Policy staff in least 21 states -- and possibly many more -- have experimented with dynamic scoring since the early 1990s. While many states regularly use dynamic models to assess the economic impact of infrastructure investments, almost all state-level efforts to dynamically score tax policies have been abandoned. The primary culprits: wildly unrealistic expectations of revenue changes and serious problems using a highly imprecise policy tool in a balanced-budget environment.
Dynamic scoring is largely concerned with the economic ripple effects from a tax change. Dynamic effects are often compared to the more traditional static revenue estimates, which typically measure only the direct effects of a tax change, although selected behavioral effects (or elasticities) may be incorporated. With a typical static estimate, a $100 million tax cut would be expected to reduce revenues by the same amount -- $100 million. A dynamic estimate, on the other hand, attempts to account for economic growth (or decline) associated with reduced (or increased) taxes, so a $100 million tax cut that is dynamically scored might only reduce revenues by $90 million - a 10 percent dynamic effect.
In a recent joint report by Georgia State University's Center for State and Local Finance and the Fiscal Research Center, experts looked in detail at the range of dynamic effects in seven states that reported the results of dynamic modeling. States reported dynamic effects ranging from 1 to 20 percent of the static revenue estimate.
So why, despite all of this interest and some apparently large dynamic effects, have almost all dynamic-scoring efforts at the state level been discontinued? The problem is twofold: First, even acknowledging economic-growth effects of tax cuts, the size of a dynamic effect -- even a large one -- is typically miniscule relative to the overall size of a state's budget. And second, given the complexity of these estimates and timing issues (when exactly does the dynamic effect occur?), the actual dynamic estimates are too imprecise and too uncertain to be built into a state's budget in any meaningful way.
Kansas' recent experiment with dynamic scoring is a case in point. In 2012, Kansas adopted major reductions in its income tax. In fiscal year 2015, the state economist's static estimate forecast that revenues from the 2012 tax changes would decline from $6.466 billion to $5.642 billion (an $824 million loss or 13 percent). However, a dynamic analysis from a pro-tax-cut research institute predicted that the state would actually lose only $714 million. That analysis forecast that $110 million, or 13.5 percent, in additional revenues would be recovered through dynamic effects.
The problem for the state is that even with such a large tax cut and a large estimated dynamic effect, the actual value of the dynamic effect would be only 2 percent of general-fund revenues. While dynamic models do not generate a margin-of-error estimate for dynamic effects, research has shown that traditional revenue estimates carry an error rate of around 3 percent.
In the Kansas case, the state's revenues actually trended below the static revenue estimates. Whether the dynamic effects failed to materialize, were incorrectly estimated or were simply lost in the normal error rate around a traditional revenue estimate is an open question and one that is unlikely ever to be conclusively resolved. Meanwhile in the past few weeks, Kansas legislators have been painfully struggling to close a $400 million budget gap. Despite the hardship, Kansas has been better served by sticking with the more conservative static estimate -- the dynamic estimate would have left the state with an even larger gap to fill.
None of this is to suggest that dynamic scoring can't be useful in comparing economic tradeoffs among different tax-policy choices or even among different tax and expenditure mixes. States such as Nebraska that have approached dynamic models in this spirit -- using the model as a source of information rather than a budget tool -- have had a much more successful and productive experience. These models can also be customized to show the distributional effects of tax-policy changes across income classes or industry types.
In the end though, the overwhelming conclusion of staff across the states was that policymakers found dynamic scoring disappointing. Not only did tax cuts not pay for themselves (and no credible model shows that they do), but even very aggressive estimates of dynamic effects were too small and too uncertain to be a source of easy money that could be used to fill a budget gap.
