Uncle Sam's Stats: Call Them Unreliable

In late June, the U.S. government reported that the trade deficit for April jumped 20%, to $8.4 billion. Predictably, the world's currency markets went nuts. Moments after the figure moved on the financial wires, the dollar collapsed and U.S. bond prices plummeted.

On July 13, another bit of official econo-data, the consumer price index, will draw its own keen following. Government retirees and Social Security recipients watch the CPI because their annual cost-of-living adjustments are based on the index. Politicians latch on to the number as they debate the need for higher interest rates.

There's just one problem: The trade-deficit is almost surely overstated. So is the CPI. That means pensioners are getting an artificial sweetener in their COLAs. And all those bond traders are betting on flawed numbers from statisticians two steps behind the economy.

In a classic example of shortsightedness, Uncle Sam has been starving his statistical agencies for more than a decade, denying funding to improve data collection and analysis. As a result, the numbers that move financial markets and policymakers are misleading at best. "The economy doesn't stand still," says former Commissioner of Labor Statistics Janet L. Norwood. "We need a statistical system that can change with it."

HIGH-TECH TANGLE. The trade deficit is a prime example. Government statistics show U.S. exports falling far short of imports. But privately, government economists believe they are underestimating the value of U.S. exports by 10% or so. If true, much of the shortfall vanishes.

One reason for the low estimate: Computer software, an industry dominated by U.S. producers, barely shows up in the reams of government trade numbers. According to the Commerce Dept., which is supposed to track such things, software exports last year were $385 million. No one knows for sure what the true number is, but $385 million isn't close: Microsoft Corp.'s international sales alone were five times that last year. That suggests the government may be undercounting as much as $4 billion in software exports--4% of the total trade shortfall.

How did the data get so bollixed up? The government calculates only the value of instruction manuals and the blank disks upon which the programs are written. The value of the software itself simply disappears. So a program that sells for $300 registers on trade books at $5 or $10.

It's the same with inflation, now officially at about 3%. The CPI may overstate price hikes by 0.5 to 1.5 percentage points. Why? Because it has been a decade since the index received a complete scrubbing, and it no longer reflects real buying patterns.

The CPI still assumes, for instance, that personal computers represent less than 1% of household purchases--too low now. It also fails to reflect the difference between doctors' charges and actual, lower reimbursements from insurers. More critical, the CPI doesn't accurately gauge the inflation-dampening effect of discounters such as Wal-Mart. "Everybody knows the CPI overstates inflation," says Kemper Financial Services Inc. economist John E. Silvia.

NEW BASKET. At least the CPI is going to get fixed. Congress is about to give the Labor Dept. $50 million--mostly to recalculate the "market basket" that people actually buy. But not because of newfound concern for accuracy. Instead, lawmakers see it as an easy way to cut cost-of-living hikes for retirees. The Congressional Budget Office figures that a 0.5% cut in the CPI would save a cool $25 billion over the next five years. "If you really want to cut the deficit," says one senior official, "the best investment you can make is to get more money into the Bureau of Labor Statistics."

Spending money to update inflation indexes and properly measure trade balances would surely meet the test of any cost-benefit analysis. But breaking down the inefficient fiefdoms that dominate the government data business would also help. For instance, raw trade data are now collected by one agency but massaged and published by two others. It's no surprise the numbers are a mess.

At a minimum, this overlap should be eliminated. And it's time to go one step further. Canada has a single statistical agency, a model of effective data collection and analysis. Its inflation market-basket is studiously revised every four or five years. The U.S. could do worse than copying that system. Done properly, it would save money and more accurately gauge the economy. That would help bring bond traders, pensioners, and politicians back to reality.

Before it's here, it's on the Bloomberg Terminal.