Sunday, August 28, 2016

Finance is Not the Economy

That's Where the Real Thinking is Done

Michael Hudson and Dirk Bezemer published a great article about a week ago on the net "Finance is Not the Economy".  Well worth a read.

The article focuses on one of AA's favorite topics: how the failure to account for the financial sector means that economic analysis is incomplete and therefore incorrect.  That is not to say the economics will ever be more than a best guess. 

A failure on at least two fronts--impact of the financial sector on the economy and the risk of financialization of the economy.

One thing did catch my eye - lack of a reference to Rudolf Hilferding.


The Arthurian said...

Hi! My first time here. I thought the "Finance is Not the Economy" article was really good. Bezemer tells a great story, and Hudson is a magnificent historian. But I think there is a problem with their analysis. Maybe you would be interested.

A response to Bezemer & Hudson's "Finance is Not the Economy"

"If you figure nominal debt to nominal GDP, or if you use the wrong "real" calculation, you can expect debt to look flat before the 1980s. I've been reading Finance is Not the Economy and I'm worried that Bezemer and Hudson find debt growing in proportion to GDP before the 1980s because they are using nominal values for debt..."


Abu 'Arqala said...


Thanks your comment. Posters are a rarity here. So أهلا و سهلا

Apologies for the unconscionable delay in responding. It’s been busy at work.

I’ve read your blogpost and understand that your central concern is the use of nominal values rather than real in the article.

I’ve got several reactions, but not enough time to post them in “one go”.

So to start …

I think that as a practical matter accurately determining the “real” value of the debt would be impossible.

(1) Finding the “Original Loan”

a. Economists would have to find the original loan amount and year to “real value”.

b. The major problem is loans are rolled-over/refinanced. Thus, there is no neat correspondence between a corporate loan and its purpose as there is say with a residential mortgage. A corporate loan made in 2016 does not necessarily finance an expenditure made in 2016. It may be refinancing a loan made in 2014 which in turn refinanced a loan made in 2009 and so on. And these refinancings may not be on a strict one to one refinancing basis. A loan or bond in 2016 may refinance a number of prior years’ loans. Also a new debt may include refinancing of existing individual loans as well as new loans for yet to be made expenditures. There could also be cross instrument refinancings, e.g., loans to bonds. Loans to CP. CP to loans.

c. To put this into context, some examples of loan tenors and thus likely rollovers.

i. As of June 2015, US Treasury marketable debt had an average maturity of 70 months according to an August 2015 US Treasury report to TBAC.

ii. The average C&I loan is shorter. All outstanding US C&I loans have an estimated average 727 day maturity according to FRB E2 as per an August 2016 survey.

iii. Most CP --roughly the kilometer to the C&I mile in terms of outstandings, that is roughly 60% of C&I volume-- is under a year and clustered around a 30 day maturity.

d. A key problem in real valuing is how the economist knows when she’s found the “original” loan and separating one original loan from others that might be included in a larger “omnibus” loan.

e. Given that existing loans are not likely to be the original loans that funded the original expenditure, one could try to find the original loans in two ways.

f. Identify major original capital investments (those likely to affect GDP long term) and then seeing if there were a public record of the initial loan. One couldn’t merely go by cost because projects are financed by debt and equity. (More on that topic later). One could ask companies to provide details of each original loan for each investment expenditure. Somehow I think that request would fall on deaf ears, assuming that the companies are able to do this.

g. Another way would be to try and trace back to the original loan through existing loans. Clearly a much more difficult path. At least two problems with that. First, there may no longer be an existing loan for the original investment which is still contributing to GDP. Second, rollovers as outlined above. A borrower may take out an initial loan to fund an investment but then later refinance the debt.

For example, a real estate developer will take out a construction loan and then refinance post construction with a bond or bank/insurance company loan. Later the building owner may refinance again as he couldn’t obtain a long enough tenor (maturity) to match the cash flow, because he wants a certain capital structure, wants to improve the building, or use the building as collateral to fund the purchase of another building, etc. One would have to be able to trace back current outstanding loans through this history back to the original loan for investment.

h. Either method would require the unlikely-to-occur acquisition of an extremely large amount of data and require an army of analysts and one or two Cray computers to compile the data.

More to come ….


The Arthurian said...

Thank you for the kind welcome, AA.

I was happy to receive your reply. You put so much thought into it. Regarding the adjustment of debt for inflation, you say "Economists would have to find the original loan amount and year to “real value”." Yes, exactly.

When real debt is calculated the way real GDP is calculated, many years' borrowings are adjusted as if all the borrowing happened in one year. For some purposes, perhaps, that is fine. But when evaluating economic growth, it is important to value the purchasing power of each year's addition to debt separately. If this is not done, the purchasing power of existing debt is undervalued by inflation. This is central to my difficulty with the "Finance is Not the Economy" paper.

My practice is to take each year's addition to debt, adjust the addition as I would adjust that year's GDP, and then add all the adjusted values together to get real debt.


You write: "The major problem is loans are rolled-over/refinanced."

Yes, it can make things very complicated, as you show. But I have managed to convince myself there is a simple solution. I am interested to know what you think of it.

If I borrow $100 in year 1, repay it in year 2, but then borrow $100 again in year 2, the inflation adjustments for the two "year 2" transactions are identical. I can safely ignore both transactions, and assume that the $100 from year 1 remains on the books.

(The borrower may get a lower interest rate because of the "year 2" transactions. This will affect his future debt service payments but it does not change the purchasing power of the original loan amount. And the year 2 borrowing is not spent into circulation as the year 1 borrowing was.)

If I borrow $100 in year 1, repay it in year 2, and then borrow $150 in year 2, I do the "real" calculation for the $50 net increase in debt for year 2.

Mine is a crude calculation, compared to the complex lending arrangements you describe. But my method works easily with the "change from year ago" values of TCMDO debt at FRED (for example). And my method is not more crude, I tell myself, than the method used to convert between real and nominal GDP.

I could certainly have this all wrong. I'm no economist. But I would be interested to know if you think my method provides an acceptable solution to the "two Cray computers" problem.

Abu 'Arqala said...


Thanks yours.

It would help me in my response if you could work through some numbers with me.


Scenario A

Year zero outstanding nominal debt is $1,000
Year one outstanding nominal debt is $1,100 and inflation is 10% per annum.
Year two outstanding nominal debt is $1,275 and inflation is again 10% per annum.
Year three outstanding nominal debt is $1,350 and inflation is 15% per annum.

Scenario B

Year zero outstanding nominal debt is $1,000.
Year one outstanding nominal debt is $1,000 and inflation is 10% per annum.
Year two outstanding nominal debt is $1,000 and inflation is again 10% per annum.
Year three outstanding nominal debt $1,000 and inflation is 15% per annum.

What would be your adjusted real debts for each of the years? Just so there's no confusion on my end or at least it's minimized, can you show the equations?



The Arthurian said...

To reduce the "real debt" question and make it manageable, I use rules and assumptions:

When a loan is refinanced, the old loan is paid off and a new loan is taken out at the same moment. The price index does not change in the interim, because there is no interim! If the nominal amounts are equal, then the real value repaid is equal to the real value borrowed anew. The "real" calculation is unaffected by the refinancing.

If a loan is repaid and a new loan is made while prices remain at the same level, only the difference in principal amounts need be considered for the calculation. In other words, the real calculation is based on change in the accumulation, and on the price level at the time of the change.

If a loan is repaid and another loan is made when the price level is higher, the two are worked out in different steps of the calculation. A separate step is required for each new price level.

Also, I think of principal and interest as two separate amounts, and my calculation is for principal only. One thing at a time.

Maybe I should say, too, that I think in terms of macro and aggregate numbers, as opposed to micro and individual loan circumstances. Could be that's just a way to cope with things that I don't know about...


I notice that scenarios A and B suffer the same inflation. I want to begin by converting your inflation rates to price index values. I'll use Year 0 for the "base year".

Year 0: Index Value 100.00 // Then 10% inflation -> 110.0
Year 1: Index Value 110.00 // Again 10% inflation -> 121.0
Year 2: Index Value 121.00 // Then 15% inflation -> 139.15
Year 3: Index Value 139.15

This reply consists of four parts.

The Arthurian said...

(The Price Index is as given above, and Year Zero is the Base Year.)


Year 0: 1000 Nominal, 1000.00 Real
Year 1: 1100 Nominal, 1090.91 Real
Year 2: 1275 Nominal, 1235.54 Real
Year 3: 1350 Nominal, 1289.44 Real


Year 0
Existing and new debt at end of period: 1000 nominal = 1000 real
Accumulated Real Debt: 1000 total

Year 1
Addition to debt: 100 nominal
Price Index Ratio: 100.00/110.00 = 0.9091
Real purchasing power: 100 nominal * 0.9091 ratio = 90.91 real
Accumulated Real Debt: 1000 balance + 90.91 addition = 1090.91 total

Year 2
Addition to debt: 175 nominal
Price Index Ratio: 100.00/121.00 = 0.8264
Real purchasing power: 175 nominal * 0.8264 ratio = 144.63 real
Accumulated Real Debt: 1090.91 balance + 144.63 addition = 1235.54 total

Year 3
Addition to debt: 75 nominal
Price Index Ratio: 100.00/139.15 = 0.7186
Real purchasing power: 75 nominal * 0.7186 ratio = 53.90 real
Accumulated Real Debt: 1235.54 balance + 53.90 addition = 1289.44 total

I hope the format I've used is a help and not a hindrance.

The Arthurian said...

(The Price Index is the same as for Scenario A, and Year Zero is again the Base Year.)


Year 0: 1000 Nominal, 1000.00 Real
Year 1: 1000 Nominal, 1000.00 Real
Year 2: 1000 Nominal, 1000.00 Real
Year 3: 1000 Nominal, 1000.00 Real


Year 0
Existing and new debt at end of period: 1000 nominal = 1000 real
Accumulated Real Debt: 1000 total

Year 1
Addition to debt: None
Price Index Ratio: 100.00/110.00 = 0.9091 (applies to changes in debt)
Accumulated Real Debt: 1000 total

Year 2
Addition to debt: None
Price Index Ratio: 100.00/121.00 = 0.8264 (applies to changes in debt)
Accumulated Real Debt: 1000 total

Year 3
Addition to debt: None
Price Index Ratio: 100.00/139.15 = 0.7186 (applies to changes in debt)
Accumulated Real Debt: 1000 total

My first thought regarding Scenario B was that I had something wrong. I came up with 1000 Year Zero dollars, the same for all the years. I am now satisfied that I got it right: If I start with 1000 Year Zero dollars, and add nothing to it, I end up with 1000 Year Zero dollars (no matter what year it is).

The Arthurian said...


For calculating real debt, the simplifying fact is that the current price index applies only to the current change in debt. The current price index does not apply to debt accumulated in prior years.

The initial debt balance is always an accumulation from prior years. This initial accumulation cannot be decomposed into annual increments, because the required information is unavailable. (This problem always arises with the real debt calculation.) My solution is to say that we are looking at the effect of inflation since the start-date.

With GDP the current price index applies to the whole of GDP because nothing remains in GDP from the prior year. With debt, there is always a balance left over from the prior year, and so the current price index is useful only for the current year's changes to the accumulation. Thus, the real debt calculation cannot be the same as the real GDP calculation. As a result, the real-to-real version of the debt-to-GDP ratio differs from the nominal-to-nominal version.

However, the popular assumption seems to be that the nominal-to-real calculation that works for GDP will also work for accumulated debt. (For example, this Wikipedia table includes columns for nominal and real measures of U.S. Federal debt. Footnotes 58 through 64 indicate that the same nominal-to-real calculation has been used for the Federal debt and for GDP.)

Under the erroneous assumption that the same calculation can be used for both debt and GDP, the real-to-real and nominal-to-nominal versions of the debt-to-GDP ratio are identical. However, real purchasing power is understated by that calculation. When an inquiry concerns economic growth or purchasing power, real values must be used. This leads to my concern with the Bezemer & Hudson paper.


I had the hardest time getting started with your scenarios because I couldn't come up with a satisfactory way to check my results. Finally I set up an Excel sheet with FRED data, and did the work in Excel as I always do. When I was satisfied with it, I copied the sheet and plugged your numbers into it.

I uploaded my Excel file to Google Drive and made it available to view and download.

I thank you for your time and attention.


Abu 'Arqala said...



This is very useful.

As is sadly usual, my response is likely to be delayed. This time by the "festive" season.


Abu 'Arqala said...

Thanks again for your posts.

I see potential issues with using the net change approach you’re advocating.

First, additions to debt are real-valued at the real value factor (RVF #1) when the net increase takes place but reductions in that same debt are real-valued at the real value factor when the net decrease takes place (RVF #2). That is, two different RVFs are being used for the same amount: one on entry and one on exit.

Scenario A – First Three Years As Agreed:

Year 0: 1000 Nominal, 1000.00 Real
Year 1: 1100 Nominal, 1090.91 Real
Year 2: 1275 Nominal, 1235.54 Real
Year 3: 1350 Nominal, 1289.44 Real

Additions: Each Year Inflation at 10%

Year 4: 1450 Nominal, RVF 1.53065, 1355 Real
Year 5: 1550 Nominal, RVF 1.68375, 1414 Real
Year 6: 1650 Nominal, RVF 1.852087,1468 Real
Year 7: 650 Nominal, RVF 2.037295, 977 Real

The “problem” we face in Year 7 is that we don’t know what years the 1,000 reduction should be applied to.

Assumptions about application of these funds lead to different results.

The net change model result is 977 Real.

But, if the 1,000 is a repayment of Year 0 in full, Real should be 468.

On the other hand, if Years 1 through 6 have been repaid in full with a partial payment on Year 0, then we have to deduct 650 from Nominal and 468 from Real. As an interim result, both Real and Nominal are now $1,000. The remaining $350 from the 1,000 Nominal payment (1,000-650 for additions in Years 1-6) results in both Real and Nominal outstandings of 650 (because the RVF0=1).

Thus, for just these assumptions we have values of 468, 650, and 977.

Second, but I see another perhaps more significant impact from using a net figure. It masks changes in value arising from new loans and repayments of old loans within a single year.

Same modified Scenario A above but with an examination of Year 5.

Assume in Year 5, the $100 net increase is actually composed of $900 in repayments of Year 0 and new loans of $1,000. Refinancings of existing debt are assumed to be 0.

The $900 should be deducted at face value (RVF0=1) from Year 4’s Real amount, giving an interim figure of 435 (1355-900). Or in other words what’s left at this stage is 100 in Year 0 debt plus 355 in real value of debt for Years 1-4.

The 1,000 in new loans should be revalued at 594 (1,000/RVF Year 5 --1.68375) for a total of 1,029 vs the model’s 1,414.

While these are “extreme” examples, they highlight the potential for the real value calculated by the net change method being imprecise. Over a long period, individual year imprecisions could accumulate making the “real” number less and less precise. And thus less useful for analysis.

I suppose one could argue that with so many transactions there would be offsetting imprecisions. So this method would produce a “reasonable” estimate of real value. I don’t think we have the information to make that assumption, particularly for the private sector transactions.

I have some more thoughts on the exercise of comparing debt to GDP which I’ll try and put down in writing in the not too distant future.

These center on, but are not limited to, the conceptual model/economic theory that underlies this exercise and which we are trying mathematically model to enable us to “test” its validity. Or in other words what is the causal relationship we are positing, are we positing more than one causal effect from "debt", is "debt" a proxy for another variable?

The Arthurian said...

AA, You have given me something wonderful this holiday season: a puzzle to occupy my mind.

My first reactions are very often wrong. It is always best for me to wait before responding. But I want to indicate what my first reactions are. You write:

"The “problem” we face in Year 7 is that we don’t know what years the 1,000 reduction should be applied to."

Hmm. True. We do not know which lenders were repaid, so we cannot say how much the lender was harmed, nor how much the borrower was helped by inflation.

On the other hand, if the entire debt was repaid in year 7, nominal debt goes to zero, but real debt falls to 658.26 by my calculation. This amount represents the full value of the "erosion" of debt -- the loss to lenders and the gain to borrowers.

AA, you also wrote:
"But, if the 1,000 is a repayment of Year 0 in full, Real should be 468."

If I understand correctly, you are subtracting 1000 nominal from 1468 real. I would subtract 1000 nominal from 1000 nominal, giving a nominal year zero balance of zero. And I would subtract 490.85 (the year 7 real value of 1000 nominal repaid) from 1000 (the year 0 real value of 1000 nominal originally borrowed) and say the year zero erosion amount is 509.15 real.

I definitely need to take more time to work through your numbers and think about what you are saying. For now, let me thank you again for this puzzle. And I hope your holiday is a good one.

Abu Arqala said...


I think your two comments go right to the heart of the conceptual issue I raised at the end of my last comment.

Here are a few more comments which I hope will assist you in working through the puzzle.

If the intent is to study a causal link between debt and GDP, then the key question is what drives income production. Debt by itself? Or what is done with the proceeds from the debt?

I'd argue debt by itself doesn't increase income.

Rather income is increased by what is done with the proceeds from the debt. Purchasing or creating an asset with the proceeds of the debt. That's the central argument of Finance is Not the Economy --some debt is not particularly effective in generating income as measured by GDP because it is used to purchase non income generating assets.

I think one can categorize those assets into two conceptual types: short term "working capital" assets or long lived assets.

(1)Short Term Assets and Financing

Working capital assets would be raw materials to be transformed into inventory then sold generating receivables and finally cash. In some cases finished goods not raw materials are purchased direct into inventory then later sold generating receivables and cash.

These transactions take place usually within a "working capital cycle" (WCC) one year or less and their income generating effects are primarily if not solely related to that WCC (one year) period. As such they should be properly valued at face/nominal value in most cases.

Some issues:

(a)Identifying this debt and segregating it from capital investment related debt would add another layer of difficulty to the already difficult task of accounting for debt according to my theoretical approach.

(b)Not all WCC cycles uniformly end 31 December. Example, a receivable for a sale in Year X may turn into cash in Year X+1 with the underlying refinancing repaid that year (X+1). Income would be booked in Year X for almost every firm (the majority of firms use accrual not cash accounting). The cash from the receivable in Year X+1 wouldn't be an income event that year and would be used to repay the WC loan.

But if we assume most WCC loans are repaid prior to 31 December, then those debts don't appear in the YE debt stock and therefore are not "counted" as factors that affect (increase or decrease)Year X income/GDP.

(2) Long Lived Assets and Financing

Long lived assets (PPE for example, or R&D)which generate income over multiple years should be valued at real value original cost.

(3) Production versus Financing

Gains or losses resulting from financial developments (inflation or deflation) are separate from production/income generating processes and the posited causal link between investment (proxied by debt) and income.

Some issues:

(a)It would be almost impossible to identify such transactions.

(b)Such gains or losses are not included in a firm's net income (unless hyperinflation accounting is being used) or national GDP. I think that reducing the "causal cost" to reflect this type of gain or loss would distort the equation. A cost is reduced but the income not recorded

(c)Unless I've missed something, it would also seem to be inconsistent with your current model which keeps Year 0 dollars at original cost and present values each succeeding year's net change in debt "back" to Year 0 real values.

Happy holiday to you as well.