Posted on 09/05/2004 4:01:55 AM PDT by dvwjr
Analysis of recent Newsweek/PSRAI post-RNC convention polling data, with derived and revised Kerry and Bush post-convention 'bounces' and poll leads. Comments welcome...
dvwjr
First, here is the relevant poll data from the Newsweek/PSRAI three-way poll taken from the Newsweek September 2nd - 3rd, 2004 poll released Saturday September 4th, 2004.
Newsweek Poll: Post-Republican Convention, Princeton Survey Research Associates International Final Topline Results (9/4/04) N = 1,008 Registered Voters, 18 and over Margin of error: plus or minus 4 Interviewing dates: September 2-3, 2004 SAMPLE SIZE/MARGIN OF ERROR FOR TOTAL ADULTS: 1,188 Total adults (plus or minus 3) SAMPLE SIZE/MARGIN OF ERROR FOR REGISTERED VOTERS SUBGROUPS: 1,008 Registered voters (plus or minus 4) 505 Thursday interviews (plus or minus 5) 503 Friday interviews (plus or minus 5) 374 Republicans (plus or minus 6) 303 Democrats (plus or minus 6) 300 Independents (plus or minus 6) 476 Men (plus or minus 5) 532 Women (plus or minus 5) 112 18-29 (plus or minus 10) 349 30-49 (plus or minus 6) 516 50+ (plus or minus 5) 283 Southern White (plus or minus 7) 559 Non-Southern White (plus or minus 5) 144 Non-White (plus or minus 8) 417 Military households (plus or minus 6) 578 Non-military households (plus or minus 5) 310 Republican states (plus or minus 6) 409 Swing states (plus or minus 6) 289 Democratic states (plus or minus 7) 541 Bush/Cheney supporters (plus or minus 5) 403 Kerry/Edward supporters (plus or minus 6) SAMPLE SIZE/MARGIN OF ERROR FOR SUBGROUPS OF REGISTERED VOTERS INTERVIEWED FRIDAY 187 Republicans (plus or minus 8) 139 Democrats (plus or minus 9) 165 Independents (plus or minus 8) NOTE: An asterisk (*) indicates a value less than 1%. **REGIONAL DEFINITIONS Republican states: AL, AK, CO, GA, ID, IN, KS, KY, LA, MS, MT, NE, NC, ND, OK, SC, SD, TX, UT, VA, WY Swing states: AZ, AR, FL, IA, ME, MI, MN, MO, NV, NH, NM, OH, OR, PA, TN, WA, WV, WI Democratic states: CA, CT, DE, DC, HI, IL, MD, MA, NJ, NY, RI, VT Results for questions 1a/b and 2a/b include leaners. 1a/b. Suppose the election were being held TODAY and you had the following three choices for president and vice president... (INSERT CHOICES AND ROTATE) George W. Bush and Dick Cheney, the Republicans; John Kerry and John Edwards, the Democrats; and Ralph Nader and Peter Camejo, the Independent or third party candidates. Who would you vote for? [IF DO NOT SUPPORT BUSH/CHENEY, KERRY/EDWARDS, OR NADER/CAMEJO RESPONDENTS WERE ASKED] As of TODAY, do you LEAN more toward...Bush and Cheney, the Republicans; Kerry and Edwards, the Democrats; or Nader and Camejo? BASED ON REGISTERED VOTERS Three-Way Trial Heat Total Bush Total Kerry Total Nader Undec./ /Cheney /Edwards /Camejo Other Current Total 52 41 3 4 =100 Thursday interviews 49 43 3 5 =100 Friday interviews 54 38 4 4 =100 Republicans 94 4 1 1 =100 Democrats 14 82 1 3 =100 Independents 45 40 9 6 =100 Men 54 39 4 3 =100 Women 50 43 2 5 =100 18-29 45 45 9 1 =100 30-49 56 37 3 4 =100 50+ 50 44 1 5 =100 Southern White 66 30 1 3 =100 Non-Southern White 51 41 4 4 =100 Non-White 34 56 4 6 =100 Military households 58 36 2 4 =100 Non-military households 48 43 4 5 =100 Republican states 56 37 2 5 =100 Swing states 51 42 3 4 =100 Democratic states 47 44 5 4 =100 Trends (8/5-10/04) 45 47 2 6 =100 (7/29-30/04) 42 49 3 6 =100 (7/8-9/04) 44 47 3 6 =100 (5/13-14/04) 42 43 5 10 =100 (4/8-9/04) 42 46 4 8 =100 (3/25-26/04) 45 43 5 7 =100 (3/18-19/04) 45 43 5 7 =100 Note: August 5-10, 2004 trend from Pew Research Center. July 29-30, 2004 poll conducted during the week of the Democratic National Convention and should not be used to estimate Bush's post-convention bounce. Trends before 7/8-9/04 based on slightly different question wording that did not include vice presidential candidates' names. 1c. Do you support (INSERT Q.1a CHOICE: Bush and Cheney/Kerry and Edwards/Nader and Camejo) STRONGLY or only moderately? BASED ON REGISTERED VOTERS
Source 1: PRnewswire - Newsweek September 2-3, 2004 poll
Source 2:Yahoo Financial News - PRnewswire: Newsweek September 2-3, 2004 poll
First check some of the raw poll numbers in a consistency check. Note that the 'plus or minus' figures in parentheses are used to evaluate the other poll questions which involved a particular poll sub-group. Note that the MoE for each sub-group is necessary for certain poll questions elsewhere in the published article, but does not imply that the raw data varies by that amount.
SAMPLE SIZE / MARGIN of ERROR FOR REGISTERED VOTERS SUBGROUPS: 1,008 Registered voters (MoE: plus or minus 4%) 476 Men (MoE: plus or minus 5%) 532 Women (MoE: plus or minus 5%) -------------------------------- 1,008 Total respondents 310 Republican states (MoE: plus or minus 6%) 409 Swing states (MoE: plus or minus 6%) 289 Democratic states (MoE: plus or minus 7%) ------------------------------------------------- 1,008 Total respondents
These match the 1,008 registered voter sample size. Other sub-groups do not total to the 1,008 registered poll respondents because there was no answer given. Next look at the political affiliation component of the poll.
374 Republicans (MoE: plus or minus 6%) 303 Democrats (MoE: plus or minus 6%) 300 Independents (MoE: plus or minus 6%) ------------------------------------------ 977 Total
which leads to 31 poll respondents who did not self-identify as to party affiliation counted as 'other'. So, with 1,008 total poll respondents we have 374 Republicans, 303 Democrats, 300 Independents and 31 'Others'. For this analysis, any 'Others' will be added to the Independent category.
374 Republicans (MoE: plus or minus 6%) 303 Democrats (MoE: plus or minus 6%) 331 Independents (MoE: plus or minus 6%) -------------------------------------------- 1,008 Total
The other important number set is the total number of Bush/Cheney and Kerry/Edwards poll respondents. This combined with the above Rep/Dem/Ind breakdowns will allow the recreation of the raw data table for this poll.
541 Bush/Cheney supporters (MoE: plus or minus 5%) 403 Kerry/Edward supporters (MoE: plus or minus 6%)
Now displayed below in an easier to read table format are the poll numbers for political affiliation along with the three-way results published by Newsweek/PSRAI as shown above.
Now given the data in the table above, next recreate the raw data table from three of the above listed data elements: Total respondents (1,008 Registered voters), party affiliation breakdowns (Rep=374/Dem=303/Ind=331) and the Bush or Kerry responses (541 Bush, 403 Kerry). The table so derived is displayed directly below:
So the Newsweek published results of 52% Bush, 41% Kerry, 3% Nader and 4% Other do not match the 'rounded' results in the raw data table. The political affiliation demographic breakdowns do not match either, so Newsweek/PSRAI must have re-weighted the Rep/Dem/Ind factors. This makes sense as a breakdown of 37.10% Republicans, 30.06% Democrats and 32.84% Independents does not appear to be a realistic weighting for a national presidential poll.
From the Newsweek/PSRAI data table, we can 'back' into the re-weighted data table, and that table is displayed directly below:
Looks like everything matches up, so we end up with Newsweek/PSRAI re-weighting their polling data as 34.62% Republican, 32.24% Democrat and 33.13% Independent. Better, but still too many Republicans for registered voters as compared to Democrats. Still, this can be used to get the "Bush Bounce" that we have all been waiting for since the "Kerry Bounce" last month. Remember that 'lead' is different from 'bounce'. A 'lead' is how far candidate A is ahead of candidate B in a current poll. A 'bounce' is defined as the difference between: the change in candidate A's poll results in two successive polls and the change in candidate B's poll results in the same two successive polls.
Now before continuing, please notice that Newsweek last conducted a national presidential preference poll July 29-30, 2004 during the Democrat convention. This should be used as the last Newsweek/PSRAI reference poll to compute the "Bush bounce", but Newsweek will have none of that... Instead they state that a Pew Research/PSRAI poll conducted from August 5-10, 2004 should be used as the base-line poll to compute the "Bush bounce". This is because Newsweek did not conduct its own poll during mid-August to gauge any slippage in the "Kerry bounce" from the Democrat convention, any effect of the "Swiftboat Vets for Truth" controversy. They just decided to use the Pew Research poll of August 5-10, 2005 because that organization teamed with PSRAI, Newsweek's normal polling partner.
Three-Way Trial Heat Total Bush Total Kerry Total Nader Undec./ /Cheney /Edwards /Camejo Other WHO Trends ----------------------------------------------------------------------------------- Pew Research/PSRAI (08/05-10/04) 45 47 2 6 =100 Newsweek/PSRAI (07/29-30/04) 42 49 3 6 =100 Newsweek/PSRAI (07/08-09/04) 44 47 3 6 =100
The following information was published by Newsweek to qualify the use of their previous polls to compute any "Bush bounce". How interesting...
"Note: August 5-10, 2004 trend from Pew Research Center. July 29-30, 2004 poll conducted during the week of the Democratic National Convention and should not be used to estimate Bush's post-convention bounce. Trends before 7/8-9/04 based on slightly different question wording that did not include vice presidential candidates' names."
For the following "bounce" analysis I will not follow the Newsweek "advice" listed above for two reasons: First, they did not arrange or pay for the August 5-10, 2004 Pew Research poll, Pew Research did so. Second, there may have been different polling requirements that Pew Research required or that were different than the polling standards/requirements that are used by Newsweek itself when it engages PSRAI to conduct its polls. If Newsweek wants a mid-August poll before the Republican National Convention, they should pay for their own - or used Gallup. The reason that Newsweek wanted the additional polling data point would be to ameliorate some of the drastic swings in their political party affiliation breakdowns for the past few polls. They need to pay the price for such manipulation...
Here are the latest political affiliation breakdowns for the past five Newsweek/PSRAI polls. If the poll result percentages in the table show below are rounded off so as not to have a fractional component, the results should match the published polls for each polling organization. The fractional differences in the table poll results are due to the calculations done to distribute the sample number of Registered voter responses which was necessary to avoid any great inaccuracies due to the rounding already performed by the polling organizations.
Also note the computation of the "Convention bounce" for both John Kerry last month and now George Bush for this month. Bush got the "big bounce" that Newsweek was trying to limit by using the Pew Research/PSRAI poll from August 5-10, 2004.
For his convention Kerry got a 4.63% 'bounce' and ended up with a 7.63% lead in the Newsweek polls. Bush got an 18.44% 'bounce' and ended up with a 10.81% lead in the Newsweek polls.
Now if you look carefully at the five Newsweek political party affiliation breakdowns shown above you will note that in the party affiliation breakdowns, Republicans are quite under-represented in the July 29-30, 2004 poll which was taken during the DNC. Also note that the Republicans are over-represented in the September 2-3, 2004 poll taken during the RNC. I decided to re-weight the last two convention polls myself, using the political affiliation weightings from the Newsweek July 8-9, 2004 poll. This allows me to use an actual set of Newsweek/PSRAI weightings that appears to have the all important 1% to 2% advantage in Democrats over Republicans in the registered voter category. The use of these re-weightings will lower both the Kerry and Bush post-convention 'bounces', but are probably a more fair comparison that the skewed Newsweek numbers.
One of my assumptions that some might disagree with is that I feel that the political party affiliation sample composition should remain fairly consistent. Some believe that from month-to-month people re-identify with another party depending on how a particular candidate is faring in the polls. If that were so, then there would not have been any "Reagan Democrats" in the 1980s, they would have converted and become "Reagan Republicans". They may have voted for Ronald Reagan, but they still voted for Democrat congressional candidates, so to me they remain Democrats. I believe that party affiliation changes are slow, not detected nor truly documented by monthly fluctuations by polling organizations.
Without further ado, here are the re-weighted Newsweek/PSRAI convention 'bounces' and leads...
Using the revised data, at the DNC convention Kerry gets a revised -2.17% 'bounce' and ends up with a revised 0.84% lead in the Newsweek polls. Bush got an revised 15.73% 'bounce' and ends up with a revised 8.11% lead in the Newsweek polls.
I have created a temporary 'ping' list to many who had expressed interest in these R/D/I poll breakdowns. If you wish to be added (or removed) from this 'ping' list, please place a reply on this thread and I will attempt to clean-up this list so as to not bother those who do not wish to read this type of poll analysis.
Just my take...
dvwjr
You do such excellent work.
Trust me on this, because
I can actually READ and UNDERSTAND
the information as provided by you. :)
Thank heavens you netted it out!..I think 8 is a nice bounce.
Oversampled Republicans?
Lead not bounce!
Swing states 51 42
Incredible numbers : )))
That's a bounce differential of 17.90 points.
"Figures don't lie, but liars can figure."
There certainly are a lot of ways to deliver the poll news. A good reminder that we need to take them with a grain of salt.
What a lot of work dvwjr.
Bump!
It looks like quite a few Democrats will vote for President Bush - 14 %.
Excellent work and I did correct myself to state 8% lead, not bounce is nice.
Please add me to your ping list!
Interesting analysis.
ping me. thank you for the analysis.
Please leave me ON your list. This is absolutely facinating to me.
Look at the last three of the five Newsweek/PSRAI polls. Note the big dip in Republicans during the Democrat Convention. Down by 5.4%, the Dems went up by 2.9% and Kerry got a 49% to 42% lead over Bush.
Now look at September, it swings the other way, now too few Dems and too many Republicans. Makes Bush look better than he is and would allow for Newsweek to portray John Kerry as 'charging' back in the next few weeks.
That's why the political affiliation breakdowns are so important, to prevent the 'whip-saw'...
dvwjr
BTW, a DU poster even called your analysis 'brilliant'. I will leave it up to whether that is a compliment or not.
I have thought the same thing - that the oversampling of Republicans will make a "Kerry comeback" apparent in the next polls. The MSM know that Bush is ahead, but want to show him with negative momentum around the time of the election, so they are reporting the most favorable Bush polls now.
Whether the oversampling is intentional or not, I expect that we will see Kerry's numbers improve a bit, and the media will plaster his comeback all over the headlines.
If DU can understand this analysis, than that is a true compliment. I don't trust a single thing coming from Newsweak or Time. Good or bad. That is one major difference between FR and DU. On this site, even good news is looked at and analyzed with a critical eye. The objective is to find the truth. On DU, anything that's "bad" news is a Karl Rove lie, and anything "good" is obviously true.
Excallent!
Ak! ka! ka! ka! ka! ka! ka! ka! ka!
Disclaimer: Opinions posted on Free Republic are those of the individual posters and do not necessarily represent the opinion of Free Republic or its management. All materials posted herein are protected by copyright law and the exemption for fair use of copyrighted works.