Are the Chickens Coming Home to Roost?
This implies that we can evaluate the Democrats’ nominating scheme as a social choice mechanism. We can ask whether it does a good or bad job at aggregating individual preferences into a social choice.
In my opinion, the Democrats’ nominating system stinks. It doesn’t stink as much as the Republicans’ bone-headed scheme, but it still stinks. Before I get into my list of grievances against it – I have to clear away a bit more theoretical underbrush. As I said, a nominating system such as this translates individual preferences into a social choice. Depending upon the quality of the system, it may do a better or worse job.
We might think of it this way.
Preferences of the Voters ->Social Choice Mechanism ->
Social Choice + “Systemic Influence”
The “Systemic Influence” would be the influence that the system itself has on the outcome. So, for instance, assume that 70% of all voters prefer Candidate A. There is some kind of nominating system that determines who wins the nomination – and Candidate A gets 95% of the vote. The difference of 25% could be called the “Systemic Influence.” Importantly, we can never reduce the expected value of systemic influence to zero. There is no perfect way to translate individual preferences into a social choice. However, that’s not to say that the Democrats couldn’t do a better job than what they do.
The problem for the Democrats might seem small at first. Obama has won about 53% of the delegates, and about 51% of the popular vote. That is a pretty small difference. The problem is that the difference is systemic. The nominating system seems to contain several biases that favor Obama.
First, there is a “small state” bias. This exists in the Electoral College: small states have a proportionally bigger sway than large states. The Democrats have imputed this bias into their delegate allocation scheme. We can appreciate this via the following graph. It compares the number of voters per state who voted Democrat or Republican in 2004 against the number of 2004 Kerry voters (our rough measure of Democrats per state) per pledged delegate to the convention. Don’t be confused by this latter statistic. It is really straightforward – Kerry voters divided by pledged delegates. This is meant to give us a sense of how well Democrats in a state are represented at the convention. The smaller the number, the better represented they are.
This is the small state bias at work. Small state voters are better represented at the convention than large state voters. Notice that the relationship between the two is logarithmic. For sparsely populated states, a small increase in the number of voters yields a big increase in the number of Kerry voters per delegate. With the more populous states, a big increase in population yields a small increase in the number of Kerry voters per delegate. Thus, we see a big difference between Wyoming and Hawaii, but a small difference between New York and California.
We can “linearize” this relationship by taking the natural log of each state’s total vote. What this does is essentially turn the r-shaped relationship into a straight line. This might help us pick up on some details that the above graph is hiding.
As we can see, the basic relationship remains. Voters in larger states are not as well represented as voters in smaller states. Thus, the states form a band that move from the bottom-left to the top-right. Look carefully at it, and you’ll notice a curiosity. States at the top of the band are almost always strong Kerry states, while the states at the bottom are almost always strong Bush states.
This implies that Bush states are better represented at the convention than Kerry states, independent of population. For instance, examine the vertical cluster of observations about a third of the way across the graph – Maine, Nevada, New Hampshire, New Mexico, West Virignia, Nebraska, and Utah. All of these states had roughly equal numbers of voters in 2004. Look carefully at their ordering. There is a general pattern there. The stronger a state goes for Bush, the better representation the Kerry voters in the state have at the convention. Maine has one delegate for every 16,500 Kerry voters. Nebraska has one for every 10,500 Kerry voters.
This pattern persists across the whole graph – states of equal size get more or less delegates depending upon how strongly they went for George W. Bush. This implies a Republican state bias. Let’s examine this directly by comparing Kerry delegates per state with Bush’s share of the two-party vote.*
This data is strongly consistent with the theory of a Republican state bias. Of course, there is evidence of the small state bias here as well. Smaller states tend to be in the bottom-right, and larger-states in the top-left. But the tight band of observations that go from the top-left to the top-right as Bush’s share of the vote increases demonstrates this Republican bias very clearly.* A few examples can amplify this point.
-Note the states we discussed before – ME, NV, NH, NM, WV, NE, and UT. They are arrayed from the top-left to the bottom-right – with voters in Bush states getting more representation at the convention. Maine and New Hampshire are in the top-left, Nebraska and Utah are in the bottom right.-Note the distance between Idaho and New Hampshire – both of which have 4 electoral votes and roughly equal populations. There is one pledged Democratic delegate from Arizona for every 16,000 Kerry voters. There is one pledged delegate from New Hampshire for every 10,000 Kerry voters.
-Compare Texas to Illinois. If there was just a “small state bias,” Illinois Democrats should be better represented than Texas Democrats. After all, Illinois is smaller than Texas. In fact, the opposite is true.
-Compare Alabama and Connecticut, and you will see the same relationship. Connecticut is smaller, but Alabama Democrats are better represented.
Finally, there is the caucus bias. We already know about this. Caucus participation is much lower than primary turnout – but the DNC does not take this into account when it allocates delegates to states. Accordingly, caucus state delegates have fewer Democrats “behind” them. For instance, for every one pledged Obama delegate from Minnesota, there are 2,862 pro-Obama caucus-goers. For every one pledged Obama delegate from Wisconsin, there are 15,381 Obama primary voters.
This has the effect of enhancing the Republican state bias. Bush’s average share of the vote in the 15 states with a caucus process was about 57% (compared to 53% across all states). So, most caucus states get a boost anyway, just for being Republican. When we factor in the caucus bias – we have to conclude that the relatively few Democrats who participate in caucuses are much better represented at the convention than other voters.
The small state bias makes sense to me. The Electoral College has a small state bias in part to protect against regional candidates from winning the White House on a sectarian campaign. It makes sense for the Democrats to have similar protective measures. However, the Republican and caucus biases seem difficult to justify. Why do they exist? Of course, there is a mathematical answer to the question (that is explored in an endnote). Delegates are allocated according to formulas that embody these biases. However, referencing the formulas only begs the question. What we are interested in is why the party has arranged matters in this manner.
I think it is due in part to the fact that politicos have taken lousy care of the parties. The best and brightest of both parties haven’t cared enough to manage the nomination process with an eye to the future. They just don’t think much about the process, or about the party organization generally.
At its core, the nominating system is a logically inconsistent hybrid. Both parties changed their fundamental orientation to how nominees should be chosen in the 1970s – but they did not bring fundamental change to their nomination systems. Instead, they added openness requirements to the old scheme. State parties still send delegates to a convention that decides on a nominee. The difference now is that they must have open selection methods. What we have then is a Progressive Era variation of a Gilded Era system. There is no internal logic, no answer to the question: if the voters should decide, why retain delegates and conventions?
As a matter of fact, the system has barely been refined since the alterations of the 1970s. Forget redesigning the system to match the times. We’re talking about tweaks to improve it at the margins. These don’t really happen, either. Politicos created this hybrid with no internal consistency, and never returned to evaluate carefully whether further reforms would be needed to avoid a “perverse” result.
And what is a “perverse” result? Let’s return to our initial schema:
Voters Prefer Candidate A ->Nomination Rules Aggregate Individual Preferences into Social Choice ->
Candidate B Wins Due to Systemic Biases in Nomination Rules
This is “perverse.” Candidate B has effectively gamed the system – which is not to say that he intended to, but only that he was the systematic beneficiary of the biases. And so, we see voters preferring one candidate and the process conferring the nomination on another.
This year, Barack Obama is benefiting from several of these biases. So, there is the potential for this kind of “perverse” result. It could happen that Clinton wins the votes while Obama wins the pledged delegates. It need not be this way. No system is perfect, but if Democrats had been forward-thinking about their system – they might not be in such a bind.
This relates to a point I have made before on this page. Our country has done a poor job maintaining its political parties. We pay dearly for our negligence. Every two years, we complain about non-competitive congressional elections. In between elections, we complain that members of Congress are irresponsible and unresponsive. We ask, why is Congress broken? Perhaps it is because the parties – the greatest mechanisms ever invented for managing governmental agents – have been stripped of their power. They have been given over to what scholars call “candidate control.” Candidates are not responsible to the parties and the voters they represent. Instead, the parties are in service to the candidates. There is no doubt that the parties of the 19th and early 20th centuries were malfunctioning, corrupt, and irresponsible. But rather than reform them, we decimated them.
I think this nomination debacle is, in part, the fault of our disregard for the political parties. They are these hollowed-out husks that cannot handle the simple task of resolving a two-way dispute. And so, many Democrats want this robust and healthy debate to end because they are worried about chaos, and rightly so. If Clinton comes all the way back, this unrefined, antiquated, foolish system is going to have to settle the matter.
The Democrats’ hope is that Clinton or Obama humbly accepts the vice-presidential nomination (you know, the one not worth a bucket of “spit”). I find this indicative of the times. As the Democrats stare down chaos, their hope hinges upon the personalities of the candidates. That’s candidate-controlled politics in a nutshell.
[*] For the sake of visual ease – I have excluded the District of Columbia from this graph. DC only gave Bush 9% of the vote in 2004 – including it would expand the horizontal size of the graph and thus reduce the ability to see individual data points in the main cluster distinctly.
[*]To confirm the visual observation – an OLS regression equation was run that regressed Kerry delegates per state on a function of the total votes per state, a function of Bush’s share of the two-party vote (to correct for some observed heteroscedasticity), the number of Electoral College electors, and whether the state was a swing state (i.e. whether the victor’s margin of victory was less than or equal to 5%). Bush’s share of the two-party vote had a statistically significant effect in the expected direction. The model seemed to satisfy the assumptions of OLS regression. It explained over 90% of all variation, and all variables were found to be statistically significant. Washington D.C. was included in the analysis.
[*] Let’s take a moment here to discuss the DNC’s delegate allocation formula. In particular, how is it that there is a “Republican bias” in it? The DNC computes a state’s base number of delegates via an “allocation factor,” which is:
Allocation Factor = 1⁄2 × ( ( SDV ÷ TDV ) + ( SEV ÷ 538 ) )
SDV is a state’s vote for the Democratic presidential candidates from 1996, 2000, and 2004, TDV is the total vote for the Democratic presidential candidates from those years, and SEV is a state’s electoral vote.
I see at least three sources of Republican bias.
First, it does not include a “discount” factor for 1996. Bill Clinton won 12 states that year that John Kerry lost in 2004. That gives the count a slight pro-Republican bias.
Second, and relatedly, it takes no account of turnout by year. Turnout was up drastically in 2000 and 2004 over 1996. This can create “perverse” quirks in the formula. For instance, John Kerry’s 13-point loss in North Carolina in 2004 is counted about as heavily as Clinton’s 18 point victory in New Jersey in 1996.
Third, and most important of all, partisan differences between states get squashed. The “allocation factor” is actually just the average of a state’s contribution to the party’s national vote and its contribution to the Electoral College. This favors states that favor Republicans, which can be see if we take a example.
Let’s look at Indiana, Missouri, Tennesse, and Washington – all of which have the same number of Electoral College votes (11). We’ll examine their share of the Democratic vote, their share of the Electoral College vote, the average of the two (i.e. the allocation factor), and the difference between the average and their share of the Democratic vote.
Look at what happens. Republican states get a boost, Democratic states get a burden. So, Washington state contributed 0.8% more to Democratic victories than Indiana, but its allocation factor is just 0.4% larger. Interestingly, the size of the boost or burden depends upon how Republican or Democratic the state is. A state with no votes for Democrats and 11 electors would enjoy a boost of 1%. A state that contributes 4% of the total Democratic vote and 11 electors would suffer a burden of 1%.
The small state bias also plays a role here. Larger states make up a proportionally smaller share of the Electoral College than small states (for instance, California’s population is 72 times the size of Wyoming’s, but its Electoral College delegate is only 18 times Wyoming’s). Thus, all small states would get some boost, regardless of how they voted. Similarly, all large states face some kind of burden. But the size of the boost or the burden depends upon partisanship. Republican states are treated better than Democratic states.