May 192020

COVID-19 has cost a lot of people, governments, businesses, and non-profits.   It has also broken down a lot of mental barriers for things that just “couldn’t be done” and slaughtered a lot of sacred cows.  And what cows are more sacred than the ivory towers of academia?   What is going to happen to education, especially higher education, because of our unfriendly visitor from Wuhan?  Find out in Eric Hiller’s latest article where he takes a hard self-reflecting look at if his own undergraduate degree could have been done just as well (or at least well-enough to justify the massive savings) online!   It’s just been published at Marketwatch.

Click here to read:  Eric Hiller examines Covid 19 effect on higher education







Mar 192020

There is an old saying that ignorance is bliss, and perhaps you’re one of the people in America that is blissfully ignorant of politics. If you are, God bless you, although, I suppose you shouldn’t be voting. However, if you are interested in politics, you may be wondering what exactly happened on Super Tuesday; I know I am.   Senator Bernie Sanders seemed to be chugging forward under a full head of steam, whereas former Vice President Joe Biden had only just gained a foothold through his first big victory in South Carolina.

However, after the dust cleared, Biden was in the lead of the total delegate count, although not by an overwhelming amount through his own earned pledged delegate.   Nonetheless, it “felt” that he had simply staged an amazing comeback and that Sanders might now be reeling on the ropes. This left me asking, do the data support this perception?

I decided to take a look at the numbers.   What I found, may surprise the reader, but somewhat makes sense once you let it sink in. It turns out that both candidates did better than expected.  Biden’s outperformance was simply larger than Sanders. How did I arrive at this conclusion, let’s look!

Biden outperformed the polls (especially polls over a week old)

I had been watching the polls develop over the last few weeks before Super Tuesday at  Using this data, I decided that the relevant period to look at would be after the Nevada primary until super Tuesday. One might ask, why not after South Carolina? That is a valid question and as we will see, it does matter. However, there just was not a lot of time to get new polls between South Carolina and Super Tuesday, so I extended the period a bit.  My hypothesis was that changes right before super Tuesday (e.g. Biden greatly outperforming in SC, most other neo-liberal competition dropping out) would have been too late to have enormous effects.

Next, I graphed the result in that State for each of the two candidates, versus what the polls had predicted.    Figure 1 shows the result for Biden. Note that the states are sorted on the graph from left to right by the which state awarded the most delegates. The solid orange square represents the actual percentage of the vote that Biden received in each state.   The markers in blue show up to the last six polls which polling periods closing substantially after the Nevada primary, but before the voting on Super Tuesday.  (Note that I have use the intensity of the color, marker size, and marker shape to indicate how near the pole was to the Super Tuesday Primary.  For example, the dark blue big open square is the most recent poll to the primary for each state.).  The grey footballs below the graph show the intermediate results of the analysis. For each state, I subtracted the actual vote percentage the candidate received from the average, the median, or the most recent of that State’s polls.

Figure 1 (Click to enlarge!)

So, how did Biden do? Well, in most states he over-performed polling expectation. However, he tended to over perform a lot more in states with fewer polls.   We would naturally expect statistically that the more polls over a longer time period that a state has, the wider the range in these polls would probably be. However, we don’t necessarily see that pattern consistently. For example, California’s six polls have a pretty tight range.  Compare this to Utah, which only had two polls, bounding an even wider range (see also CO and MA which have less polls, but wide ranges).  We also see, not surprisingly, that the most recent poll (except ME) was the most accurate to the performance.  In general, the over or under performance to the most recent poll is smaller for the states that had more delegates, and therefore, got more attention.

Sanders slightly beat recent poll expectations in States with lots of delegates, but under performed smaller States

Let’s consider senator Sanders and his performance as well.  As indicated earlier, people have the feeling that Senator Sanders under performed expectations. There is some truth to that in the data. However, most of the under performance was versus older polls or versus polls in States with lower numbers of delegates. This is why there is a greater amount of under performance to the average poll.

Figure 2 (Click to enlarge!)

However, if we look at the most recent polls in four out of five of the biggest delegate states, Sanders beat expectations, including Massachusetts.   Considering California and Texas represent almost 50% of the delegates that were awarded in super Tuesday and included in this analysis (Figure 3), Sanders actually overperformed the expectation of the last poll before the election, and in the case of Texas only slightly underperformed the second to last poll.   Interestingly, the biggest underperformance of Sanders was in his own home state of Vermont! However, what really hurt Sanders was his big underperformance to earlier polls in California.

California and Texas dominate delegates, but the rest of the field has a long tail

It matters which states have the most delegates. As we can see on Figure 3, like many things in nature, the political power of the Super Tuesday States follows a fairly steep Pareto trend. California and Texas hold almost 50% of the delegates by themselves. Obviously, they are less powerful when taken in context with all States in the nation.  The fact they are included in super Tuesday is a choice, perhaps an engineered political choice.  But we can also see a fairly long tail of delegate drop-off, after you get past Texas (approximately half of the states in super Tuesday represent 81% of the votes).

Figure 3 (Click to enlarge!)

On a weighted-average basis by number of delegates, both Biden and Sanders overperformed

Given the Pareto, in order for us to understand the overall meaning of this analysis, it’s important that we use a weighted average by delegates to interpret the results. Figure 4 shows a graph with the weighted average of the under or over performance for each candidate versus the average, median, and recent poll.   Because people were so negative in polls on Biden, up until the last few days before Super Tuesday, this gives Biden the appearance of over performance (almost 10% considering the weighted average of each states poll average). This is not dissimilar to non-weighted average or median of all state averages.   Biden’s almost 10% overperformance to expectations would be much higher, if we excluded the most recent two or three polls. However, because polling organizations will tend to wait until close to a state’s primary to poll them if they only want to expend resources for one poll, Biden’s overperformance from a median point of view was much more modest.  However, the ex-Vice President was able to overperform where he needed to, so that on a weighted average basis he did a lot better than the median, over performing bye 3.5% versus the most recent poll.

Figure 4 (Click to enlarge!)


Senator Sanders also outperformed on a weighted average basis, regardless of whether you weight average the average, median, or most recent for each state’s individual over performance. In other words, by the time that Super Tuesday happened, the polls did somewhat accurately correct expectations, at least in the states that really mattered.   However, it is likely the older poll expectations were still in many people’s minds.

Without the weighted average, it does appear that Sanders missed expectations by a little bit. Once again, if we look back to Figure 2, Sander’s underperformance was worse versus older polls and in states where there were few polls (many Southern and New England states).   This probably led to the perception that he had performed badly overall, although this is not borne out by the weighted average results.

Biden is in a commanding position, including his gifted delegates from others, but 2/3 of the game remains to be played

Every main player at the poker table folded his or her hand before or right after Super Tuesday, except Sanders and Biden.  Everyone except Warren openly tried to give their cards and chips to Biden.    Figure 5 shows the delegate counts after Super Tuesday as of press time for this article. The first thing we note is that most of the delegates pledged delegates were indeed earned by the two candidates we are discussing. This is shown in dark blue.   However, what happens to the delegates from the candidates who have dropped out, which at this point is everyone but Tulsi Gabbard. According to the article listed in the sources, candidates are allowed to choose what person is a delegate. However, they cannot officially control who delegates vote for. On the other hand, according to the DNC rules, the delegate is supposed to agree to vote for whoever they, in good conscience, believe their candidate wanted them to vote for.

Figure 5 (Click to enlarge!)

Given that, I assumed that all the delegates for Bloomberg, Buttigieg, and Klobuchar will vote for Biden. Not only did these candidates endorse the Ex-VP, but the candidates themselves are much more ideologically similar to Biden than Sanders. On the other hand, Elizabeth Warren dropped out endorsing no one.  I assumed that her candidates delegates would go to Biden and Sanders in a ratio of 36% versus 46%, based on the article listed in the sources. That leaves us with the two delegates held chosen by Gabbard.   I made the bold assumption that Sanders will get.  And, just because he seems to need the help, I assigned Sanders those candidates.

Even before we reapportion the other candidates’ delegates, Biden has a definite, but not insurmountable lead over Sanders (6%). Add in the delegates from the candidates who dropped out, and the gap widens to something that is starting to look big (12%). But once again, it is still fairly early.  There is a total of 3,979 pledged delegates to be allocated based on the primary and caucus process. So far, 34% of those have been claimed. Leaving 66% still to be earned by the candidates. In general, one would think that as candidates drop out there will be voters who will become uninvolved, because there are less change the remaining candidates matches them ideologically.  In that scenario, perhaps Senator Sanders supposed passionate base could help him.   Now that the other neo-liberal candidates are gone, some of their voters may simply not vote at all.  

How is the prediction playing out?

As of press time (March 20, 2020), Figure 6 shows the new delegate counts.  The model in Figure 5 underpredicted Bidens power.  He now has 57% of the pledged delegates, not including picking up other candidates delegates, and if the predictions above on delegate transference hold true, he now has 59% of the votes.

Figure 6 (Click to enlarge!)

Biden needs to not mess up; Sanders needs to super-activate his activists, but they will be hampered by other events

We can synthesize several strategic conclusions from the analysis of Super Tuesday’s results and the subsequent primaries in Figure 7.

First, the neo-liberal wing of the party is definitely more loyal to ideology than to a candidate, because some many voters were clearly willing to coalesce on Joe Biden almost instantly after their candidate dropped out. However, Biden’s wing of the DNC may be pretty much out of dry powder. All their candidates are gone, so from this point Biden is on his own (no more gifted delegates). Also, we don’t know how many of Warren’s candidates will actually go to one candidate versus the other, or if she will have a change of heart and actually endorse someone later.    

Figure 7 (Click to enlarge!)

Also, momentum does matter. It seems that Biden’s overperformance in South Carolina had a much more powerful effect than many people would have expected, given how strong Sanders seemed going into the day.   The Chinese Corona Virus is a gift to Biden and the neo-liberal wing of the Democrats.  It has effectively blocked Sanders from increasing his momentum.

However, there are other wildcards such as the age of the people left in the primary.  Both Biden and Sanders are older men with known health problems. Will a health problem or perhaps an unforced or forced error on the debate stage stop one or the other?   Biden survived the latest debate with Sanders 1:1.  Biden looked visibly more engaged and the lack of a real audience may have helped him focus; it certainly hurt Sanders who is able to engage a crowd.

The response of the Democratic primary voters may be of concern to the Republican Party, as well. Many may find it surprising the Democratic voters, even primary voters, were listening so attentively that with such short notice of other candidates dropping out, they were able to shift support to Biden. These people are clearly watching the campaign with rapt attention and picking up on the signals that their candidate and others are giving them. The question is, who will start talking to them now with the other candidates losing authority?  Will other light houses to the Democrats, such as the Clintons and the Obamas, step up to advocate to voters?

Finally, Senator Sanders is legendary for having a passionate fanbase and great ground game. However, his progressive support is clearly not as strong as people thought. He did not get his full vote out that he was predicted to have just one or two weeks before super Tuesday.   He did slightly overperform, but he should have done far better according to the analysis.   Although there are theoretically over half of the pledged delegates left, Sanders would have to win great majorities.  Some organizations would become demotivated from the loss, but that is the modis operandi of Sanders active supporters. He might have activated them into a 2-minute drill mode, in which they doubled down on their efforts to get him elected.   They probably feel like they were robbed of victory on Super Tuesday, but can they get non-active Sanders voters to the polls?  Sanders voters are known to be younger people, and statistics show that older voters tend to vote in larger percentages.

But although Senator Sanders has not been personally afflicted by the Chinese Corona Virus, his campaign has. 

  • It effectively neutralizes Sander’s best weapon – his passionate army of ground infantry.
  • No one is going to go to Sander’s rallies.
  • No one is going to go to Biden’s rallies, but that is probably a welcome gift to Biden’s handlers. It allows Biden to avoid many appearances that give him more opportunities to gaffe.
  • In a crisis, people tend to want safety and stability, which probably favors the moderate Biden over Sanders, who although his ideas are very old (he has been consistent), they are new for the American voter base to actually consider.

To those who are passionate about either of the two front-running remaining candidates, best of luck in your campaigning and in your decisions.   For those who are part of the opposing party waiting to see who will face the President, sit back with the popcorn and enjoy the show.  And for everybody else, I guess you can just Netflix and chill, and ignore it all.   

Skip to toolbar