States will decide in the coming weeks how to spend roughly three-quarters of a trillion dollars from their taxes on personal income, sales and other activities. It’s a big number. That means even relatively small errors in revenue forecasts can make a substantial difference for state and local government budgets, and the services they support.

In the wake of the Great Recession, we’ve seen how miscalculations in revenue projections can force states to make painful choices. A new report by the Rockefeller Institute and the Pew Center on the States finds that in fiscal 2008-09, the typical state’s tax collections fell 10 percent or more below projections. The results: a total of $49 billion in unexpected revenue shortfalls and midyear cuts to education, health care and other services. Why do such errors occur, and what might states do about them?

Given the economic environment in 2008 and 2009, it’s no surprise that revenue analysts did not foresee how sharply tax collections would decline. The recession that began in December 2007 was the worst since the Great Depression. The nation lost more than 8 million private-sector jobs, and capital gains fell by 60 percent — each contributing directly to precipitous drops in personal-income tax receipts. Retail sales also fell sharply, driving steep drops in state sales-tax revenues.

It’s important to recognize, though, that errors in revenue forecasts can occur even in less turbulent periods. Such gaps between projected and actual revenues have been growing larger and more frequent — thus creating bigger problems for states — over the last two decades.

The Pew-Rockefeller Institute study examined 23 years of tax collections, comparing actual receipts to the estimates that states used when adopting annual budgets. States erred on the low side in 16 of those years, leading to higher-than-projected revenues and budget surpluses. In itself, such extra revenue is a fine thing — but policymakers sometimes use unexpected resources to make lasting commitments that are hard to sustain.

In the other seven years we studied, states’ overall revenue estimates were too high, so that policymakers were required to deal with unexpected shortfalls. These negative surprises, too, have become larger and more troubling over time. During the three main years of the 1990-92 fiscal crisis for states, revenue was at least 5 percent below forecast in 25 percent of states. During the comparable years of the downturn that began in 2001, 45 percent of forecasts were off by at least that amount. And in 2009 — admittedly the most-difficult-to-forecast year of the current crisis — fully 70 percent of all projections missed the mark by similar proportions, while half of the states had truly big problems with collections coming in more than 10 percent below forecast levels.

In good times or bad, projecting tax revenues is not easy. State officials have to start by assessing key trends in the national economy such as employment, personal income, retail sales and corporate profits. No one predicts those things with total accuracy. States must also take account of issues unique to their regional economies and tax bases.

But if revenue forecasts are prone to more and larger errors than was the case in an earlier era, something must have changed. Indeed, that’s the case: The relationship between state tax systems and the economic activities to which they apply has evolved significantly over the past few decades.

In 1978, the personal income tax represented just over one-quarter of total state tax revenue. In 2008, it was above 35 percent. During that time, the nature of the income that is typically subject to state tax has evolved. Decades ago, taxable income was overwhelmingly dominated by salaries and wages. These may decline during recessions, but usually by relatively modest amounts. Although such regular income is still the largest share of the taxable base, capital gains and bonuses play a much more important role now.

These subsets of income rise and fall much more sharply over the course of the business cycle than does income from salaries and regular wages. Over the half-century ending in 2010, capital gains represented a median 2.8 percent of the nation’s gross domestic product. In boom years for the financial sector, such as 2000 and 2007, that proportion has more than doubled, reaching 6.5 percent or more. From a short-term perspective, this was good news for states that rely heavily on income taxes because the spike in capital gains drove collections higher. But when such gains fell back to earth, so did the revenues based on that income. Meanwhile, the unexpected jump in tax collections had encouraged many states to expand programs or cut taxes – steps that would prove unsustainable within just a year or two.

The Rockefeller Institute included an extensive discussion of capital gains and state tax revenue, and the extent to which they vary across states, in the April 2010 State Revenue Report, beginning on page 20. The table below shows states that depend heavily on income taxes and capital gains; a more complete version appears in the April 2010 report.

Some states rely heavily on severance taxes based on natural gas, oil and mineral production. This represents an especially significant problem when energy markets weaken. At such times, the state may lose revenue from both the taxes imposed directly on energy firms and those on workers’ wages and consumer sales — with both driven largely by the same industry trends.

For example, in July 2008, the prices of oil and natural gas both reached record high levels. While states such as New York and Connecticut were suffering from the crisis in the financial sector later that year, Oklahoma was enjoying a flood of revenue. Collections came in nearly $200 million higher than expected. But by late December 2008, the prices of both oil and natural gas had fallen by more than half. That set the stage for 18 months of the worst revenue shortfall in Oklahoma’s history. Montana, Alaska, Colorado, Louisiana, Texas, West Virginia and Wyoming experienced similar troubles.

What can state policymakers do to guard against damage from budget shortfalls that are driven by errors in revenue estimating? We will address that question in a second Observations, to be posted shortly.