Here is an example.
EXAMPLE 4 The case of the missing vans
Auto manufacturers lend their dealers money to help them keep vehicles on their lots. The loans are repaid when the vehicles are sold. A Long Island auto dealer named John McNamara borrowed more than $6 billion from General Motors between 1985 and 1991. In December 1990 alone, Mr. McNamara borrowed $425 million to buy 17,000 GM vans customized by an Indiana company, allegedly for sale overseas. GM happily lent McNamara the money because he always repaid the loans.
189
Let’s pause to consider the numbers, as GM should have done but didn’t. At the time GM made this loan, the entire van-customizing industry produced only about 17,000 customized vans a month. So McNamara was claiming to buy an entire month’s production. These large, luxurious, and gas-guzzling vehicles are designed for U.S. interstate highways. The recreational vehicle trade association says that only 1.35% (not quite 2800 vans) were exported in 1990. It’s not plausible to claim that 17,000 vans in a single month are being bought for export. McNamara’s claimed purchases were large even when compared with total production of vans. Chevrolet, for example, produced 100,067 full-sized vans in all of 1990.
Having looked at the numbers, you can guess the rest. McNamara admitted in federal court in 1992 that he was defrauding GM on a massive scale. The Indiana company was a shell company set up by McNamara, its invoices were phony, and the vans didn’t exist. McNamara borrowed vastly from GM, used most of each loan to pay off the previous loan (thus establishing a record as a good credit risk), and skimmed off a bit for himself. The bit he skimmed amounted to more than $400 million. GM set aside $275 million to cover its losses. Two executives, who should have looked at the numbers relevant to their business, were fired.
John McNamara fooled General Motors because GM didn’t compare his numbers with others. No one asked how a dealer could buy 17,000 vans in a single month for export when the entire custom van industry produces just 17,000 vans a month and only a bit over 1% are exported. Speaking of GM, here’s another example in which the numbers don’t line up with each other.
EXAMPLE 5 We won!
GM’s Cadillac brand was the best-selling luxury car in the United States for 57 years in a row. In 1998, Ford’s Lincoln brand seemed to be winning until the last moment. Said the New York Times, “After reporting almost unbelievable sales results in December, Cadillac eked out a come-from-behind victory by just 222 cars.” The final count was 187,343 for Cadillac, 187,121 for Lincoln. Then GM reported that Cadillac sales dropped 38% in January. How could sales be so different in December and January? Could it be that some January sales were counted in the previous year’s total? Just enough, say, to win by 222 cars? Yes, indeed. In May, GM confessed that it sold 4773 fewer Cadillacs in December than it had claimed.
190
In the General Motors examples, we suspect something is wrong because numbers don’t agree as we think they should. Here’s an example where we know something is wrong because the numbers don’t agree. This is part of an article on a cancer researcher at the Sloan-Kettering Institute, which was accused of committing the ultimate scientific sin, falsifying data.
EXAMPLE 6 Fake data
When reporting data about the Minnesota mouse experiments, the cancer researcher mentioned above reported percentages of successes for groups of 20 mice as 53, 58, 63, 46, 48, and 67. This is impossible, because any percentage of 20 must be a multiple of 5!