Thursday, October 27, 2016

A Review of CREG Forecasting

Given how important the Consensus Revenue Estimating Group's (CREG) reports are for the state's lawmakers and for our governor, and given the problems in the latest report that I pointed to the other day, it would be a good idea to take a closer look at how CREG's forecasts evolve over time.

Let me first of all, again, stress that economic forecasting is among the toughest things an economist can get involved in. Contrary to what conventional wisdom would suggest, there are no good methods for predicting where the economy is heading at any given point in time. The most common forecasting techniques, which are based on econometrics, can very well predict the future of the economy so long as nothing of any significance happens. Once the economy hits a recession, econometricians are notorious for their failed forecasts.

The main reason why econometrics is unable to predict disruptions in stable economic activity is that its models must produce "rigorous" solutions; roughly, this means that forecasts where all the relations between the variables are statistically significant. If there is no rigorous solution, the econometrician typically backs away from making a forecast. 

While this might be a good idea for the econometrician, and strictly speaking the right thing for him to do, the people who rely on economic forecasts are not being helped by the economist's confinement to rigorous forecasts. Under some circumstances it is necessary for economists to present forecasts based on other methods. 

Wyoming and its economy present such circumstances today, and have done so for at least a couple of years now. Without getting lost in the higher layers of economic theory (though for the interested audience, here is a good introduction) the main problem with predicting the Wyoming economy today is that its future is hidden behind thick layers of uncertainty. Under such circumstances, it is not a good idea to stick to forecasting techniques, or even methodologies, that apply to highly stable macroeconomic conditions. 

With this in mind, I hope to have made clear that I am not picking on CREG specifically. My concern is, rather, with the kind of forecasting techniques that are exhibited in their reports - and what I see as their unwillingness to adapt to changing economic conditions. 

Figure 1 offers a good example of this point. It reports the forecasts of General Fund revenue over four or five years, as published by CREG in their October reports since 2012. For example, the grey function represents the total revenue forecasts for the period 2015-2019 published by CREG in October 2014: 

Figure 1
Source: CREG October reports 2012-2016

There is a striking pattern in the revenue trends that CREG puts forward every year: they all suggest that revenue will increase in the future. Take a look, especially, at the difference and the similarity between the functions for the 2014 (grey) and 2015 (yellow) reports: they essentially predict the same trend in future revenue, but suddenly CREG found it necessary to lower the forecast for the first year of the 2015 forecast, compared to 2014. 

In plain English: in October 2014, CREG predicted that General Fund revenue for 2016 would be $1,145 million; in October 2015 they predicted that the revenue for 2016 would be $1,019 million. This represents a not-insignificant eleven-percent drop in forecasted revenue.

When I say that forecasting is tricky, I refer specifically to situations like this. CREG had good reasons to adjust their forecast downward from one year to the next. 

The problem - and this is a big problem - is that they continued to predict rising revenue over the next few years. But when a major change in forecasting conditions forces you to reduce your predicted number for a variable in the near future, then logic - and a nice package of economic theory - dictates that you also question the trend in your forecast. 

My question, then, to CREG would be why, despite an eleven-percent downward adjustment in predicted General Fund revenue, they chose to assume that this event would have no effect on the revenue trend in the next few years. The prudent choice would have been to remove that assumption and start from scratch, using different, less rigorous but for uncertainty more useful methods. (I have myself predicted a decline in revenue for the next 3-4 years.)  

To be fair, the October 2016 report, represented by the blue function in the bottom-right of Figure 1, is indicative of some change of heart in CREG forecasting. They deserve kudos for this, but this only reinforces my question about how they arrive at their trends; their 2016 report predicts that General Fund revenue in 2020 is going to be a whopping 21 percent lower than what they predicted in 2014. 

Consider that number once again: in 2020, one in five dollars will disappear from the General Fund compared to what CREG predicted two years ago. 

Such a dramatic decline in forecasted revenue should cause all alarm bells to go off, both at CREG and at the legislature. Hopefully, CREG will put out a special report focusing on how they have changed their revenue forecasts, what the implications are for the state budget - and whether or not it is a good idea to continue to believe that the underlying revenue trend is always going to be pointing upward. 

No comments:

Post a Comment