Predictive Analytics and the Fiscal Cliff

The U.S. economy is in poor shape, with higher-than-desirable unemployment rates and lower-than-desired GDP growth rates. Meanwhile, federal expenses continue to exceed revenues (primarily taxes). While the size of the national debt may be worrisome, it is currently not a major crisis. Historically low interest rates have mitigated the interest rate burden, but an unpredictable (and entirely possible) event could lead to a sudden interest-rate spike that would create another economic crisis.

On Jan. 1, 2013, the Bush era tax cuts are scheduled to end, and tax rates will return to what they were before those cuts were put in place. Moreover, mandated cuts in federal spending would go into effect. While the combination of higher revenues and lower government expenses would effectively raise revenues and reduce the growth rate of the national debt, the worry is that higher tax rates would encourage businesses to shed workers, raising the unemployment rate and pushing the U.S. economy back into recession. Despite the political jockeying for position currently going on in Washington, virtually no one wants to go over the fiscal cliff. Predictive analytics may help.

Using Predictive Analysis in Economic Modeling

Predictive analysis is a hot topic in advanced analytics technologies, especially with the rise of interest in big data. Predicting the future is useful, as it can be an aid (but only an aid) to decision making. Now, predictive analysis may be used to analyze actual data (such as in a trend analysis comparing a current situation to historical data) or to investigate proposed policy actions (such as an increase in tax revenues) to predict what impacts they might have.

The Congressional Budget Office (CBO) has long used two economic models to analyze the medium- and long-term impacts of federal tax-and-spend policies. The Solow-type growth model estimates the impact of proposed policy changes, such as the example given on tax increases and labor supply changes at a given point in time. In contrast, the life-cycle model estimates the impact on the supply of labor that depends on people's expectations on how they expect their after-tax compensation to change over time.

Despite the promotion of analytics as a useful tool, all models have their limits.

• Models are only as good as their assumptions. That may seem trite, but economies (like businesses) are not static, and the value of some parameters may change or may have more significance than anticipated. Virtual reality is not physical reality, so surprises are not only possible but likely, especially over long time frames.

• Models can only deal with endogenous conditions, which means those that they presume respond to the changes made; they cannot respond to exogenous conditions, which arise outside the ability of the model to predict but can nevertheless affect the model's results. One example would be the spillover effects of, say, a European recession on U.S. economic output.

Why Use Models?

While models have limitations, a well-crafted economic model is much better than a manual analysis, which cannot consider the impact of many variables and is often the product of wishful thinking rather than objective analysis. For example, a simpler (although still quite complex) business model that I built for a division of a Fortune 500 company many years ago enabled senior management to make decisions over a period of years that increased revenues, changed pricing policies, cut costs and increased the return on capital investments. The model sometimes provided results that seemed at first to be counter-intuitive, but that improved overall understanding of what could and could not be done.

The output of economic models is not as easy to accept as business models, which may give results (such as increased revenues) that everyone finds acceptable. Instead, the outputs from every policy action may have some positive benefit, but also have an unpalatable side effect. Those tradeoffs are likely to be painful to someone. However, the models can serve as a common point of reference and discussion.

Now, there may not be better economic models than those used by the CBO, but the CBO has the imprimatur of the Congress itself and is a non-partisan organization. The CBO is generally considered to be well-run and independent, although in a political domain, there will always be critics.

Next page: How Can Models Help?The responsibility for action on taxation is a joint responsibility of Congress and the president. The challenge they face today is not only to come up with a short-term fix, but also a long-term solution. In the short term, failure to come to agreement could result in a self-fulfilling prophecy of a recession if businesses think that a recession is coming and make cutbacks (such as number of employees or budget cuts) that really aren't necessary.

The fix applies only to short-term issues and does not address long-term concerns, and it could induce paralysis among enterprises that are unsure how they should act, leading them to take no action to expand (including employment) as much as they should.

The 17th century French economist Jean Baptiste Colbert is credited with the maxim, "The art of taxation consists in so plucking the goose as to obtain the largest amount of feathers with the smallest amount of hissing." That statement probably applies not only to taxation, but also to budget or expense reduction. One result would be to ensure that actions are staggered incrementally over time to prevent any response of the economic system that has overly negative consequences (and prevent much "hissing," though each impact in and of itself does not represent a tipping point).

Analytical models can also serve as an independent reference point to display what the expected impact of policy decisions might be. Although they cannot literally force decisions, they can serve as a focal point for discussion and negotiation. By no means are they a panacea or predictors of a certain future, but they are the best that we can do.

Mesabi Musings

Recall the wisdom of Dr. Samuel Johnson: "Depend upon it, sir. When a man knows he is to be hanged in a fortnight, it concentrates his mind wonderfully." Hopefully, such concentration will result in policy decisions that will prevent having to deal with a crisis that could and should be avoided.

Also, the use of predictive analysis tools, such as the CBO computer-based economic models, will--we hope--help the decision makers concentrate their thinking and lead to decisions that, unlike a condemned man unable to avoid his fate, will allow all of us to avoid being hung out to dry. Let's hope that the politicians turn into statesmen who recognize that no solution they implement will be perfect, but that the overall benefits should more than make up for any defects.

And that raises a challenge to IT in general and IT vendors in particular. Using predictive analytics to increase transparency to proposed governmental policy decisions should be an important point, and that will become an ongoing goal.

So questions arise: What other solutions, in addition to economic modeling tools viewed as independent and unbiased, should be brought into the discussion? Are there any advanced analytics tools that are not currently being employed that could be used? Can big data play a role, with its ability to analyze masses of data from multiple sources? What can be done not to predict the exact nature of an unexpected shock to the economic system, but what its impacts could be? What politically independent organizations, including IT vendors, can ably and effectively assist in these efforts?

All of us (literally, not only in the United States) have a big stake in the promotion of sound economic policies. We know our government leaders have the tools. Let's hope they use them well.