that financial crisis and to some extent oil shock made intersest rate channel of monetary policy impotent, so massive doses of QE (credit easing) were urgently needed in Q4 2008.

  5. Gravatar of ssumner ssumner
    3. October 2009 at 12:07

    Aurthurian. Good question, as there is a lot of misunderstanding about this issue. NGDP growth is much more like inflation than it is like RGDP growth. So whereas many people feel we should target inflation, and believe that that target has nothing to do with the trend rate of growth in real GDP, I feel the same about NGDP. Under our current money regime I favor something like 5% NGDP growth, regardless of whether trend RGDP growth is 1%, 3%, or 6%. I don’t care about price inflation, I care about wage inflation.

    George Selgin and Bill Woolsey have convinced me that if we get a good, forward-looking monetary policy regime, then 2% or 3% NGDP growth is probably even better.

    I agree that the trend rate of real GDP growth in the US may have slowed from 3% to somewhere around 2% or 2.5%. But it doesn’t influence my views on monetary policy at all.

    123, I agree the interest rate channel was ineffective once rates hit zero (although they still haven’t quite got there, but close enough.) But I don’t think we needed a lot of QE. We needed to target NGDP expectations, and then the money supply would be endogenous. With 5% NGDP expectations, nominal rates would probably rise above zero, and there would be little hoarding of currency or reserves (assuming the Fed paid no interest on reserves.
    If we didn’t target NGDP expectations, then I agree that a massive amount of QE would have been necessary to boost expectations. But if we weren’t looking at expectations at all, that policy could have led to excessively high inflation.

  6. Gravatar of JimP JimP
    3. October 2009 at 15:40

    Buiter on the need for the Bank of England to start charging interest on reserves.

    http://blogs.ft.com/maverecon/2009/09/what-can-be-done-to-enhance-qe-and-ce-in-the-uk-and-who-decides/#more-5626

    And will the Fed listen? No. The deflationists and moralists are running the show.

  7. Gravatar of StatsGuy StatsGuy
    3. October 2009 at 16:11

    scott:

    “We needed to target NGDP expectations, and then the money supply would be endogenous.”

    This, btw, is your one-line answer to Krugman’s assertion that monetary policy (through the lense of the Taylor rate) would require 10 trillion dollars be added to base money.

  8. Gravatar of StatsGuy StatsGuy
    3. October 2009 at 17:10

    On Beckworth’s VARs…

    I dare not foray into the “which measure of which aggregate is best” debate, but with respect to these specific VARs, they seem to suffer from collinearity. (Do we have error bounds on the predictions? With n = 10, I’d suspect the marginal errors are pretty wide due to collinearity but the joint errors might be tigher.)

    My guess is there’s high collinearity between the measures during the period when the residuals are large (which is what we’re trying to predict) and little statistical leverage (since there are only 3 data points in the post-crisis world, 4 if you count Q3 of 08). If you look at the different models, the explanatory power of the different measures in the final 3 quarters depends largely on the relative explanatory power of those measures in the earlier several quarters. In other words, the model is mathematically using the leverage in the first 7 quarters to help parse the causality in the final 3 quarters. That presumes the relationships are stable during these time periods, which is highly questionable.

    Here’s an example of a specific sensitivity: In the final model, the one that attributes the largest portion of causality to the monetary policy changes, note that much of the induced correlation stems from the fact that the monetary policy change syncs well with a positive NGDP residual in Q2 of 2008, while the bond spread continues on a general downward path. Well, it just so happens that Q2 of 2008 was goosed by stimulus phase 1 spending (rebate checks). Without that data point (for example, if you threw in a stimulus dummy – perhaps leaving the oil price variable out) what would happen to the final model?

    In fact, probably largely because of this data point, the last model seems to tell us that the financial crisis had a POSITIVE projected impact on NGDP… Uh, does this pass the smell test? I can imagine a small effect, but a positive effect?

    As to the meaning of different variables across the 10 quarters, do we presume that the corporate bond spread “meant” financial crisis? In the early period, corporate bond spread seems to “mean” merely that there was a downturn coming, and investors know that in downturns it is the more vulnerable firms that suffer disproportionately. The interpretation of this isn’t that the credit markets _aren’t_ working, but rather that they _are_ working. The financial crisis narrative was that the credit markets stopped working, and this systemic failure triggered the collapse of AD by restricting the supply of credit that banks _should_ have (rationally) been offering (even given the banks’ predictions of plummeting AD and expected asset values at that time). That’s Fisher’s “Credit Channel Blob” story (and that is forevermore what I will call it).

    Finally, I’m not sure what measure to use for “financial crisis”… one possible thought is to use the average spread between the conforming and non-conforming home loans during that period. Conforming home loans are repurchased by Fannie/Freddie, and starting in Q4 of 2007, we some some action in that spread.

    Actually, now that I think of it, here’s a better suggestion:

    Use the LIBOR/Fed Funds spread.

  9. Gravatar of David Beckworth David Beckworth
    3. October 2009 at 20:09

    Thanks Scott for the reply. First, let me better explain the historical decomposition of the forecast error. In this exercise the VAR model is estimated across the entire sample (1971:Q1-2009:Q2) and then this fitted model is used to forecast the period in the graphs, 2007:Q1 – 2009:Q2. The difference between the forecasted and actual path of nominal spending can then be traced to non-forecasted movements in other variables in the VAR during this same time.(Note that VAR is a system of equations where every variable is being forecasted and can be influenced by the other variables in the VAR…it is an interacting dynamic system.) Utlimately, then, this exercise is all about trying to explain the forecast error of the nominal spending growth rate.

    It is typical to report the impulse response functions (IRFs)with VAR which show the typical dynamic response of one variable (e.g. nominal spending) to a non-forecated movement in another one (e.g. expected inflation). They also typically come with standard error bands to see if they IRFs are significantly different than zero. I left this out for the sake of brevity, but they show the effects of shocks to expected inflation, my monetary policy stance measure, spreads, and oil to be significant.

    With that said, let me move to something Scott said:”This is a two part question; can money stabilize aggregate demand, and can stable AD stabilize RGDP?” My exercise only attempts to tackle the first question. Yes, I could easily throw in RGDP but I wanted to narrowly address the first part because (1)some people do not believe so and (2) it seems obvious to me that given there are nominal rigidities(i.e. sticky prices), a sudden change in nominal spending has to affect RGDP.

    Regarding TIPs I got grief from ECB, a regular commentator at my blog, for using them in an earlier post to get expected inflation since they are contaminated with a liquidity premium. This was a fair criticism and ECB was kind enough to point me to the Philadelphia Fed quarterly survery of economic forecasters which has expected inflation. Another advantage in using this survey is that it goes back to the early 1970s. TIPs only go back to the late 1990s.

    I like your idea of using expected NGDP. It would be useful on two fronts as you note: (1) construct a better measure of the stance of monetary policy and (2) put it into a VAR instead of expected inflation. The good news is that the Philadelphia Fed survey has this information.

    StatsGuy noted something bizarre in the third graph: it shows a slight positive contribution to the forecast error for the financial crisis for 2008:Q4. A related but more general point is that the financial crisis shock largely has no effect in this figure as it is close to the forecasted valule over most quarters. This is notably different than in the previous graphs. The only difference between these graphs is the measure of the stance of monetary policy. Technically, then, what was a forecast error for the financial crisis in the previous graphs is now being accounted for by forecast errors in monetary policy. The implication, then, from the last figure is that monetary policy was the main original source of the problems. This may be a stretch to attribute almost everything to monetary policy.

  10. Gravatar of StatsGuy StatsGuy
    4. October 2009 at 03:39

    Just noticed that the data runs from 1970q1 (heh…), so that’s nearly 200 data points… much better, and so much for the single data point concerns. (and so much for night reading) With many data points we ought to be able to get error bounds on the predictions. But now we are depending on the stability of the relationships over 40 years – essentially, we’re assuming that this recession is like previous recessions, and using that to pull apart the collinearity in the latter period. But the argument of Fed people like Fisher is that this recession is _unlike_ previous recessions, and so the relationships would be different. (aka, “This one is different.”)

    Likewise, when I wrote Fed Funds/Libor spread last night, I mean Discount Window/Libor spread (wasn’t that obvious?), or alternatively some other measure of overnight interbanc lending rate vs. official lending rate. The forecast errors for this spread should be a more direct measure of systemic shock to the banking system than forecast errors for corp bond spreads, which could vary due to reasons other than a banking crisis.

    But if you have multiple measures of fiscal crisis or monetary policy expectations, you could of course use composites of the forecast errors from each measure.

    Anyway, sorry about missing the 1971 starting point of the data, and focusing on the 10 data point graphs.

  11. Gravatar of StatsGuy StatsGuy
    4. October 2009 at 04:14

    Another thought – if the VARs included a measure of the forecast error for fiscal spending, one could test the notion that deficits saved the world (at least, in the short term)…

  12. Gravatar of ssumner ssumner
    4. October 2009 at 05:03

    JimP, At least other people are talking this way, for the first 4 months of the year I was the only person calling for an interest penalty—so that’s progress. But it is a little late in the game. Much like the Great Depression, by the time they figure out monetary stimulus, most of the damage will have been done. Still, it’s not too late.

    Statsguy, Yes, that’s the one line answer, although I ramble on endlessly as I’m not sure everyone gets the point.

    Statsguy and David, Just thinking out loud, how about the ratio of an index of bank stocks to an index of all stocks as a measure of financial distress. Banks stocks alone wouldn’t work, as they would fall in an ordinary recession as well. But the ratio of bank to other stocks might pick up shocks specifically hitting banking. I’m no expert here, but I think others have also mentioned the Libor/ffr spread (is that the Ted spread?) so it might be worth trying Statsguy’s idea.

    I forgot that the TIPS only go back to 1997, so I guess you will have to use the forecast index.

    Statsguy#4, My employment data shows government employment falling over the past 12 months. So how can something have “saved the world” if it hasn’t been implemented yet?

    Thanks for your extensive comments on David’s work. I found them very helpful as I am not used to reading VAR papers.

  13. Gravatar of StatsGuy StatsGuy
    4. October 2009 at 06:51

    I think (?) Krugman’s argument was that the Fed deficits largely went to backstop state budget deficits and prevent state budget cuts (the “50 little hoovers”, since many states have balanced budget requirements). Some of it also went to very diluted tax cuts. So he would probably argue that total government employment (including state, which probably saw the bigger decline) is not the right measure.

    I’m not saying I agree, just anticipating what Krugman’s argument might be. I’m not even sure how to measure forecast errors in federal budget since the budgeting process is so long. (The last stimulus, which was proposed in Q4 08, was passed in Q1 of 09, and much of the spending is occurring just now, and will continue to be spent into next year.) Maybe focus on funds that were budgeted outside the standard appropriations process. Alternatively, OMB or CBO might have a series for historical budget projections. No matter what, some snotty nitpicker (like myself) is going to poke holes in it.

  14. Gravatar of Jon Jon
    4. October 2009 at 12:20

    The interpretation of this isn’t that the credit markets _aren’t_ working, but rather that they _are_ working. The financial crisis narrative was that the credit markets stopped working, and this systemic failure triggered the collapse of AD by restricting the supply of credit that banks _should_ have (rationally) been offering (even given the banks’ predictions of plummeting AD and expected asset values at that time).”

    The stuck credit channels theory simply does not comport with all the available evidence. The nail in the coffin is the Fed’s Senior Loan Officer survey data which consistently indicated that decreased lending was due to decreased demand (which the loan officers reported was due to a pullback from capital investment and reduced inventory holding). And to sweeten it up, most respondents indicated explicitly (over the course of many quarters) that neither liquidity, nor reserve position, nor capital constraints were a factor–answering “not important”.

    Use the LIBOR/Fed Funds spread.

    LIBOR should not be considered a reliable statistic. It’s nothing more than cartelization device at this point. The yield on commercial paper issued by banks to finance their balance sheets is much reliable. The data is collected by the Fed under a rigorous reporting regime, not a voluntary ‘guestimate your borrowing costs’ survey.

    This is not the first time this has happened. The “prime-rate” used be a real-statistic. Big borrowers used to actually use it as a benchmark index. Now its just a marketing scam played on home buyers. Through 2007, most major borrowing was indexed to LIBOR. Now that has also been abandoned by large borrowers as dishonest. OIS is one of the preferred replacements.

  15. Gravatar of Greg Ransom Greg Ransom
    4. October 2009 at 19:16

    “what do we mean by a “constant stance of monetary policy?”

    The original definition of “neutral money” was Hayek’s, which goes something like:

    “a market coordinating money rate of interest which did not create a boom and bust cycle by falsely misdirecting investment through the time structure of production goods”

  16. Gravatar of David Beckworth David Beckworth
    5. October 2009 at 04:58

    Jon:

    What maturity of commerical paper issued by banks would you recommend to be used?

  17. Gravatar of ssumner ssumner
    5. October 2009 at 05:45

    Statsguy, You said;

    “I think (?) Krugman’s argument was that the Fed deficits largely went to backstop state budget deficits and prevent state budget cuts (the “50 little hoovers”, since many states have balanced budget requirements). Some of it also went to very diluted tax cuts. So he would probably argue that total government employment (including state, which probably saw the bigger decline) is not the right measure.”

    Good points, but it really doesn’t undercut my argument at all:

    1. Krugman argues that tax cuts are ineffective, and that only more government spending can help.
    2. If the effects of more federal spending were offset by less state and local spending, then it would mean that no fiscal stimulus occurred. Which is exactly my point. In Keynesian models what matters is the combined government fiscal stance. So although the actions of state and local governments might provide a sort of “excuse” for fiscal stimulus to have not happened, nonetheless fiscal stimulus did not happen.

    Jon, Don’t forget to identify the person you are quoting (especially if it’s not me.)

    Greg, Good point, I addressed this in the most recent comment section.

  18. Gravatar of 123 123
    5. October 2009 at 11:40

    Scott, you wrote:
    “I agree the interest rate channel was ineffective once rates hit zero (although they still haven’t quite got there, but close enough.) But I don’t think we needed a lot of QE. We needed to target NGDP expectations, and then the money supply would be endogenous. With 5% NGDP expectations, nominal rates would probably rise above zero, and there would be little hoarding of currency or reserves (assuming the Fed paid no interest on reserves.
    If we didn’t target NGDP expectations, then I agree that a massive amount of QE would have been necessary to boost expectations. But if we weren’t looking at expectations at all, that policy could have led to excessively high inflation.”

    On September 17, 2008 3 month treasury yields were below zero. If we targeted NGDP expectations last year, fed funds rate would drop to zero and QE would be needed to prevent the drop of NGDP expectations. QE is needed until financial system gets stronger and can expand the money supply. Last September this was not possible without QE because there were serious doubts about the solvency of largest banks.

  19. Gravatar of StatsGuy StatsGuy
    5. October 2009 at 13:00

    Jon: “The nail in the coffin is the Fed’s Senior Loan Officer survey data which consistently indicated that decreased lending was due to decreased demand (which the loan officers reported was due to a pullback from capital investment and reduced inventory holding).”

    That is a very good nail indeed. Nice data (http://www.federalreserve.gov/boarddocs/snloansurvey/);

    In the Jan 09 survey, for example, banks overwhelmingly cut credit lines where they could and tightened lending standards. This has been cited as face-vale evidence of non-functioning capital markets. Your point – that banks did not cite capacity constraints as a reason – reinforces that this was less about the capacity to lend, rather than the desire to lend (or, alternatively, a lack of _creditworthy_ demand). The counter-argument is that the bank tightening was excessive (most banks tightened across almost all categories), but importantly this is not an argument about _capacity_ (i.e. artifical blockages). It’s an argument about bank judgment about economic future. Some people cite the collapse of the securitization markets as a capacity constraint, but you are absolutely right that this is only a constraint to the degree that banks can no longer act as a middleman in making bad loans that they then resell. The banks had plenty of their _own_ money to lend, but didn’t want to lend it.

    Another more minor point in the survey – banks indicated that they were issuing more contracts with a minimum rate on FLOATING rate loans. In other words, an artificial spread-increasing mechanism that is designed to preserve a minimum NOMINAL rate. How, exactly, do people who argue that it’s all real explain minimum contractual nominal rates on FLOATING rate loans (which have limited if any future inflation risk)? How can they call this “cheap” money?

    What a great data source, Jon.

  20. Gravatar of StatsGuy StatsGuy
    5. October 2009 at 15:23

    123: “Last September this was not possible without QE because there were serious doubts about the solvency of largest banks.”

    But the question is, to what degree did the insolvency result from the difference between the asset prices that existed in Jan 09, and asset prices that WOULD have existed if the Fed had credibly sustained NGDP growth (or even 2% inflation from the beginning of the crisis phase in September).

    Those who think this is a hokey distinction probably believed the Mark-To-Market credit derivative prices in Jan 09 were the “real” prices (and by real, I mean the nominal value that was _truly_ recoverable).

    But consider the ABX index, which arguably measures ground zero of the debacle…

    http://static.seekingalpha.com/uploads/2009/8/2/saupload_abx1.png

    And since then, it’s gone up more. Saying that monetary policy couldn’t fix those loan losses because they were “real” ignores the fact that the “reality” of the losses is contingent on beliefs about the stance of Federal monetary policy. The turnaround began when those beliefs changed from highly contractionary to only modestly contractionary (relative to the anticipated 2% trend). Yay.

    That’s not to say the crisis was not real. It was real, and many real and serious structural problems remain, but it was MUCH worse than it should have been.

  21. Gravatar of Jon Jon
    5. October 2009 at 17:27

    Scott: sometimes, it isn’t my intention to elicit a particular person’s response. Please.

    David: I’d be hard pressed draw a distinction between 1mo, 2mo, 3mo paper for these purposes. For my own work, I’ve used 1mo.

    For an illustration of the ‘bank-confidence’ crisis: plot ffeff, against financial 1mo, nonfinancial 1mo, and LIBOR.

  22. Gravatar of Jon Jon
    5. October 2009 at 18:51

    But the question is, to what degree did the insolvency result from the difference between the asset prices that existed in Jan 09, and asset prices that WOULD have existed if the Fed had credibly sustained NGDP growth (or even 2% inflation from the beginning of the crisis phase in September).

    This reminds of remarks from Turkey. A banker was asked if he was concerned about the financial crisis. His answer: no, the inflation-rate is so high, the nominal yield on loans generates enough free cash-flow as to preclude the possibility of insolvency.

  23. Gravatar of ssumner ssumner
    6. October 2009 at 16:40

    123; You said;

    “On September 17, 2008 3 month treasury yields were below zero. If we targeted NGDP expectations last year, fed funds rate would drop to zero and QE would be needed to prevent the drop of NGDP expectations. QE is needed until financial system gets stronger and can expand the money supply. Last September this was not possible without QE because there were serious doubts about the solvency of largest banks.”

    It’s much harder to predict thins than you’d think. The most expansionary monetary policy in American history (March to July 1933 was associated with a fall in the monetary base. That’s because the banking system would immediately get much stronger in 5% NGDP growth was expected, because those sorts of nominal growth expectations would dramatically raise asset values. Much of the bloated monetary base was a reaction to the liquidity trap, a reaction to the deflationary environment. I’m not saying you are necessarily wrong. If QE is the tool you are using to boost expectations, you might need a lot of it, as you say. FDR had a different tool (he raised the price of gold sharply) and thus didn’t need much more base money.
    I am saying its a very complicated problem, and Ben Bernanke said the same thing in 1995 when he coined the term “multiple monetary equilibrium.” Come to think if it, that’s worth a post! The same equilibrium could be associated with two very different money supplies.

    Statsguy and Jon, That’s great information, I wish I had it when I presented at GMU. Sorry I didn’t notice it in your previous comment Jon. But I can’t find it from the link. Is there another link?

    Those are all very good points both of you make.

    Jon, I wasn’t criticizing you. I try to answer every single comment. And 98% of the unidentified quotes are mine. So I just assume someone is quoting me unless otherwise indicated. Now that I know your purpose I’ll try to remember that and only respond if it seems clearly directed at me. Your input is very valuable.

  24. Gravatar of Jon Jon
    7. October 2009 at 17:14

    Scott:

    See for instance page 19 of the full report from January:

    A. “Possible reasons for tightening credit”
    a. current or expected capital position (Not Important 73.5%)
    b. less favorable economic conditions (Very Important 71.4%)
    h. current or expected liquidity position (Not Important 87.8%)

    4. Demand for loans
    A. Moderately weaker (54.7%), Substantially weaker (15.1%, About the same (20.8%)

    5.B. If weaker why
    a. inventory financing needs decreased (62.2% somewhat important, 18.9% very important)
    b) plant investment decreased (45.9% somewhat important, 48.6% very important)

  25. Gravatar of ssumner ssumner
    8. October 2009 at 06:21

    Thanks Jon.

  26. Gravatar of 123 123
    17. October 2009 at 12:12

    In a sense I agree with you that it is hard to predict. Maybe a heavy dose of QE last October would have turned things around quickly (and QE would be discontinued in December), maybe not. It all depends on the question was the problem with financial sector a liquidity issue or a solvency issue (from a perspective of summer 2008).

    I hope that multiple monetary equilibrium are just theoretical curiosities.

  27. Gravatar of ssumner ssumner
    18. October 2009 at 10:05

    123, I think a credible and explicit nominal target, level targeting, can make then a curiosity.

Leave a Reply