Lying With Statistics (Financial Services Lobbying Edition)

An excellent post by Elizabeth Warren at Credit Slips reviews some of the canards that have been successfully presented to promote the interests of various business interests against hapless consumers. The latest involves payday lending. The very fact that the Pentagon has come out against payday lending (they’ve proposed a usury ceiling of 36%) should be a sign of who is on the side of right here.

From Credit Slips:

When the credit industry lobbied Congress for adoption of the bankruptcy amendments, they made a powerful claim: Bankruptcy costs every American family $400. The number was pure fabrication, but the number was repeatedly quoted in newspapers, magazines and in Congress. It offered elected representatives a lot of cover to explain to the folks back home how they could vote to squeeze more money from working families and put it in the hands of a dozen or so credit card issuers. Adam Levitin shows us that another number has been drawn out of thin air: the Mortgage Bankers Association claims that any amendment to the bankruptcy laws to deal with subprime mortgages will increase mortgage rates for all homeowners by two percentage points–recently dropped to 1.5 points. Adam is doing a great job fighting back, but, as it was with the $400, academics don’t have the same PR machine.

Now there’s a third data claim: Payday lending is good for families. Once again, the claim is wrong, but the industry is pushing it hard in the media. Maybe only a small part of the academic world realizes the importance of data in legal policymaking, but private industry seems to understand very well the power of numbers.

A report from the Center for Responsible Lending points out that the study’s claim is based on the rate of bounced checks, but the dataset mixes together returned check data from states that permitted payday lending and from states that prohibited it. The study also looks at the number of FTC complaints filed, but the higher reporting rate in North Carolina was true both before and after payday lending was banned. (I’m not sure what a higher rate of filing FTC complaints means anyway–maybe just more activists in the state who urge prople to write the FTC?)

The University of North Carolina put together a detailed analysis of payday lending by studying low-income households after the ban on payday lending went into effect. Instead of trying to read the tea leaves of regional rates of bounced checks, the CRL took the same approach as Credit Slips’ own Angie Littwin took in her research: they went directly to the people who were the targets of payday lending and surveyed them. The result? Payday lending had no discernible effect on the availability of credit. In addition, twice as many borrowers reported they were better off without payday lending than they had been with it.

The payday loan study takes on additional credibility because it is written by a researcher at the Federal Reserve and a grad student. (That makes it sound like a study from the Federal Reserve, but the paper says it is not.) There is no reason to believe the mistakes in the study are intentional, but they are severe. This isn’t one of those academic interpretation questions or quibbles at the margins. These mistakes go right to the heart of the only support for the claim that payday lending is beneficial.

Last year the Pentagon went to Congress to say that payday lending was interfering with troop readiness, and Congress outlawed payday loans to military families. Some state legislatures are now looking hard at exending the same protection to all their citizens. They should be abke to do so without being fed bad numbers. And if bounced check charges are out of control, then they should take a look at those as well–not turn loose payday lenders so they can compete with banks to squeeze families harder.

Numbers are powerful. But wrong numbers can do a lot of damage.

Print Friendly, PDF & Email


  1. Anonymous

    The work by these two authors is not just shoddy. It’s dangerous and wrong. Indeed, so dangerous and damaging that there really should be some tort for ‘wrongful numbers’. The two people have just commited that tort and hundreds of thousands of people, including children, will suffer real damage as a consequence. Yet, our stenographic media will blithely, even in too many cases, cynically, spread this nonsense — and thereby, join in the tort.

  2. Anonymous

    We have a lotto mentality world that is addicted to returning to the stock market after nursing hangovers; there is zero connection with reality in the market in regard to fundamental logic, thus one needs to factor in addictive behavior into this chaos!

    Future value is all about confidence levels!!!!!!!!!!

    The confidence level is the interval estimate in which the VaR would not be expected to exceed the maximum loss. Commonly used confidence levels are 99% and 95%. Confidence levels are not indications of probabilities.

    Consider a trading portfolio. Its market value in US dollars today is known, but its market value tomorrow is not known. The investment bank holding that portfolio might report that its portfolio has a 1-day VaR of $4 million at the 95% confidence level. This implies that under normal trading conditions the bank can be 95% confident that a change in the value of its portfolio would not result in a decrease of more than $4 million during 1 day. This is equivalent to saying that there is a 5% confidence level that the value of its portfolio will decrease by $4 million or more during 1 day. A 95% confidence interval does not imply a 95% chance of the event happening, the actual probability of the event cannot be determined.
    The key point to note is that the target confidence level (95% in the above example) is the given parameter here; the output from the calculation ($4 million in the above example) is the maximum loss (the value at risk) at that confidence level.

    ▪ VaR at a particular confidence level is calculated using the percentile function. For example, if we computed 5000 simulations, our estimate of the 95% percentile would correspond to the 250th largest loss; i.e., (1 – 0.95) * 5000.
    ▪ Note that we can compute an error term associated with our estimate of VaR and this error will decrease as the number of iterations increases.
    Monte Carlo simulation is generally used to compute VaR for portfolios containing securities with non-linear returns (e.g., options) because the computational effort required is non-trivial. Note that for portfolios without these complicated securities, such as a portfolio of stocks, the variance-covariance method is perfectly suitable and should probably be used instead. Also note that MC VaR is subject to model risk if the market model is not correct.

    According to “Quantitative Risk Management”, McNeil, Frey, Embrechts, 2005 pg. 38.
    Given some confidence level the VaR of the portfolio at the confidence level α is given by the smallest number l such that the probability that the loss L exceeds l is not larger than (1 − α)

    In probabilistic terms VaR is a quantile of the loss distribution.

  3. Anonymous

    The key point I like to point out to people of below average means, is this: the bank can be 95% confident that a change in the value of its portfolio would not result in a decrease of more than $4 million during 1 day.

    Ok, Hello (out there), the key factor in this model, as with many models is time, time in this case, refers to 1 day, e.g, the possible loss to the model in one days time.

    Thus, if you make the wrong bet and the time-based result impacts of performance related to 1 day, your model must assume that your confidence remains the same in regard to VAR going forward.

    Thus your model can gain in negative momentum in the form of trying to anticipate the probability of either more good days or bad days, but still related to the next day VAR.

    This is how a drunk gambler at a casino justifies making bigger bets after tossing “good money” after “bad money” by using theoretical statistical deviations from reality to justify the next bet, which more than likely will be the wrong bet.

    The solution to this problem is more vodka!

  4. Anonymous

    Let us say, a daytrader pondering future market results, who feels invigorated and filled with lust and passions bets the family farm on financial/banking stocks or beat up HB-type securities, believing for some reason of ill conceived logic that these securities are now under valued in relation to current 10 year Treasury notes — that are yielding some variable interest rate, which is marked as being greater than the out come of this potential bet.

    If I were a crazy Russian hedge fund strategist, I would suggest looking more closely at the theory relating to Brownian motions and thus the application and theory of random walks (which they are about to undertake)!

    You people that make assumptions as these, to bet farms, need to think in terms of a drunk, who has consumed high quantities of vodka, as you proceed on your “random walk”>

    The daytrader, as a drunk, strikes out upon a walk Monday, away from the lamp post/computer terminal, which contains capital efficiency, i.e, your utility for future value, future consumption; the farm.

    From this PC daytrade station, you will lurch forward by making a step to call a broker via your terminal screen, and thus take random bets every 1000/second (OR WHATEVER); after two SUCH steps or lurches, you may wobble about and return to your original position from which you started.

    Your random pattern of daytrading, however, is subject to a statistical pattern, where the losses or gains you make in relation to your starting point, ARE proportional to the square root of the number of chips you have placed upon the DAYTRADE CASINO table.

    Thus, your current position is somewhat related to your level of cognition and ability to discern how much vodka you should drink and thus how wild your next bets will be, because the distance to your next gain will be related to your loss and you will have to return to your starting point with either empty pockets and a shirt covered in FOUL SMELLING mucus, or stay where you are and wait until sober, and thus be able to make a more educated guess as to the reality associated with yield returns!

    I propose a drink, a toast to all our daytrader friends that will be lost in the great blizzard of 2008. Za vashe zdorovye!!!!

    I love you crazy guys, I embrace you AND HOPE TO SEE YOU AT THE NEXT LAMP POST!!!!!

  5. Anonymous

    But concerns of Bachelier were very far from Pearson’s; In his thesis entitled “ Théorie de la Spéculation”, he was studying the fluctuations of stock-market prices as they vary up and/or down. Let us just translate the daily quotations in particle positions at equal time intervals, we are faced with a series of random numbers. Applying the central limit theorem, Bachelier describes these values through a Gaussian dispersion law. Calculating the probability P that the price be equal or larger than a given value on a given day, he showed P obeys Fourier equation. He describes the elementary process as a law of diffusion of probability. As a fine mathematician, the second part of his derivation relates to the limit case where the variable is a continuous one and still obeys the same equation . Later on this first model received considerable developments, a srandom walk is just the sum of random variables. A famous theorem on random walks on an integer lattice was derived in 1921 by G. Pólya (1887-1985) — a renown mathematician who coined the name of the “central limit theorem” — giving a strong basis to the classical model. But let us remember that the movement of material particles in suspension in a liquid were thestarting point of our speculations. It is interesting at this stage to mention some later developments of the theory of the random walk. Starting with the stricto sensu Brownian motion of a particle in suspension in a fluid, people got interested in the random walk of …anything.

Comments are closed.