Modeling Financial Liquidity and Solvency

Too often in debates regarding the recent financial crisis, the event was regarded as a surprise that no one could have anticipated, conveniently forgetting those who pointed out sloppy banking, lending and borrowing practices in advance of the crisis.  There is a need for a well-developed model of how a financial crisis works, so that the wrong cures are not applied to the financial system.

All that said, any correct cure will bring about a predictable response from the banks and other lending institutions.  They will argue that borrower choice is reduced, and that the flow of credit and liquidity to the financial system is also reduced.  That is not a big problem in the boom phase of the financial cycle, because those same measures help to avoid a loss of liquidity and credit availability in the bust phase of the cycle.  Too much liquidity and credit is what fuels eventual financial crises.

To get to a place where we could have a decent model of the state of overall financial credit, we would have to have models that work like this:

  1. The models would have to have both a cash flow and a balance sheet component to them — it’s not enough to look at present measures of creditworthiness only, particularly if loans do not fully amortize debts at the current interest rate.  Regulatory solvency tests should not automatically assume that borrowers will always be able to refinance.
  2. The models should try to go loan-by-loan, and forecast the ability of each loan to service debts.  Where updated financial data is available on borrowers, that should be included.
  3. The models should try to forecast the fair market prices of assets/collateral, off of estimated future lending conditions, so that at the end of the loan, estimates can be made as to whether loans would be refinanced, extended, or default.
  4. As asset prices rise, there has to be a feedback effect into lowered ability to finance new loans, unless purchasing power is increasing as much or more than asset prices.  It should be assumed that if loans are made at lower underwriting standards than a given threshold, there will be increasing levels of default.
  5. A close eye would have to look for situations where if the property were rented out, it would not earn enough to pay for normalized interest, taxes and maintenance.  When asset prices are that high, the system is out of whack, and invites future defaults.  The margin of implied rents over normalized interest, taxes and maintenance would be the key measure, and the regulators would have to have a function that attributes future losses off of the margin of that calculation.
  6. The cash flows from the loans/mortgages would have to feed through the securitization vehicles, if any, and then to the regulated financial institutions, after which, how they would fund their future liabilities would have to be estimated.
  7. The models would have to include the repo markets, because when the prices of collateral get too high, runs on the repo market can happen.  The same applies to portfolio margining agreements for derivatives, futures, and other types of wholesale lending.
  8. There should be scenarios for ordinary recessions.  There should also be some way of increasing the Ds at that time: death, disability, divorce, disaster, dis-employment, etc.  They mysteriously tend to increase in bad economic times.

What a monster.  I’ve worked with stripped-down versions of this that analyze the Commercial Mortgage Backed Securities [CMBS] market, but the demands of a model like this would be considerable, and probably impossible.  Getting the data, scrubbing it, running the cash flows, calculating the asset price functions, implied margin on borrowing, etc., would be pretty tough for angels to do, much less mere men.

Thus if I were watching over the banks, I would probably rely on analyzing:

  • what areas of credit have grown the quickest.
  • where have collateral prices risen the fastest.
  • where are underwriting standards declining.
  • what assets are being financed that do not fully amortize, including all repo markets, margin agreements, etc.

The one semi-practical thing i would strip out of this model would be for regulators to score loans using a model like point 5 suggests.  Even that would be tough, but even getting that approximately right could highlight lending institutions that are taking undue chances with underwriting.

On a slightly different note, I would be skeptical of models that don’t try to at least mimic the approach of a cash flow based model with some adjustments for market-like pricing of collateral and loans.  The degree of financing long assets with short liabilities is the key aspect of how financial crises develop.  If models don’t reflect that, they aren’t realistic, and somehow, I expect that non-realistic models of lending risk will eventually be the rule, because it helps financial institutions make loans in the short run.  After all, it is virtually impossible to fight loosening financial standards piece-by-piece, because the changes seem immaterial, and everyone favors a boom in the short-run.  So it goes.