“The phenomenon that forecasters are trying to accurately predict has just become . . . harder to accurately predict,” says Brenna Erford, who manages state budget policy work at the Pew Charitable Trusts, which issued the report last month in conjunction with the Rockefeller Institute of Government at the State University of New York.
The mistakes are even more dramatic around recessions: The forecasting errors around the past two, in the early and late 2000s, were more than triple the size of the errors surrounding the recession in the early 1990s, they found.
Increasingly volatile revenue is to blame, the authors argue, and it’s driven by myriad causes, including a growing reliance on an interconnected global economy, Erford says. In Hawaii, for example, state officials closely monitor the value of the yen because Japanese tourism is a key market for the state. In Nevada, officials keep an eye on Asia, whose high-rolling baccarat players can affect sales and gambling revenue.
The “Chinese New Year goes into my [forecasting] model and that’s impacted by the health of the Asian economy,” Janet Rogers, the state’s chief state economist, told the report’s authors.
The report itself focused on the largest sources of revenue: corporate income, personal income and sales taxes. The median error rate for corporate income taxes over the 27-year period was largest, at 2.8 percent, while it was 1.8 percent for personal income taxes and 0.3 percent for sales taxes.
Rising income inequality may be making things worse
When it comes to volatility in personal income taxes, decades of rising income inequality may be to blame. At least, that’s a hypothesis put forth by economists at Standard & Poor’s, a credit rating agency, in a report last fall.
Not only has it depressed economic growth, but as income has pooled at the top, states have consequently become more reliant on top earners for revenue. Unfortunately, those high-income earners are different from the rest. Unlike others whose incomes rely more on wages, top-earners’ incomes tend to come more on capital gains, which are, you guessed it, more volatile.
Although the Pew report doesn’t go so far as to blame income inequality, it does credit capital gains as the most volatile component of personal income taxes — a problem that California, it notes, has had to reckon with recently:
“California’s reliance on capital gains revenue for its general fund and on taxes paid by a small, wealthy portion of its population led Governor Jerry Brown to propose tying the state’s rainy day fund to volatile capital gains revenue, which California voters approved Nov. 4,” the Pew and Rockefeller authors note.
There are other sources of error, too
Sales taxes have traditionally been among the easiest to forecast, but the recession complicated even those predictions. Typically, consumers are consistent in their purchases of things such as food and toiletries, even during a downturn. “But consumers dramatically curtailed spending even on staples during the recession, causing the sharpest drop in sales tax receipts in 50 years,” the report’s authors write.
Having a small population or relying on relatively few economic sectors can make it harder to forecast revenue, and both affect North Dakota, which had the largest forecast errors of any state. Fortunately for North Dakota, the error has underestimated revenue and not the other way around.
But even that has its downside. How do you determine the size of tax cuts or spending increases if your budgetary windfall came as a surprise? If revenue came in above estimates one year, how can you trust your estimates for the next?
So what can states do?
The report’s authors argue that states should do a better job of building reserve funds. One way to do that is to harness the volatility itself to fund those rainy-day backups, as 13 states already do in some form. Alaska, California, Louisiana, Massachusetts and Texas, for example, each use revenue from especially volatile taxes to bolster their reserve funds.
Another way to avoid errors is to reduce the lag time between when the forecast is issued and the period it covers. And another suggestion may seem obvious, although it is in practice difficult to do well: regularly test and alter forecasting models.