Individual Income Tax Revenue
The latest numbers from the state OMB showed some interesting information regarding taxes, again. Rather than focus on sales tax this time I thought we could branch out into income taxes. Why? Sales tax revenues get the bulk of attention in the media and from me generally and I think branching out is important. Another reason is as the impacts and effects of the oil decline transmit into other sectors more fully, and the state budget cutting starts to fully take effect, we will see other revenues exhibit declines. My interest in the overall forecasting process, if we can call it that, also means we need to branch out into other revenue areas to understand the complete picture of the revenue process. The individual income tax revenue was also almost $20 million short of the forecast in May. With that in mind here is the picture for actual versus forecast individual income tax revenues in North Dakota:
From a forecasting perspective the frequent peaks are a bit of a concern, especially when the actual is so much higher than the forecast. The consequences of the actual in such excess above the forecast might seem trivial, since it is a surplus, but in terms of the process of legislating and making spending decisions it creates a false sense of security if it goes on for too long. The peaks just speak to an inaccurate forecast process that seems unable to adjust or correct for the issue.
Rather than making a big deal about the size of this month’s deficit let’s take a different look at this by focusing on the forecast errors, the actual value less the forecast value.
This graph shows the average forecast errors by month over the last eight years. What we see is that the forecast is consistently overestimating the May value. Also of interest is that the average April error is incredibly high, with the actual very often well in excess of the forecast value. Also of note is that the average forecast is well below the actual value. This is what I mean by the persistent type of error leading to erroneous decisions on the part of lawmakers. This actually speaks to how bad the process is right now. Considering the misses made in the last two years where actual came in well below the forecast it seems that there had to be incredible misses with the actual well in excess of the forecast to offset that. It is a very easy thing to incorporate monthly effects into a statistical model to control for and correct for those types of issues. My students learn this all the time, for example, holiday effects in November or December, or maybe Valentine’s Day in February. They even learn how to account for effects like Easter crossing various months. We are not currently getting that in the North Dakota forecast process.
For completeness sake I also generated the standard deviation of the forecast errors. We can further see that the forecast process exhibits significant volatility.
Forecasts need to be accurate across all months, at least the way they are conventionally structured, and if the statistics will be reported on a monthly basis. In fact they should be structured to take into account the seasonal or monthly effects likely occurring within the area under study. Taken together the three graphs show a situation of a forecast process that hardly meets these standards and does not serve the lawmakers and ultimately the taxpayers well.
Taxing and spending decisions should take forecasts into account. However, when the forecast is inaccurate and volatile it seems almost impossible for the decisions made to be optimal, or for there to be even a high level of certitude about the correct course of action.