TCS Daily

How Would a Computer Pick the Prez?

By Nelson Hernandez - September 29, 2004 12:00 AM

TCS contributor Douglas Kern's recent article ("President Elect - 2004") regarding the success of Commodore 64-era political game President Elect 1988 in predicting elections prompted a search by TCS staff for the designer/programmer of that game, Nelson Hernandez, Sr. We tracked him down. In this article, the man who banged out the original BASIC source code in 1981 on his Apple II+ computer explains who he thinks will win -- and why.

-- The editors

My comments on Doug Kern's experimentation with my game must be general; a detailed critique of his methodology would be an impenetrably esoteric discussion for most readers. But the main point I would like to make is that the game indeed projected the 1988 election with uncanny success well in advance, but it cannot be applied to the 2004 election.

In real life as well as in President Elect 1988, each presidential election takes place within a certain contextual background wherein the electorate subjectively evaluates the relative success or failure of the incumbent party, which is then politically rewarded or punished. In every election cycle the voting population arrives at a collective answer to candidate Reagan's famous 1980 debate question, "are you better off today than you were four years ago" well before the election takes place. PE 1988 knew the actual situation in 1984 with perfect hindsight and could quantify the incumbent party's relative success or failure in 1988 based on hypothetical economic/situational inputs using a fairly simple mathematical formula I created to compare the current overall "state of the union" to what it was in the previous election.

However using PE 1988 to project 2004 is problematic because the economic and national security/foreign situation inputs Kern was plugging for 2004 were being compared to the state of the union in 1984 instead of the one which prevailed in 2000. This mismatch alone renders his experiment moot.

There are other mismatches. Since the 1988 election, 30 electoral votes have moved from the northeast quadrant of the nation and shifted elsewhere. Of these, 19 were distributed to California, Texas and Florida. The 11 states of the old Confederacy have increased their electoral representation from 138 to 153 electoral votes, now constituting 57% of the votes needed to win (270).

Additionally, most states deviate from the national average according to long historical trends. Some states are politically stable, consistently Democratic or Republican-leaning. Some states are slowly evolving into a new pattern due to demographic trends. Vermont is the nation's best example: the most heavily GOP state in the nation in the 1930s, it is now among the most heavily Democratic. Sixteen years having elapsed since 1988, projected state biases from 1988 do not apply to 2004. Voting patterns have changed in many states.

Finally I diverge from Kern's assessment of this year's candidates (shown here on a 1-9 scale, where 9 is equal to the very best candidates in history in each category and 5 is merely average).                           

 Bush Speaking  4 3
 Bush  Magnetism                                      7 6
 Bush  Poise                    5  8
 Kerry  Speaking 
 5 6
 Kerry  Magnetism   4 4
 Kerry  Poise
 5 7

In PE 1988, Magnetism is the most important attribute, in that a charisma advantage puts the inferior candidate at a disadvantage every week of the campaign. Poise is shorthand for how blunder-prone a candidate is on the campaign trail and under the intense pressure of a televised debate. Speaking, a measurement of rhetorical skill, is the least important factor; it is mainly used as an effectiveness multiplier during debates.

Hence I drop the magnetism gap between the candidates from 3 to 2, helping Kerry, and increase the speaking gap, also helping Kerry. As for poise, my working assumption is that Bush's coolness under debate pressure has been underappreciated; witness his narrow but unmistakable victories in three successive televised debates with Al Gore in 2000, primarily thanks to charismatic intangibles. Kerry is at a disadvantage in this area because blunders can take many forms; in his case it could be excessively opaque or contradictory policy pronouncements, or striking an excessively dour note in a debate, etc. Kerry's biggest vulnerability is probably stylistic.

The War on Terror is clearly the decisive issue of this election. In all "war" elections, national security/foreign policy considerations outweigh economic and social policy. (In "peace" elections the reverse is true.) This being the case, the game allows the player to subjectively measure the overall mood toward the war on a 1-9 scale, where WWII received an unequivocal '9' and Vietnam in 1968 received a '3'. (That war was so divisive by March 1968 that President Johnson announced he "would neither seek nor accept" the Democratic nomination that year.) Kern does not identify how he measured the nation's attitude toward the War on Terror. For my part I would rate it a provisional '6', with the number climbing +1 each time there is noteworthy forward progress toward an independent Iraq and -1 whenever U.S. troop casualties notably increase in tempo. A '6' rating offers the incumbent only a slight advantage, all other things being equal.

Finally it is unclear how Kern rated Kerry and Bush ideologically. In the game, candidates are offered subjective ratings on a 0-100 scale regarding their economic, social and national security/foreign policy positions, with 100 signifying an ardent left-winger and 0 an arch-conservative. Candidates outside the 25-75 band are punished in the game's internal formulas in proportion to their extremism. This year ideology is a ticklish question because Sen. Kerry is currently the most liberal Senator in Congress based on his voting record, certainly in the 90s overall. However he has assiduously tried to mitigate this disadvantage in 2004 by seldom discussing his Senate record and offering multiple nuanced policy variants on high-profile issues, particularly the War on Terror.

On the other hand, this election has seemed less about articulated policy issues and political ideology than any in the past. To editorialize, political campaigns are now more about entertainment and political theater because substantive, intellectual discussions of complex public policy topics result in poor television ratings and apparently have no positive effect on "swing voter" behavior. Try to imagine the show-stopping absurdity of Bush and Kerry earnestly arguing the particularities of an issue as technical and specific as the fate of the islands Quemoy and Matsu (as Kennedy and Nixon did in 1960) and you get a sense of how far we have come toward presidential politics becoming just a high-stakes reality television show, where the tactical objective is to simply to entertain, titillate and seduce the fickle "swing voter".

Turning to my assessment of this year's election, I have believed all year that Bush will win re-election by a comfortable margin -- though certainly not the 16 points Kern stipulated in his experiments. (Dr. Ray C. Fair of Yale, the nation's foremost presidential election forecaster, has advanced a similarly outlandish Bush margin of victory.) I personally would guess the final victory margin in the 8% range barring any major intervening events (I am writing this on September 24). The reasons are conventional and straightforward: Bush has greater personal charisma, the economy is not in a visible tailspin and is not dramatically better or worse than it was in 2000, the War on Terror is mildly divisive yet about 60% of the voting population supports its continuance. Those things factored in, an incumbent has a natural advantage (being better able to shape events) and finally the aforementioned shift of electoral votes to the Sunbelt favors the GOP.

My sense is that none of the Democrats who ran this year would have been likely to defeat Bush under the above circumstances. I think Dick Gephardt would have presented the most formidable opponent: he could have picked up a few close Midwestern states (e.g. Missouri, Iowa, Minnesota) while retaining all the states Kerry will win this year. In addition he would have been more likeable, less liberal and less vulnerable to attack than Kerry. Combined with moderate-to-conservative running mate from a battleground state capable of definitively swinging his home state into the Democratic column, and this election would have been very tight indeed. But even the optimal ticket (from a purely tactical standpoint) from the roster of candidates that ran this year would not have offered the Democrats a cakewalk.

I am also inclined to think the Democrats will almost certainly return to the White House following the 2008 election. The reasons I feel this way are entirely speculative and beyond the scope of this discussion, but I'll give a hint: the economy, which this year is practically a non-issue, may be the only relevant issue in 2008. I believe the incumbent party in that election year (presumably the GOP) will be at a serious disadvantage. In a perverse way the GOP might be politically better off in the longer run to let Kerry win this year while holding onto both houses of Congress.

All the above said, it's awfully flattering to hear someone still singing praises for a simple game I designed 23 years ago.


TCS Daily Archives