GEEK ALERT: When Mean Isn't Median AND Median Isn't Mode

GEEK SPEAK WARNING!  However, discussed in a way all can understand.


Quick definitions -- Median = Half Above/Half Below; Mode = Most Commonly Observed Data Point;  Mean = Average.  Take your standard bell curve.

On this curve,the vertical line in the center represents the mean, median, and mode.  But in the "real" world the results we observe don't always fit this curve.  Let's focus this discussion on the question of Median and Mode.  I'll save the Mean, no pun intended, discussion for a later time.


In the 2008 Presidential Election, our model generated the following chart:

Forecast: Tsunami Warning Issued for House; Storm Clouds Gather in Senate

Post for October 29, 2010 --


New polling in both the House and Senate served to both clarify and obscure expected results for this coming Tuesday.
  
Polling in Maine by Critical Insights shows potentially close races in both of the state's two Congressional districts: 

• ME-1:   Republican Dean Scontras  leading (but well within the margin of error) Rep. Chellie Pingree by a margin of 45-41.  Two weeks ago the same pollster showed Pingree leading by 45-40.  The significant note is that Critical Insights' poll from the middle of May showed Pingree with a 24 point lead.
• ME-2:   Republican Jason Levesque trails incumbent Democratic Rep. Mike Michaud by a margin of 44-40 (also within the margin of error).  Two weeks ago Critical Insights' polled the district and found Michaud leading by 19.

Significance: If Democrats were to lose both of these seats in a state where polls close relatively early, Republican gains will be massive.


As it is, our model's evaluation of InTrade data suggests only "huge" gains of approximately 60 seats from their post 2008 election level.  Friday's analysis gives Republicans 237 seats in the next Congress.  The Republican probability of taking the House 92.3% under our model.  On our 2006 vintage InTrade analysis on the Friday preceding the election, Democrats' chance of taking the House was 72%.  Democrats gains 31 seats (compared to post 2004 elections) that year.    The probability of that size of gain (30 or more seats) on Friday before the election was 35.4%.  That is roughly the same probability our model generates for gains of 65+ using InTrade data based on Friday data.   Non-scientific observation: Friday before election data appears to have UNDERESTIMATED election day results in the House. 


Friday's Daily Snapshot and House Pictures

Our Senate analysis makes Democrats the favorites to maintain control of the Senate.  Our analysis of InTrade data suggests that Republicans will hold at least 49 seats.  Politically, this means Democrats control; for InTrading purposes it means Neither Party to Control contract would prevail.  In fact, our modelling suggests that on a tighter trend analysis, Republicans may well end up with 50 seats or better.

Here are the Senate pictures.

Increasingly, the data seems to suggest that Pennsylvania's appearance on the screen ten days ago was a "false warning."  Data increasingly suggests Pennsylvania may be well beyond Democratic efforts to save the seat.   There are increasing signs that recent polling data showing Washington within a point either way are no longer the exceptions, but the consensus among pollsters.


Questions of the day.   Add your input and observations.




1.   Will Senator Lisa Murkowski's write-in bid cost Republicans control of the Senate?

InTrade data currently favors Lisa Murkowski to win, but the situation is volatile.  Democrats are still an underdog in the seat, BUT the DSCC is spending money there.  The Republican Senate nightmare ... win two of WV, WA, and CA ... and then find out you don't control the Senate because of Alaska.  Think it's possible? Probable? Or No Way that Happens? 

2.   What Senate race will be the biggest surprise on Tuesday?

Republican House 90%; 49 Republican Senators

90% Probability of a Republican House


Pundits are almost universal in predicting a Republican House takeover next Tuesday.  Our model based on InTrade market data agrees pegging Republicans as a 90% favorite to do so.  The House debate right now centers around the question of how high the Republican wave in House elections will be.  Some pundits argue that Republicans will only capture the House by a narrow majority.  Others argue that the GOP will end with their highest seat count since the 1940s or earlier.  Right now our model projects Republicans holding 236 seats in the next Congress.  An important caveat -- our model suggests that the risk for greater losses by Democrats exceeds the risk of significantly smaller Republican gains.

Is Today's Republican Uptick in Races and Micro Model Beginning of a Rally?


Republicans saw an uptick in their prospects in our 7 key races, yet remain an underdog in their battle to claim the Senate.  Our macro and micro model projections are only separated by 1/10th of a seat in today's analysis of InTrade data.  The micro model projects 48.9 seats and the macro model 49.0 seats.  In short, our micro model is catching up with the macro model.  We would expect them to move in close tandem from this point forward.

As mentioned earlier, Republicans started to retrace lost ground in all seven of the most competitive -- or key -- Senate races.  Illinois, Washington, and West Virginia all saw Republican improvement of more than 3%.  However, in West Virginia and Washington as sizable gain is still required to push those seats into "Republican Favored" territory.  The table below summarizes our model's pricing of Republican probabilities in Thursday nights data run.

Here is our daily chart of this month's pricing movement for the seven races included in the above table.  As you can see, volatility remains and is likely to pick up in at least 4 of the 7 seats over the weekend.

Be sure to check out our InTrade Portfolio on our InTrade portfolio page.  Additional charts on House and Senate control probabilities are posted below.

Mid Day Update

House

GOP Control    89.7%

Seats              236

Senate

Micro Model GOP Control    6.9%

Macro Model GOP Control  11.6%

Micro Model GOP Seats     48.7

Macro Model GOP Seats    49.0

2 A Days Start Today

Starting today we will starty posting multiple updates each day.  The "mid-day" update will consist generally just of our models' topline estimates. 

Republicans HIt New Highs in House Projection; Senate Path Narrowing

House GOP Seat Projection at New High

Based on current InTrade data, our model projects a Republican seat count range of 236 to 237.  The Midpoint estimate comes in at 236 -- an increase of 59 over the post-2008 election seat count.  The low end estimate, also 236, and the high end estimate of 237 represent new highs in our model's estimate.   The probability of a Republican takeover of the House of Representatives stands at 89.5%.

Based on prior observation, we believe that we are likely to see continuing changes in our model's seat count estimate, but would expect trading in the House control contract markets to trade in a fairly narrow range over the next few days.

New polling data has been a mixed bag of news over the last few days.  In some cases, it appears previously believed endangered Dems may be holding on -- even if by a thread.  However, other races have delivered surprising information for Democrats.  Of particular note, Rep. Rick Bouncher (VA-9) has seen his lead in polling evaporate over the past month.  A previously reported 10 point advantage now registers at a one point deficit according to Survey USA polling.  In California, Rep. Jim Costa (CA-20) in what was once viewed as a safe Democratic seat now finds himself trailing and in serious jeopardy of defeat.

Republicans' Narrow Senate Path

Our micro model's simulation suggest a midpoint estimate of 48.5 Republican seats (see graph at left), while our macro model simulation pegs Republicans as emerging with 49 seats.  This is consistent with our model's analysis of individual races in which our midpoint estimates peg Republican probabilities at above 60% in enough races to bring the Republican seat count to 49.  Unfortunately, for Republicans, the model suggests that the three seats most in play for Republicans have current probabilities of under 40%.  Republicans need to win two of those three seats to win control.  The following table and graph illustrate our model's current evaluation of the state of those and four other races that are generally regarded as this year's Senate battleground.  Recent action has moved Pennsylvania away from Democrats, but we will continue to include it in our list based upon our model's view of volatility and volatility potential in the InTrade markets.

Random Thoughts



As far as pundit and pollster analysis goes, the last two days have seen some interesting observations and data on the electoral environment.  Yesterday, Nate Silver offered the following at FiveThirty Eight.com in answering the question Are Democrats Overachieving in the Senate?. 

Right now, among the 37 Senate elections, we have Republicans favored in 25, Democrats favored in 11, and one (Colorado) that’s too close to call. If Democrats have a relatively good election night, they will win about one-third of the available Senate races. And if anything, the states that are voting for Senate this year are slightly blue-leaning. If the entire Senate were up for re-election in this political climate, the Republicans would be favored to earn a filibuster-proof majority, and might even earn a veto-proof majority! Fortunately for Democrats, that’s not how the system works. (Maybe some of our readers could go though the list of 63 senators that are not up for re-election and guess which ones they’d expect to lose if they were. It could be kind of fun.)  emphasis added

Data released by Gallup today seems to offer a similar view of just how favorable the makeup of this year's electorate may be to Republicans.  Gallup noted at 2010 Electorate Still Looking More Republican Than in the Past that:

Gallup's latest figures on the composition of the 2010 electorate suggest that, consistent with an earlier Gallup report, those voting in this year's congressional elections across the country will be similar in gender, age, and education to 2006 voters. At the same time, they will be substantially more Republican in their party orientation, and more conservative than has been the case in the past several midterms.

The Gallup post is worth a read.  As noted here several times in the past, this year's election will be a REAL test of pollsters' likely voter models.  It is unfortunate that all of the public pollsters tracked by Real Clear Politics and others don't provide as much transparency as Public Policy Polling (PPP), Rasmussen and SurveyUSA in terms of the demographic composition of survey respondents contained in their likely voter polls.   The reality is that such data would allow analysts and others more flexibility in placing "value" on a particular poll.  

A Republican Senate Retreat? Or Preparing for a Final Assault?

Republicans were at a general standstill in our models' evaluation of their prospects to take control of the Senate.  The projected midpoint estimate of Republican Senate seats was unchanged in both the micro and macro models at 48.3 and 49.0 seats respectively.   While the toplines were stable there was considerable variance in individual races.

Same Pollster Different Day and State

A couple of days ago, Democratic polling firm Public Policy Polling (PPP) released a poll showing the Democrat with a significant lead.  The other day the state was West Virginia.  On Tuesday, the state was California.  Their poll showing Senator Boxer with a 9 point lead and news of Republican Carly Fiorina's hospitalization triggered a 13.8% decline in our model's midpoint probability estimate for the Republican.   One interesting aspect of this poll is that unlike Survey USA and Rasmussen Reports recent surveys, PPP estimates an electorate in California that is not only more Democratic than in the 2006 midterm elections, but more Democratic than the California electorat was in 2008.   PPP's partisan split in this poll was 47% Democrat, 34% Republican, and 19% Independent/Other.  In 2006 exit polls Democrats made up 41% of the electorate and in 2008 made up 42%.  At 34%, their estimate of the Republican share of the electorate is consistent with Rasmussen (34%) and Survey USA (35%) as well as 2006 exit polls (35%). 

We will have more to say on this at a later time.  But here is a table illustrating the differences.

In the House, our model currently projects between 234 and 236 Republican House seats following the election and a probability of House control that is just shy of 90%.  Our graphs of the trendlines for House and Senate seat estimates follow.  Also, check our InTrade Portfolio page to check out additional trades we made today.

Lastly, by popular demand we've updated our charts from a while back showing the Iowa Electronic Markets' probability of control change in the House and Senate compared with 2006.

West Virginia Takes Center Stage Today

West Virginia took center stage today in delivering volatility to the InTrade markets.  Our model's midpoint probability for a Republican win dropped 19.6% to 41.4%.  The move appears primarily related to the release of a new poll by Democratic polling firm Public Policy Polling (PPP) giving Manchin a 6 point edge.  This is up from a 3 point lead in PPP's last poll and compares to the most recent Rasmussen Reports survey in West Virginia that showed Manchin trailing by 7 points. 

This move in West Virginia dropped the number of seats that Republicans are above 50% in to 26 from 27.  If Republicans were to win all 26, they would end up with 49 seats in the Senate.  Based on current volatility where both candidates are bouncing around between 40% and 60%, we could well see  this seat bounce back and forth more than once over the next week.

Here are our two graphics on seven closely contested Senate races.

As an aside, several readers have asked how our projections compare with those of Nate Silver at FiveThirtyEight (now with New York Times).  In general, our projections have not varied greatly over the last two weeks.  Presently, our forecast is a little more positive for Republicans than his.  Here's the present comparison.  We've listed both our micro and macro models for the Senate.

Before turning to the House side, here is the updated trend chart from our Senate micro model.

On the House side of the ledger, our forecast did drop back one House seat to 234 in today's data run.  The probability of Republican House control also declined marginally, but Republicans still remain overwhelming favorites to take control of the House of Representatives.

Please take the time to check out our InTrade $play portfolio on the InTrade portfolio page.  No new trades today, but we will be making some aggressive moves tomorrow.

Tomorrow we will be putting up a new series of polls asking readers for their views on how they view the outcome of our seven key Senate races.  Make sure to vote and encourage friends and colleagues to contribute to their wisdom to the crowd.





Clock Ticking -- Republicans Still Coming Up Short for Senate Control

Republicans continue to poised to take control of the House of Representatives in the upcoming midterm elections.  Based on our model's analysis of data from the InTrade prediction markets, Republicans are projected to increase their seat count to 235 seats.  This would represent a gain of 58 seats from the base of 177 they held following the 2008 elections.

Republicans continue to enjoy a better than 50% chance of winning 5 out of 7 closely contested races.  Were they to win all 5 they would end up holding 50 Senate seats in the new Congress.  At present Democrats have better than a 70% chance of winning in both Washington and California.  While Democrats have to be considered favorites to hold those seats polling data suggests the races remain close.

Our micro model is currently projecting 48.3 Republican Senate seats, while the macro model clocked in at 48.9 seats.  

We can expect to see some renewed volatility in the markets this week as they respond to new polling data and rumours surrounding developments in the individual Senate races.  In particular, we will be watching for new polling data in our list of key races and also new data from Delaware and Connecticut.  While Republican candidates are currently priced at less than a 10% probability of winning, we could see an uptick in prices if there appears to be any favorable movement in the polls in those two races.