Election Forecasts - MH Prognostics

Election Polls vs. Election Result Forecasts

auf Deutsch

Election polls are undoubtedly an important basis for all election forecasts. But the significance of polls depends - besides usual methodological factors (sample size etc.) - to a great extent on factors like time remaining until election day and stage of the election campaign, but also on the expected decisiveness of voter turnout and the definitiveness of voting choice intentions.

Therefore it is also very critical if a voting choice intention reflects only a current tendency ("if you are not sure: towards which party/candidate are you leaning currently?") that might or might not materialize on election day, or if it mirrors a definite voter preference that will materialize on election day with almost absolute certainty ("I would cast my vote even if it rained cats and dogs")

If for example in a serious poll - with a margin of +/- 3% - party A is leading party B by 7 points, then one might think that party A will nearly certainly come out on top of party B on election day. But if a considerable proportion of supporters of party A is not definitively determined or still unsure if they will cast a vote on election day, whereas a significant higher proportion of supporters of party B are more determined to cast a vote - and follow through their intention on election day - then it should be no major surprise if party B ekes out a close win on election eve.

MH election forecasts therefore include a string of other factors which might have a substantial impact on factual voter choice behavior - and therefore on the election result, including - without limitation - fundamental data (party preferences and their consolidated changes over a medium to longer period of time, i.e. excluding the "noise" of short-term poll fluctuations), momentum of campaigns, key campaign success factors (skills of campaign team, minimizing of risk factors, media trends and tendencies...) etc.

The most important differences at a glance

Major Factors in usual (aggregate) Election Polls	Major Factors in MH election (result) forecasts
---	Weighting of multiple influencing factors
---	fundamental data
polling data	polling data
corrective factors (*)	corrective factors
time series of single polls	weighting of polls
(linear) trend estimation(*)	"Momentum" of election campaings
expected turnout(*)	expected turnout by sub-groups
---	campaign success factors
---	campaign risk factors
---	tendencies of published opinion
---	public mood regarding crucial issues

Best validity for day(s) of last conducted poll

Election forecast for actual election day

(*) partially (but not always) considered

As election day is approaching, the number of undecided voters generally decreases, therefore - at least theoretically - actual polling data and election forecasts should converge towards election day.

Additional note:
Election result forecasts of MH Prognostics are based on numerous factors (see above) whose individual weighting have to be monitored regularly (and adjusted accordingly if necessary), as well as on a prospective assessment of the remaining campaign season, before a single vote has been cast or counted.

By contrast Election result computer predictions/extrapolations are mathematical/statistical methods which make it possible to project final election results based on partial (= real-time) results - with increasing accuracy as the ballot count progresses.
Therefore these two instruments should not be confused!