This is a history of illegitimacy in America. And here is my text,drawn from three Russians, who understand illegitimacy better thananyone else.

In June 2008, the U.S. Food and Drug Administration rocked the foodindustry—and food lovers—with its warning about tainted tomatoes.Consumers in New Mexico and Texas were contracting a rare, sometimesfatal strain of salmonella and the FDA feared that salmonellacontamination from tomatoes was the cause. In the weeks thatfollowed, a major outbreak spread across the country and worriedconsumers abandoned tomatoes. Fourth of July cookouts were not thesame, and BLT lovers complained that their favorite sandwich wasimpossibly dry. By the end of July, the outbreak had infected morethan 1,200 persons in 42 states.

The October 2008 issue of PS published a symposiumof presidential and congressional forecasts made in the summerleading up to the election. This article is an assessment of theaccuracy of their models.

On September 8, 2008, the Trial-heat Forecasting Model predicted thatin-party candidate Senator John McCain would receive 52.7% of thenational two-party popular presidential vote. The forecast of avictory of modest proportions for Senator McCain reflected hishaving a five-percentage-point lead over Senator Barack Obama inGallup's early September, post-convention poll (49% to 44%) and theBureau of Economic Analysis' August report of a 3.3% GDP growth ratein the second quarter of the year, about average for a secondquarter election year economy.

The October 2008 issue of PS published a symposiumof presidential and congressional forecasts made in the summerleading up to the election. This article is an assessment of theaccuracy of their models.

The Primary Model predicted that Barack Obama would win a narrowvictory with 50.1% of the major-party vote. Obama surpassed thatforecast by just a little more than one standard error (2.5). Howcome the model came as close as it did with a forecast issued asearly as January? What prevented it from coming even closer? Andwhat might be done to improve the model?

The October 2008 issue of PS published a symposiumof presidential and congressional forecasts made in the summerleading up to the election. This article is an assessment of theaccuracy of their models.

In summer 2008, our Jobs Model forecast a Democratic presidentialcandidate two-party popular vote share of 56.6%, which would deliverthe incumbent party the biggest defeat of any post-World War IIcontest (Lewis-Beck and Tien 2008). However, we argued, from our analysis ofdifferent experimental and observational evidence, that thisunprecedented victory would be prevented by racially intolerantvoters. We estimated the net racial cost of being a black candidateand corrected our overall forecast downward to 50.1% for BarackObama. The unparalleled economic crisis, initiated after the releaseof our summer forecasts, prompted a reconsideration; the uniqueshock to the economy was no ordinary campaign perturbation. Wecalculated that the ensuing boost to anti-incumbent economic votingwould add approximately two percentage points to the opposition;therefore, we issued a public revision of our forecast to 52.0 % forObama (Lewis-Beck 2008).We are pleased that this final forecast fell so close to the actualresult of 53.5%. Nevertheless, we contend the actual result shouldhave been much closer to our original forecast. Given the dismalstate of the polity and the economy prior to the election, the Obamavictory should have been much bigger, as we show below.

The October 2008 issue of PS published a symposiumof presidential and congressional forecasts made in the summerleading up to the election. This article is an assessment of theaccuracy of their models.

The Time-for-Change Model proved one of the most accurate of the 2008presidential election forecasts run in the OctoberPS symposium. Using three predictors—thepresident's approval rating at mid-year, the growth rate of real GDPduring the second quarter, and the time-for-change dummyvariable—the model predicted that Barack Obama would win thepresidential election with 54.3% of the major-party vote. Accordingto nearly final tabulations compiled by uselections.org, as ofDecember 8, Obama has received just over 53.6% of the major-partyvote. However, it is likely that Obama's final total will reach53.7% of the major-party vote. Therefore, the model's current errorof 0.9 percentage points is likely to decrease further. The modelhas now correctly predicted the winner of the popular vote in allsix presidential elections since its creation in 1988.

The October 2008 issue of PS published a symposiumof presidential and congressional forecasts made in the summerleading up to the election. This article is an assessment of theaccuracy of their models.

Prior to the 2008 presidential election we provided forecasts of thefinal vote relying on a model containing only two variables: (1) thecumulated weighted growth in leading economic indicators (LEI)through the thirteenth quarter of the sitting president's term; and(2) the incumbent party candidate's share in the most recenttrial-heat polls. The novelty is the reliance on the advancedreading of the economy from the quarter ending in March of theelection year. (The exact equation and the exact forecast change asthe poll readings get closer to the election.) Our final forecast(Erikson and Wlezien 2008)based on trial-heat polls in August was that Barack Obama would win52.2% of the two-party popular vote. This turned out to be quiteclose to the Election Day outcome of 53.5% (as of December 2), alittle more than one percentage point above what we predicted.

The October 2008 issue of PS published a symposiumof presidential and congressional forecasts made in the summerleading up to the election. This article is an assessment of theaccuracy of their models.

The Incumbency and National Conditions Model predicted that JohnMcCain would receive 44.3% of the two-party presidential vote; heended up with 46.6% (at this writing, November 13, 2008), yielding aprediction error of 2.4 points. This error is slightly smaller thanthe mean absolute value of the out-of-sample forecasts from 1952 to2004 (2.68), so the 2008 election result is not anomalous in thecontext of this model. The addition to the model of the 2008 outcomedoes not generate many important changes to the coefficients, exceptthat the significance of incumbency interaction went from 0.08 to0.02 and the standard error and adjusted R2 of the modelboth indicate a slight improvement in fit.

The October 2008 issue of PS published a symposiumof presidential and congressional forecasts made in the summerleading up to the election. This article is an assessment of theaccuracy of their models.

At its most basic level, the Economic Expectations andTime-for-a-Change Model performed well in that it successfullyforecast a Barack Obama victory. It was in estimating the two-partyvote that the model underperformed. The point estimate was off byjust under five percentage points. While not terrible, the model didnot perform as well as it did in earlier years.

The October 2008 issue of PS published a symposiumof presidential and congressional forecasts made in the summerleading up to the election. This article is an assessment of theaccuracy of their models.

The presidential election forecast made with the Fiscal Model threemonths before Election Day put the incumbents' share of thetwo-party vote (VOTE2) at 48% (Cuzán and Bundrick 2008). As of the time ofthis writing, it appears the incumbent-party ticket of John McCainand Sarah Palin took a little less than that (around 46.5%). At1.5%, the error is less than 2.3%, which is Campbell's (2008, 680) “benchmark” for a“quite accurate” forecast. This is the second time in a row that theFiscal Model ranks among the best performers (Campbell 2005,23).

The October 2008 issue of PS published a symposiumof presidential and congressional forecasts made in the summerleading up to the election. This article is an assessment of theaccuracy of their models.

Ninety-nine days before the elections I utilized national- anddistrict- or state-level information to forecast the 2008 U.S.presidential, House, and Senate elections (Klarner 2008). Overall,the presidential model performed very well, while the House andSenate models understated the extent of the Democratic wave.

The 2008 presidential nomination was marked by the most aggressivefrontloading in recent history; the process was a mess from theoutset. Frontloading is the trend in recent presidential nominationsin which states schedule their primaries and caucuses near thebeginning of the delegate-selection season to have a greater impacton the process. In 1976, 10% of the delegates had been chosen byMarch 2. In 2008, 70% of the delegates had been chosen by that samedate. As part of their ongoing efforts to address frontloading andother problems, both the Democratic National Committee (DNC) andRepublican National Committee (RNC) revised the schedules and rulesfor 2008 presidential primary elections and caucuses.

The current method for selecting presidential nominees by the twomajor parties went into place mostly in 1972 and certainly by 1976,after Buckely v. Valeo. It was the naturalculmination of reform efforts over the history of the republic inthat, while prior reforms consistently invoked greater openness anddemocratic governance as rationales for their adoption, this methodactually empowered voters as the central figures in determining whowould be nominated (see Aldrich 1987). This fact became fully evident almost at once.The selection via primaries of senator George McGovern in 1972 andgovernor Jimmy Carter in 1976 as the Democratic presidentialnominees arguably not only would not have happened, they would nothave even come close to winning nomination without successful appealto the voting public.

There are perhaps many good arguments for Iowa maintaining its “firstin the nation” status, in terms of the presidential nominationprocess. The strongest, however, would seem to be an argument thatit is representative of the nation as a whole. That is, somehow,Iowa is a microcosm of the national political forces, faithfullymirroring the relevant electoral structures and choices of themacro-stage. This belief is certainly held by some. Palo AltoCounty, in northwestern Iowa, has long been considered apresidential bellwether, faithfully voting with the winningcandidate in a series beginning in 1916. But as media worthy as thatfact might be, it seems most likely a product of chance, for itsheavily rural, northern European-descended population make it farfrom demographically representative of contemporary America(Lewis-Beck and Rice 1992,4–6). A similar charge is commonly made today against the state as awhole, by political commentators across the land. But is it true? IsIowa really unrepresentative? That is the question we seek toanswer.

This article explores how results from Iowa may affect outcomes insubsequent nomination contests. We suggest that how Iowa matters maybe determined, at least in part, by how voters and the news mediaassess whether or not candidates meet or exceed expectationsthere.

Just as radio and television revolutionized the presidential electionprocess in earlier decades, today the “new media” are making asignificant imprint on how campaigns are conducted by candidates,covered by journalists, and evaluated by the voters. While the samebasic goals for campaigns apply, the tools to accomplish these goalshave expanded.

Forty years ago, violent protests at the Democratic NationalConvention captured the attention of the nation as rioters ventedtheir anger over a nomination process they felt excluded theirvoices. The disastrous 1968 convention spawned a cascade of reformsin the presidential nomination system, many of which were intendedto create greater opportunity for meaningful participation of theparty's rank-and-file members. Forty years later, where do we stand?Does the nomination process meet the goals of encouraging broadparticipation and connecting rank-and-file preferences to nominationoutcomes? We offer some tentative answers to these questions bytracing the history of the nomination process, its evolution overthe last 40 years, and the implications of several key changes inthe system for citizen participation.

In a book published four years ago, Andrew Busch and I dividedproposals to reform the presidential nomination process into twocategories, which we called comprehensive and incremental (see Mayerand Busch 2004, chapter5). Like any attempt to classify a complex reality, this distinctionblurs a bit at the edges, but the basic idea should be clear.Comprehensive proposals call for major, far-reaching changes in thebasic operations of the presidential nomination process; incrementalproposals make more limited, marginal changes in the rules, whileretaining the fundamental structure of the existing system.

As part of their ongoing efforts to address frontloading and otherperceived problems, both the Democratic National Committee (DNC) andRepublican National Committee (RNC) proposed revised schedules andrules for 2008. The major changes for the Democrats were that twonew states were allowed to join Iowa and New Hampshire in violatingthe official February 5 start date. The idea was that thesestates—Nevada from the West and South Carolina from the South—wouldenhance participation by more diverse populations (Latinos andAfrican Americans). While the Republican rules called for states tolose half of their delegate vote if they violated the timing rules,the Democrats implemented a “death penalty” requiring any stateviolating the timing rules to lose all of its delegates. TheNew York Times called these changes the biggestshift in the way Democrats have nominated their presidentialcandidates in 30 years. Yet in the end these changes did little tolessen frontloading, as 70% of all delegates were actually chosen bythe beginning of March. Two large states (Michigan and Florida)defied both national parties and voted before February 5.