2020 Primary Tracker Version 2.0: Adding Campaign Finance Dollars
Now that 2019 First Quarter fundraising reporting is in the books, I wanted to add campaign finance data to the Primary Tracker model. At this stage in the campaign, fundraising is probably the single most important measure of a candidate's strength and ability to endure in the primary race. Candidates who haven't raised at least a few million dollars at this stage are in danger of running on fumes before we get through the summer debate season.
The most important fundraising metric is of course total dollars raised, so that's obviously going into the model. Beyond that, I wanted to capture a measure of a candidate's breadth of support. I was planning on using the number of individual donors for each candidate, but that number is not included in the FEC first quarter reporting summaries. The FEC filings do break out donations into some useful categories: total dollars from small individual donations (up to $200), total dollars from large individual donations (>$200), PAC money, and candidates' personal contributions. The small donations can serve as a surrogate for broad support in this case.
Now to put the fundraising on the same 10-point scale as the rest of the model's data inputs. As described in a previous post on model scoring, each data point is divided by a theoretical, or in some cases feasible, maximum, and multiplied by 10, so that 10 represents the upper limit a candidate can reasonably expect to reach for each metric. That means I need a practical maximum for fundraising dollars. I took a quick look at quarterly fundraising numbers from the 2016 primaries, and the highest number I found was about $36 million for Hillary Clinton toward the end of the primary season. So I'm going with a quarterly maximum of $40 million to compare candidates' numbers against. As for small contribution dollars, a quick review suggests that the highest number likely to be achieved is 75% of total dollars raised. So my maximum there will be $30 million for quarterly numbers. Finally, I wanted to weight total dollars raised considerably higher than small contributions. We want to see breadth, but depth of the coffer is ultimately most important. So I averaged the two scores with total dollars counting for two-thirds of the average, and small dollars one-third.
One final word about incorporating fundraising dollars. Only official candidates have reported to the FEC, so we don't have data for Biden, Abrams, or other potential candidates. In that case, I'm simply leaving the fundraising metric out of their input score averaging. That keeps them on the same scale as the others.
While I was updating the model, one more change was made. As noted, each score is based on a theoretical/feasible maximum value which is calibrated to 10. In most cases the candidates are far from reaching that maximum at this early stage in the campaign. But that's not true in one case, due to an ordinal ranking problem. I'm referring to my scoring of pundit power rankings. To illustrate, here are the maximum values for each data input, and the highest value any candidate is currently sitting at:
As we can see, this early in the campaign candidates are far from reach the maximum in most metrics. But Pundit Rankings are ordinal, meaning someone has to be ranked first, and if that candidate is at or near the top in all pundit rankings, they will already be near the maximum for that metric. Is that a problem? Not necessarily, not if it accurately reflects the distance a candidate has to cover before they've locked up the nomination. But being ranked first by the pundits is not a sufficient condition for the nomination, as the other metric maxima are intended to represent. The practical problem with all this is that pundit rankings are effectively overweighted in the total model score.
So, I've inserted an adjustment to the 30 point maximum for these pundit rankings. The maximum now starts at 60 points on February 1, 2019, a year before primary season begins. That maximum will decrease incrementally each day until it hits 30 points on February 1, 2020. That puts the pundit ranking score on a more level playing field with the other metrics. I've also applied this retroactively to the rankings since the beginning of March.
So, I can now present version 2.0 in the Tracker graph, this shows candidate trends with adjusted pundit scores, and the fresh new fundraising numbers:
I'll point out three things here. First, you may notice that most candidate scores are now lower than they were in the previous model version. Biden was cruising along at around 4.3 for example. That reduction is due to the change to pundit ranking scores. Second, you'll see several big jumps for the most recent scores representing today (April 16, 2019). That's due to the addition of fundraising data. Candidates like Sanders, Warren, Gillibrand, and Delaney simply have performed better in first quarter fundraising (relative to maximum remember) than they've been doing on most other metrics. Third, you're now seeing Stacy Abrams on the graph. She's not officially running, but she is now showing up in two of the Tracker model's metrics: an Iowa poll, and the Washington Post power rankings. That earns her a spot here at Hats in the Ring. We're still waiting on the likes of Ryan, Schultz, and others to appear in two data inputs.
Finally, here's the current Tracker scoring:
So does the fundraising data help us separate the pretenders from the contenders? Possibly. I think Castro, Yang, Hickenlooper, and Inslee have a limited time to make their case to voters and start rising in the polls, before the funding will run out. They simply won't be able to sustain the level of infrastructure required in a modern campaign unless they gain more support soon. Also note that Delaney's cash is impressive, but not so much when we see that $11.7 million of it comes from Delaney's own personal wealth. That keeps him in the game for now but is not sustainable.
The most important fundraising metric is of course total dollars raised, so that's obviously going into the model. Beyond that, I wanted to capture a measure of a candidate's breadth of support. I was planning on using the number of individual donors for each candidate, but that number is not included in the FEC first quarter reporting summaries. The FEC filings do break out donations into some useful categories: total dollars from small individual donations (up to $200), total dollars from large individual donations (>$200), PAC money, and candidates' personal contributions. The small donations can serve as a surrogate for broad support in this case.
Now to put the fundraising on the same 10-point scale as the rest of the model's data inputs. As described in a previous post on model scoring, each data point is divided by a theoretical, or in some cases feasible, maximum, and multiplied by 10, so that 10 represents the upper limit a candidate can reasonably expect to reach for each metric. That means I need a practical maximum for fundraising dollars. I took a quick look at quarterly fundraising numbers from the 2016 primaries, and the highest number I found was about $36 million for Hillary Clinton toward the end of the primary season. So I'm going with a quarterly maximum of $40 million to compare candidates' numbers against. As for small contribution dollars, a quick review suggests that the highest number likely to be achieved is 75% of total dollars raised. So my maximum there will be $30 million for quarterly numbers. Finally, I wanted to weight total dollars raised considerably higher than small contributions. We want to see breadth, but depth of the coffer is ultimately most important. So I averaged the two scores with total dollars counting for two-thirds of the average, and small dollars one-third.
One final word about incorporating fundraising dollars. Only official candidates have reported to the FEC, so we don't have data for Biden, Abrams, or other potential candidates. In that case, I'm simply leaving the fundraising metric out of their input score averaging. That keeps them on the same scale as the others.
While I was updating the model, one more change was made. As noted, each score is based on a theoretical/feasible maximum value which is calibrated to 10. In most cases the candidates are far from reaching that maximum at this early stage in the campaign. But that's not true in one case, due to an ordinal ranking problem. I'm referring to my scoring of pundit power rankings. To illustrate, here are the maximum values for each data input, and the highest value any candidate is currently sitting at:
As we can see, this early in the campaign candidates are far from reach the maximum in most metrics. But Pundit Rankings are ordinal, meaning someone has to be ranked first, and if that candidate is at or near the top in all pundit rankings, they will already be near the maximum for that metric. Is that a problem? Not necessarily, not if it accurately reflects the distance a candidate has to cover before they've locked up the nomination. But being ranked first by the pundits is not a sufficient condition for the nomination, as the other metric maxima are intended to represent. The practical problem with all this is that pundit rankings are effectively overweighted in the total model score.
So, I've inserted an adjustment to the 30 point maximum for these pundit rankings. The maximum now starts at 60 points on February 1, 2019, a year before primary season begins. That maximum will decrease incrementally each day until it hits 30 points on February 1, 2020. That puts the pundit ranking score on a more level playing field with the other metrics. I've also applied this retroactively to the rankings since the beginning of March.
So, I can now present version 2.0 in the Tracker graph, this shows candidate trends with adjusted pundit scores, and the fresh new fundraising numbers:
I'll point out three things here. First, you may notice that most candidate scores are now lower than they were in the previous model version. Biden was cruising along at around 4.3 for example. That reduction is due to the change to pundit ranking scores. Second, you'll see several big jumps for the most recent scores representing today (April 16, 2019). That's due to the addition of fundraising data. Candidates like Sanders, Warren, Gillibrand, and Delaney simply have performed better in first quarter fundraising (relative to maximum remember) than they've been doing on most other metrics. Third, you're now seeing Stacy Abrams on the graph. She's not officially running, but she is now showing up in two of the Tracker model's metrics: an Iowa poll, and the Washington Post power rankings. That earns her a spot here at Hats in the Ring. We're still waiting on the likes of Ryan, Schultz, and others to appear in two data inputs.
Finally, here's the current Tracker scoring:
So does the fundraising data help us separate the pretenders from the contenders? Possibly. I think Castro, Yang, Hickenlooper, and Inslee have a limited time to make their case to voters and start rising in the polls, before the funding will run out. They simply won't be able to sustain the level of infrastructure required in a modern campaign unless they gain more support soon. Also note that Delaney's cash is impressive, but not so much when we see that $11.7 million of it comes from Delaney's own personal wealth. That keeps him in the game for now but is not sustainable.
Comments
Post a Comment