This is an appendix for our earlier post, “3 Underappreciated Indicators to Guide You through a Debt-Saturated Economy.” We’ll share a few extra charts and describe our use of the Fed’s flow of funds data, in Q&A format.
In the last chart of the main article, you showed that recessions normally follow a decline in risky borrowing, but can’t the same be said for borrowing financed by non-money savings?
No. Changes in borrowing financed by non-money savings typically lag changes in risky borrowing. In most of the recessions in our data set, borrowing that’s financed by non-money savings peaked either during or immediately before the recession. Here’s the chart:
What types of credit do you include in total borrowing and lending?
Our borrowing and lending figures include the Fed’s “credit market instruments” category, repos and some types of interbank lending. (Update: we clarify which types of interbank lending in the third question of this post.) The idea is to track any loan or debt security to its ultimate source of funding. For example, if a money market fund loans money to a dealer through the repo market and the dealer invests the proceeds in a Treasury bond (the collateral for the repo), then there’s no effect on the dealer’s net position as a lender/borrower. The dealer’s asset and liability flows can be netted out, leaving a borrowing by the Treasury that’s ultimately funded by money market fund shareholders through the repo loan.
Note that our analysis doesn’t help to evaluate mismatching of assets and liabilities. If the objective was to assess risks caused by mismatched assets and liabilities, then repos and interbank lending would be considered separately. This is a worthy objective, but it’s not our aim here. The purpose of our analysis is to track the amount of borrowing that’s not backed by non-money savings.
What do you do with bank deposits?
Deposits are treated separately for reasons discussed in the main article. Essentially, when deposits across the whole banking system are rising, that tells you that banks are lifting the supply of credit without a prior increase in savings. This is a riskier source of financing than non-money savings. It requires a banking license in a fractional reserve system, which is basically a license to print money.
Although changes in deposits aren’t shown directly in any of the charts in the main article, they’re reflected in net lending by banks. As shown in the chart below, these are approximately the same:
The close correlation is explained by the fact that a new deposit enters the banking system every time a bank makes a loan. If you could track the extra deposit as it flows through the economy, you would see that it splinters into smaller pieces that bounce from bank to bank as money is spent (or invested or gifted) and then splinter again. At any point in time, some pieces will sit in savings accounts or time deposits and others in checking accounts, while still others are converted to currency although usually not for very long. The pieces will have so many owners that our tracking exercise is impossible, but together they’ll remain equal to the extra deposit brought to life by the original loan. With some exceptions, the extra deposit doesn’t disappear until the loan is either paid down or sold off and moved out of the banking system. Therefore, deposits (plus currency in circulation) increase when banks expand net lending and decrease when net lending contracts.
Isn’t net lending by banks also reflected in the money supply aggregates?
Not exactly, because money supply aggregates don’t isolate the total amount of deposits. (See this Wikipedia link for definitions.) Economists who put huge faith in the aggregates in the late 1970s – resulting in many forecasting errors and a tumultuous, short-lived experiment with money supply targeting at the Fed – may have had more success with greater focus on bank lending and total bank-created money. To be clear, we’re not suggesting that any particular measure should be followed robotically to the exclusion of all else, just that bank net lending is more economically relevant than any of the “M”s. Here are the correlations:
How do your categories compare to those in the flow of funds report?
The Fed reports on financial flows for 29 different sectors, which it groups into broader categories including the personal sector (households, nonprofits and nonfinancial noncorporate business) and financial business (consisting of banks, the Fed and 18 other types of financial institutions). Although our bank sector is the same as the Fed’s, we divide up the nonbank financial sector differently. We group the following institutions in a category that we call “leveraged nonbank lenders”: government-sponsored enterprises (GSEs), agency- and GSE-backed mortgage pools, issuers of asset-backed securities, finance companies, mortgage REITs, security brokers and dealers, holding companies and funding corporations. These are institutions that borrow money to fund either lending activities or transactions with financial subsidiaries. Proceeds of loans or debt security issuance at these institutions are invested mostly back into the credit markets.
We combine the remaining financial institutions with the Fed’s personal sector, including pension funds, insurance companies, closed-end funds, ETFs and equity REITs. (The Fed adds items from the last three sectors to either households or its personal sector in certain parts of its report as well.) The investments made by all of these institutions are funded from income, not borrowing, with a few exceptions such as a modest amount of leverage in equity REITs. When these institutions buy a bond or other credit market instrument, the purchase is most likely funded from non-money savings.
How do you treat money market funds and mutual funds?
We divvy up money market funds and mutual funds according to the Fed’s data on how much of the funds’ assets are held by each sector. We then use investment allocation data to credit each asset type to the different sectors in proportion to their overall holdings. In other words, we implicitly assume the same mix of money market and mutual fund assets for each category of shareholder, varying only the ownership shares. By our estimates, this approach is neither badly inaccurate nor does it have a meaningful effect on our results.
For information, 75% of money market fund shares and 90% of mutual fund shares were held by domestic nonfinancial sectors (mostly households, nonfinancial businesses and pension funds) as of the end of 2013. The rest of the world held 6% of money market fund shares and 9% of mutual fund shares. Nonbank lending institutions (funding corporations, in particular) held the remaining 19% of money market fund shares.
How do you measure total borrowing?
We add personal and nonfinancial corporate borrowing to the government’s net borrowing. The idea is to isolate borrowing that’s being spent rather than invested in another loan or debt security.
Our figures are almost the same as the Fed’s figures for nonfinancial borrowing, with the exception that we net out the government’s credit market assets because most of those assets are lending programs. If we added, say, government-financed student loans to government borrowing without netting out the government’s lending, we would be double counting the amount of borrowing that flows into the economy through additional spending.
How do you estimate the amount of borrowing that’s funded by non-money savings?
We combine the lending that’s financed by our expanded personal sector with lending financed by nonfinancial corporations. Like the borrowing figures, lending is comprised of credit market instruments and repos, although taken from the asset side of the ledger not the liability side.
Isn’t there more than one way to calculate the same figure using the flow of funds report?
Flow of funds data has the nice characteristic that flows add up to zero when you total them across sectors while including any reported discrepancies (more on this below). There are no discrepancies for the Fed’s “credit market instruments” category, but the repo and interbank lending categories have small discrepancies. In any case, we calculated total borrowing and non-money savings twice – once by adding flows for the relevant sectors and then again by adding flows (with discrepancies) for all other sectors. These methods give the same answers.
How do you treat “leveraged nonbank lenders”?
The borrowing and lending flows of our “leveraged nonbank lender” category mostly balance out when netted and aggregated across the whole category. When the small net flows that remain are positive (showing a net lending position), these can theoretically be added to non-money savings (while negatives could be subtracted). We chose not to, for two reasons. First, the net flows are strongly negatively correlated (about -0.6) with the repo category discrepancy. This tells us that much of the net flow data is uncorroborated by reported flows in other sectors. Second, net flows are often balanced by “miscellaneous” financial statement items that appear unrelated to non-money savings, such as government bailout programs.
Because of the ambiguity in the data, we combine the net flows for leveraged nonbank lenders with the repo and interbank lending discrepancies. We show the total under the “not categorized” label in the main article’s first chart. For the second chart, we ignore the “not categorized” amounts entirely and base our calculations of risky borrowing percentages on the remaining data.
In the main article, you mentioned links between personal savings, public finances and the amount of foreign funding in our credit markets – can you see these links in the data?
Here’s a chart showing personal savings rates and funding from outside the U.S.:
Here’s a chart comparing U.S. government deficits to funding from outside the U.S.:
If I’m a central banker, are you saying I should pay more attention to risky borrowing and move away from short-term targets for inflation and unemployment rates?
Can you share your calculations and a finer breakdown for the different categories of funding?
We’re happy to provide spreadsheets, add detail and even extend the analysis to issues not considered here, although on a more commercial basis than the blog. If interested, e-mail firstname.lastname@example.org.