Rep. Paul Ryan's 2012 budget has ambitions far beyond 2012. It aims not just to set priorities for a single year, but also to wrench the country back into the black. The theory is straightforward enough: Tax cuts to wealthy Americans foster prosperity that moves millions of (less wealthy) Americans back to work, with increasing wages. High earnings and employment bolster tax revenue. When combined with huge cuts in domestic spending and radical changes to Medicaid and Medicare, the budget balances out in about 20 years.
Ryan's plan relies on economic forecasting from the Heritage Foundation's Center for Data Analysis. Unfortunately, at least for Ryan, some of its numbers have been exposed as a bit fantastical. Which raises at least two questions: How did Heritage get it wrong? And can we trust its other numbers?
There were a few improbable figures in the Heritage account of how Ryan's budget plan would affect the economy. For instance, it saw homebuilding ticking up more than 50 percent from current levels by the end of 2012—though housing is currently undergoing a double dip. There also were Pollyannaish employment projections, which the conservative think tank later adjusted. Initially, its economists forecast that without the GOP budget, unemployment would drop to 8.4 percent next year and 5.2 percent in 2021. With it, unemployment would fall to 6.4 percent in 2012 and 2.8 percent in 2021.
Immediately, economists from across the spectrum gave the accounting the smell test and called it spoiled. The 6.4 percent unemployment figure for 2012 was based on the economy adding about 1 million new jobs. But if the economy added 1 million new jobs, the unemployment rate would be about 8 percent, some protested. Or, take the claim that unemployment would hit 2.8 percent in 2021. It could,of course. The rate averaged 2.9 percent in 1953. It dropped to 1.2 percent in 1944. But short of an enormous, fate-of-the-world type ground war—heaven forbid—the Federal Reserve would never let it into the 2-percent range again. Starting in the 4 percent range, maybe higher, the Fed would tighten monetary policy to cool the economy off.
After the outcry, Heritage modified the GOP plan's unemployment numbers upward, to 7.8 percent in 2012 and 4.3 percent in 2021. "I just didn't think they were within an acceptable range," says Bill Beach, director of Heritage's Center for Data Analysis. "It was a judgment call—so, we reissued them."
But if the unemployment rate is higher, that presumably means lower income-tax revenue and higher government spending on programs like unemployment insurance. Wouldn't there be effects throughout the economy? Beach says it is an unusual bug of the model at use: The unemployment rate remains fairly independent from other variables. "The normal way of doing unemployment statistics or rates is to take the civilian labor force and find out the number of people who are working—then to calculate the unemployment rate from there. That's the way the BLS does it," he says. But the Heritage model calculates it differently.
The model derives the unemployment rate from two variables: a wage-price variable (essentially, how much workers cost) and the full-employment unemployment rate (essentially, the lowest the unemployment rate can get without spurring inflation). Beach says he felt that latter variable had been set too low. He set it higher, re-ran the whole shebang, and came out with new, higher unemployment rates—with no impact on the rest of the model. Total employment, public employment, and private employment were never affected, he says.
Mark Zandi of Moody's Analytics, an eminent economic forecaster himself, says that does not make a ton of sense to him, and other economists I talked to were similarly perplexed. "In my model, if the unemployment rate were to change at all, it would have a significant impact," says Zandi, "on wages, inflation, monetary policy, revenues, deficits, hours worked, how big the economy can be. Everything in the model is very sensitive to and depends on the unemployment rate."
At this point it may be helpful to explain a little bit about economic modeling, that part-art, part-science process so central to economic planning. Models are essentially big math puzzles: Economists put in a whole bunch of Xs and come up with a whole bunch of Ys. First, the economist inputs variables. There are the ones she knows or wants to guess at—say, tax rates over time. And there are the ones she wants the model to figure out—say, sweater sales. Then, she describes how those variables influence or relate to one another. Our economist might say that lowering taxes 5 percent increases sweater sales 10 percent, for instance. The models allow for all sorts of complicated relationships and constraints. For example, our number-cruncher might tell her model to assume that sweater sales will fall if the weather is unusually warm or if fleece sales fall into a given range or if interest rates rise above a certain level. With everything in place, the model spits out outputs—usually ranges of numbers over time.
These economic models include a lot of moving pieces. The quality of the data and a thousand other things affect how well the forecast stands up. Of particular importance is the soundness of the assumptions of the economist at the helm. If the weather has no impact on sweater sales, our economist's forecast is not going to prove very useful.