HOW THE GENDER GAP, SEXISM, AND MISOGYNY CHANGE THE OUTCOME IN INDIVIDUAL STATES

The gender gap in presidential elections is a well documented phenomenon. Almost without exception, women vote for Democratic candidates at a higher percentage than do men, while men vote for Republican candidates at a higher percentage than women. Moreover, men’s preference for Republican candidates is more pronounced than women’s preference for Democrats. In the 2004 election, nationally men gave Bush an 11 point lead (Bush 55%, Kerry 44%), while women gave Kerry a mere 3 point lead (Bush 48%, Kerry 51%).

Data from a recent (conducted Feb 26-28, released March 7) 50 state Survey USA (SUSA) poll of registered voters shows more than just a gender gap; it provides evidence that sexism and misogyny can have a profound impact in shaping electoral outcomes. The gender gap is nearly double the size when the matchup is John McCain vs. Hillary Clinton than when it is McCain
and Barack Obama. McCain/Clinton shows a 13.5% gap in registered voters and a 7% gap when it is McCain/Obama.

The source of the gap is unmistakable – it is due to male voting patterns. Women remain more loyal to the Democratic nominee than do the male voters regardless of the nominee while men are more likely to switch parties and vote for McCain if the nominee is Clinton than if the nominee is Obama.

Part I of this series focused on overall national trends, Part II looked at how those trends varied between regions. Part III will examine the way misogyny and sexism affects outcomes in specific states as a means of demonstrating how the dynamic of misogyny and sexism works.

The individual state cross tables from the SUSA 50 state general election poll demonstrate how the gender gap determines the winner in a very large number of states. (An alphabetical listing of data from all states can be found at Table G8 here)

In 17 states, the gender gap plays no role in determining the winner in a McCain v Clinton contest.

· Clinton loses both the male and female vote in 11 states
· Clinton wins both the male and female vote in 6 states.

In the remaining 33 states, the size and distribution of the gender gap determines who wins a McCain v Clinton matchup.

· In 17 states, the gender gap favors McCain
· In 13 states the gender gap favors Clinton.
· In 3 states the gender gap results in a tie, with male preference for McCain balanced by female preference for Clinton

In a McCain v Obama matchup, the gender gap plays a less significant role.

· In 14 states, Obama loses both the male and female vote,
· In 14 other states, Obama wins both the male and female vote..
· In 11 states the gender gap favors McCain
· In 9 states the gender gap favors Obama
· In 2 states the gender gap results in a tie.

.

But there are only 16 states where the gender gap results in the same outcome for both Clinton and Obama. (For details see Table G8A here)

Table G8A Summary— States where the outcomes
 are the same for Clinton and Obama, and gender
 support was not dependent upon the Democratic
 candidate (margins are affected) 
Num of	Winning	Clinton	Clinton	Obama	Obama
States	Party	Male	Female	Male	Female
6	GOP	lose	lose	lose	Lose
3	Dem	win	win	win	win
5	GOP	lose	win	lose	win
2	Dem	lose	win	lose	win

· In 6 states (GA, WY, UT, ID, IN, and AL) both Democrats lose both the male and female vote.
· In 3 states (RI, MA, and OH) both Democrats win both gender categories.
· In 5 states (KS, SC, NC, MO, and TX both Democrats lose while winning the woman’s vote and losing the men’s vote.
· In 2 states (WI, MN) both Democrats win by winning the woman’s vote and losing the men’s vote.

There are also 18 states where the outcomes are the same, but one candidate wins both the male and female vote, while men and women vote differently for the other candidate. In these states, the size of the gender gap results in a win or loss for one or both of the Democratic candidates. (For details see Table G8B here)

Table G8B Summary— States where the outcomes
 are the same for Clinton and Obama, but gender
 support was dependent upon the Democratic
 candidate (margins are affected) 
Num of	Winning	Clinton	Clinton	Obama	Obama
States	Party	Male	Female	Male	Female
5	GOP	lose	win	lose	lose
4	GOP	lose	lose	lose	win
8	Dem	lose	win	win	win
1	Dem	win	win	lose	win

· In 5 states (AZ, KY, LA, MS, and OK) both Democrats lose, with Clinton losing the male vote and winning the female vote, and Obama losing both the male and female vote.
· In 4 states (AK, MT, NE, and SD) both Democrats lose, with Obama losing the male vote and winning the female vote, and Clinton losing both the male and female vote.
· In 8 states (CA, CT, DE, HI, IL, MD, ME, and VT) both Democats win, with Clinton losing the male vote and winning the female vote, and Obama winning both the male and female vote.
· In 1 state (NY) both Democrats win, with Obama losing the male vote and winning the female vote, and Clinton winning both the male and female vote.

There are 5 states where one of the Democrats tie with McCain. The size of the gender gap determines the outcome in all of these states. (For details see Table G8C here)

Table G8C-- States Where the Gender Gap results in a tie 
 for one candidate 
	HRC v	Clinton	Clinton	BHO v	Obama	Obama
States	McCain	Male	Female	McCain	Male	Female
	margin	margin	margin	margin	margin	Margin
MI	tie	-19%	17%	1%	-8%	10%
NM	tie	-11%	12%	7%	-2%	16%
TN	tie	-6%	4%	-16%	-14%	-11%
VA	-10%	-19%	0%	tie	-5%	5%
NJ	5%	-9%	17%	tie	-16%	13%

· In Tennessee, while the gender gap results in a tie for Clinton, it has no impact on Obama because he loses both the male and female vote.
· In Michigan, winning the female vote and losing the male vote results in ties for Clinton, and wins for Obama.
· Winning the female vote and losing the male vote results in a loss for Clinton in Virginia, a win for Clinton in New Jersey, and ties for Obama in both states.

Finally, there are 11 states where Obama wins and Clinton loses (7 states) or Clinton wins while Obama loses (4 state). In two of these states, the winner wins both genders, and the loser loses both genders. Iin the other 9 states, the distribution of the gender gap determines who wins and loses (For details see Table G8D here)

Table G8D Summary— States where the outcomes differ
 for Clinton and Obama because of the gender gap
	Clinton	Clinton	Clinton	Obama	Obama	Obama
States	outcome	Male	Female	outcome	Male	Female
4	lose	lose	win	win	lose	win
2	lose	lose	win	win	win	win
ND	lose	lose	lose	win	win	win
AR	win	win	win	lose	lose	lose
FL	win	win	win	lose	win	lose
PA	win	lose	win	lose	lose	tie
WV	win	lose	win	lose	lose	lose

· Obama wins 4 states (CO, IA, NH, and OR) that Clinton loses where both lose the male vote, and win the female vote.
· Obama wins 2 states (WA and NV) that Clinton loses where Obama wins both the male and female vote, and Clinton wins the female vote but loses the male vote.
· In North Dakota, Obama wins by taking both the male and female vote, Clinton loses while losing both the male and female vote.
· In Arkansas, Clinton wins by taking both the male and female vote, Obama loses while losing both the male and female vote.
· In the other 3 states (FL, PA, and WV) that Clinton wins but Obama loses, Clinton wins the female vote, while Obama loses or ties for the female vote.

HOW MISOGYNY AND SEXISM DETERMINES THE OUTCOME IN INDIVIDUAL STATES

(Note: The data is “weighed” to reflect the gender distribution used in the SUSA polls, and reflects the percentage of all voters, rather than distribution within separate gender categories. This “weighing” provided a clearer picture of how the gender gap influences the outcome in the various states. It should be noted that apparent discrepancies in the data are due primarily to increases/decreases in the “undecided” vote, and to a lesser extent to the extension of the weighed data to an extra tenth of a percent – all SUSA data was in whole number percentages. See appendix 1 below for a fuller explanation of the “weighing” process. Weighed and unweighed data for all states can be found at Table G9. Note this is a very large file, and takes time to load.)

The following sections looks at the states where sexism and misogyny result in different outcomes when Clinton and Obama are matched against McCain. While there are many states where sexism and misogyny affect only the margin by which McCain wins or loses when matched against Clinton or Obama, the states below were chosen to illustrate where and how McCain wins when matched against Clinton, but loses when matched against Obama, due to the impact of the gender gap, sexism, and misogyny . (Links to SUSA polling data can be found at the beginning of each state’s analysis under their respective tables. Tables for all 50 states can be found here, and links to SUSA data for all 50 states can be found at the end of Part II of this series.)

The best place to start is in a state where there is practically no evidence of sexism, yet the outcome changes when the Democratic candidate changes. In Iowa, while there is a significant gender gap, the size of that gap does not change substantially when the outcome changes – both men and women switch their preference from the Republican (when matched McCain against Clinton) to the Democrat (when McCain is matched against Obama) in almost equal numbers

IOWA
	McCain v Clinton
	Column	A	B	C	D
line	Cand.	All	Male	Female	G. Gap
1	McCain	46%	25.5%	20.5%	+05.0%
2	Clinton	41%	18.5%	23.0%	-04.5%
3	Margin	-05%	-07.0%	+02.5%	+09.5%
	McCain v Obama
4	McCain	41%	23.0%	18.5%	+04.5%
5	Obama	50%	22.5%	27.0%	-04.5%
6	Margin	+09%	-00.5%	+08.5%	+09.0%
	Clinton margin minus Obama margin
7	Change	-14%	-06.5%	-06.0%	+00.5%

In Iowa, both men and women both men and women prefer Obama to Clinton when matched against McCain by nearly identical margins (Row 7, Col B and C)

· Women preferred both Democrats to McCain (Col C)
· Obama does significantly better among women than Clinton, gaining an extra 6 pt greater advantage of the overall vote (Row 7, Col C)
· Men preferred McCain over either Democrat (Col B)
· The difference among males in Obama v McCain is almost non-existent, comprising 0.5% of the overall vote (Row 6, Col B)
· The difference among males in Clinton v McCain is significant, comprising 7.0% of the overall vote (Row 3, Col B)
· The gender gap when its McCain v Clinton is 9.5% of registered voters (Row 3, Col D)
· The gender gap when its McCain v Obama is 9.0% of registered voters (Row 6, Col D)
· The gender gap is only 0.5% of registered voters when its Obama, rather than Clinton, running against McCain (Row 7, Col D) (

While there is clearly gender based voting patterns in evidence in Iowa, the reason Clinton loses to McCain by 5 points (Row 3, Col A) and Obama beats McCain by 8 points is a nearly identical shift of the votes in both genders. As will be shown below, this is the exception, rather than the rule, in states where the outcome changes.

The next three states show the clearest examples of how major shifts in male voter preference can affect the outcome of the vote, even as the change in women’s vote is relatively small. In all three states, women prefer both Democrats to McCain, and men prefer McCain to both Democrats, but it is the difference in the male vote when the Democrat is a male or a female that determines the outcome. The most extreme case is Colorado, which will be examined first, followed by the least extreme case, New Hampshire, and then Oregon, which falls between the two. In each state, the gender gap is about twice that when its Clinton v McCain rather than Obama v McCain.

COLORADO
	McCain v Clinton
line	Cand.	All	Male	Female	G. Gap
1	McCain	48%	28.8%	19.2%	+09.6%
2	Clinton	42%	12.5%	29.6%	-17.2%
3	Margin	-06%	-16.3%	+10.4%	+26.7%
	McCain v Obama
4	McCain	41%	23.0%	18.2%	+04.8%
5	Obama	50%	20.2%	29.6%	-09.5%
6	Margin	+09%	-02.9%	+11.4%	+14.3%
	Clinton margin minus Obama margin
7	Change	-15%	-13.4%	-01.0%	+12.4%

In Colorado women significantly preferred the Democrat over McCain

· Women provide both Democrats with an advantage in the overall vote of more than 10%
· There is no difference in how women voted for Clinton and Obama
· Slightly fewer women voted for McCain when matched against Obama than against Clinton
· The gender gap between Clinton and McCain is 26.7% of registered voters

There is a massive difference in how men voted that results in a 15pt shift in the overall vote.in Obama’s favor

· Men preferred McCain over either Democrat
· The difference among males in Obama v McCain is minimal, comprising 2.9% of the overall vote
· The difference among males in Clinton v McCain is huge, comprising 16.3% of the overall vote
· The gender gap between Obama and McCain is 14.3% of registered voters

In Colorado, the gender gap in McCain v Clinton is nearly double that in McCain v Obama, and as a result Obama leads McCain by 6 pots, while McCain leads Clinton by 9 pts.

NEW HAMPSHIRE
	McCain v Clinton
line	Cand.	All	Male	Female	G. Gap
1	McCain	49%	27.9%	21.4%	+06.5%
2	Clinton	41%	15.7%	25.0%	-09.3%
3	Margin	-08%	-12.3%	+03.6%	+15.8%
	McCain v Obama
4	McCain	44%	24.0%	20.4%	+03.6%
5	Obama	46%	20.6%	25.0%	-04.4%
6	Margin	+02%	-03.4%	+04.6%	+08.0%
	Clinton margin minus Obama margin
7	Change	-10%	-08.8%	-01.0%	+07.8%

New Hampshire resembles a less extreme version of Colorado. As in Colorado women significantly preferred the Democrat over McCain

· Women provide both Democrats with an advantage in the overall vote of 3.6% or more
· Both Democrats received the same level of support from women
· Slightly fewer women voted for McCain when matched against Obama than against Clinton, giving Obama an addition 1% edge in the overall vote

There is a significant difference in how men voted that results in a 10 pt shift in the overall vote.in Obama’s favor

· Men preferred McCain over either Democrat
· The difference among males in Obama v McCain is small, comprising 3.4% of the overall vote
· The difference among males in Clinton v McCain is quite large, comprising 12.3% of the overall vote

The near doubling of the gender gap (from 8% to 15.8% of registered voters) when Clinton, rather than Obama, is matched against McCain allows Obama to achieve a small lead over McCain of 2 pts, while McCain has a far more significant lead (8 pts) over Clinton.

OREGON
	McCain v Clinton
line	Cand.	All	Male	Female	G. Gap
1	McCain	48%	28.5%	20.0%	+08.5%
2	Clinton	43%	16.5%	26.0%	-09.5%
3	Margin	-05%	-12.0%	+06.0%	+18.0%
	McCain v Obama
4	McCain	41%	23.0%	17.5%	+05.5%
5	Obama	49%	22.5%	26.0%	-03.5%
6	Margin	+08%	-00.5%	+08.5%	+09.0%
	Clinton margin minus Obama margin
7	Change	-13%	-11.5%	-02.5%	+09.0%

Oregon shows the same general pattern as Colorado and New Hampshire, and is between those two states in terms of the impact of changes in the gender gap. As in those states women significantly preferred the Democrat over McCain

· Women provide both Democrats with an advantage in the overall vote of 6% or more
· Both Democrats received the same level of support from women
· Fewer women voted for McCain when matched against Obama than against Clinton, giving Obama an addition 2.5% edge in the overall vote

There is a significant difference in how men voted that results in an 11.5 pt shift in the overall vote.in Obama’s favor

· Men preferred McCain over either Democrat
· As in Iowa, the difference among males in Obama v McCain is minimal, comprising 0.5% of the overall vote
· The difference among males in Clinton v McCain is quite large, comprising 12.0% of the overall vote

In Oregon, the gender gap literally doubles (from 9% to 18% of registered voters) when Clinton is on the ballot, and as a result Obama achieves a significant lead over McCain of 8 pts, while McCain has a smaller but still significant lead of 5 pts over Clinton

The next three states to be examined show what happens when men change their preference depending upon who the Democratic candidate is, but women remain with the Democrat (Nevada, Washington), or when both men’s and women’s preference change with the change in the Democratic candidate (North Dakota). In all three cases, the gender gap expands as male voters switch in large numbers to McCain when Clinton is the Democrat.

NEVADA
	McCain v Clinton
line	Cand.	All	Male	Female	G. Gap
1	McCain	49%	29.1%	20.6%	+08.5%
2	Clinton	41%	16.8%	24.5%	-07.7%
3	Margin	-08%	-12.2%	+03.9%	+16.2%
	McCain v Obama
4	McCain	41%	21.9%	19.1%	+02.8%
5	Obama	46%	21.9%	24.0%	-02.1%
6	Margin	+05%	+00.0%	+04.9%	+04.9%
	Clinton margin minus Obama margin
7	Change	-13%	-12.2%	-01.0%	+11.3%

Unlike the previous four states that have been examined so far, Nevada’s gender gap when Obama is matched against McCain (+4.9%) is smaller than the 50 state average of +7%.

· Women in Nevada showed essentially the same level of support for whichever Democrat was matched against McCain
· Obama’s margin among women increases by 1% because of an increase in undecided women voters. While a slightly smaller number of women supported Obama than supported Clinton, McCain’s support dropped three times that amount when McCain was matched against Obama.
· Male support was evenly split when Obama was matched with McCain.
· Obama got a net gain of 12.2% of registered voter over Clinton due to the male vote
· The undecided voters in Obama v McCain totaled 13% of the sample, significantly higher than the 9.3% 50 state average.
· Only two states, NJ and AZ, had a higher percentage (14%) of “undecided”, three other states, KY, MT, and NE were also at 13%.
· NV is the only state with such a high percentage of undecided voters in Obama v McCain where Obama has a lead (he ties in NJ)

The reduction in the gender gap in Obama v McCain to less than a third of that in Clinton v McCain makes it possible for Obama to be 5 pts ahead of McCain in Nevada, where Clinton lags behind McCain by 8 pts.

WASHINGTON
	McCain v Clinton
line	Cand.	All	Male	Female	G. Gap
1	McCain	46%	25.5%	20.0%	+05.5%
2	Clinton	44%	19.0%	25.0%	-06.0%
3	Margin	-02%	-06.5%	+05.0%	+11.5%
	McCain v Obama
4	McCain	38%	20.0%	17.5%	+02.5%
5	Obama	52%	25.5%	26.5%	-01.0%
6	Margin	+14%	+05.5%	+09.0%	+03.5%
	Clinton margin minus Obama margin
7	Change	-16%	-12.0%	-04.0%	+08.0%

In Washington,there is a significant net increase from both men and women when Obama is matched against McCain.

· Women support both Obama and Clinton over McCain.
· Obama’s greater support among women relative to Clinton comes from an increase of 1.5% of registered voter in actual support for Obama, and a decline in support for McCain representing 2.5% of registered voters.
· Among men in Washington, the level of support for McCain and the Democrat essentially “flips”
· Obama’s gets the same level of male support that McCain does when he is matched against Clinton, and McCain gets only 1% more of the vote against Obama, than Clinton did against McCain.
· As a result of this “flip” Obama’s greater support than Clinton among males is equal to 12% of registered voter.
· The gender gap in Obama v McCainin Washington is only half that of the national average of 7% of registered voters.
· The gender gap in for Washington in Clinton v McCain (11.5% of registered voters) is only slightly lower than the national Clinton v McCain average of 13.5%

In Washington, the gender gap increases over 250% (from 3.5% to 11.5% of registered voters) when Clinton, rather than Obama, is the Democratic candidate. This increase is caused by the “flip” in male voters, and is the primary reason why Clinton is close to, but remains behind McCain (by 2 pts) in Washington, while Obama’s lead (at 14 pts) is well into the double digits.

NORTH DAKOTA
	McCain v Clinton
line	Cand.	All	Male	Female	G. Gap.
1	McCain	54%	29.5%	24.5%	+05.0%
2	Clinton	35%	14.0%	20.5%	-06.5%
3	Margin	-19%	-15.5%	-04.0%	+11.5%
	McCain v Obama
4	McCain	42%	22.0%	19.5%	+02.5%
5	Obama	46%	22.5%	24.0%	-01.5%
6	Margin	+04%	+00.5%	+04.5%	+04.0%
	Clinton margin minus Obama margin
7	Change	-23%	-16.0%	-08.5%	+07.5%

If there is one state that falls into the category “shocking”, North Dakota would probably be it. Bush beat Kerry here in 2004 by a 63% to 36% margin, and the fact that any Democrat could look competitive here, let alone be ahead of McCain, has to be seen as a tribute to Obama’s cross-cultural appeal.

· The shift in the female vote represents a net loss for Clinton of 8.5% of registered voters
· North Dakota is the only state where such a large shift in the female vote helps change the outcome between Clinton v McCain and Obama v McCain.
· While three other states (HI, ID, UT) have larger female vote shifts toward Obama, in none of these states does the preference of the majority of women change, nor does the overall outcome change.
· In Hawaii, women still favor both Democrats, and both Clinton and Obama defeat McCain, in Idaho and Utah, women favor McCain over either Democrat, and McCain beats both Democrats.
· The shift in the in North Dakota male vote toward Obama is almost twice that of the female vote, representing 16% of registered voters.
· These shifts mean that Democrats go from a huge loss of the male vote, and a small loss of the female vote when its Clinton v McCain, to a very small win among males, and a significant win among females, when its Obama v McCain.
· Only nine states, including ND, have vote shifts greater than 20%, and ND is the only one that provides Obama’s with a margin of victory. (The two large shifts in Clinton’s direction in Arkansas and West Virginia both change the outcome in her favor.)

While the gender gap nearly triples between Obama v McCain and Clinton v McCain, this change in the gap is only partially responsible for Obama’s leading McCain while Clinton is well behind McCain. In North Dakota, the major shift in the female vote is equally responsible for Obama’s lead.

These seven states – Iowa, Colorado, New Hampshire, Oregon, Nevada, Washington, and North Dakota – are the only ones where Obama wins a match-up with McCain, while Clinton loses against McCain. And while this is a specific subset of states, it accurately describes the impact of sexism and misogyny on the 2008 general election.

And while these seven states demonstrate the dynamic of sexism and misogyny, in many states this dynamic is complicated by the presence of racism. But racism does play a role in how people vote, and the interplay between racism and sexism is especially complex. Part IV of this series will examine that phenomenon.

APPENDIX 1 – WEIGHED AND UNWEIGHED DATA

In the above state-by-state analysis, the original polling data was “weighed” according to the proportion of males and females in each state’s samples. This is done for a number of reasons:

· The weighed data provides a clearer picture of how the overall electorate behaves by providing percentages of the electorate in general, rather than providing isolated views of demographic categories.
· This process is not crucial when looking at gender distribution, because the difference in the relative size of two categories is insufficient to have a truly major impact on the overall outcome. One can simply take a simple average of the male and female vote, and get a fairly good representation of overall voter behavior.
· This is especially true with regard to the sample being used by SUSA, which looks at registered voters, rather than “likely voters, and there is far less variation in gender distribution among registered voters in each state than there is among actual (or likely) voters.
· However, weighing is crucial in other demographic categories such as race and party affiliation, where there are usually large differences in the size of each group. You cannot simply take the average of how Democrats, Republicans, and Independents voted, and get a number that is close to the overall results.
· In general, weighing the data provides a far more accurate picture of changes in the electorate. While apparent discrepancies do arise when using weighed data, those are usually caused by increases/decreases in the “undecided” vote.
· It should be noted that in 4 of the 7 states examined above (IA, ND, OR, WA), “weighing” had no impact on the outcome, because the of a 50-50 split in gender categories
· Finally, it should also be noted that this data from the SUSA poll was from registered voters, and that the gender distribution in the actual election will be different. The table below shows how SUSA broke down the vote by gender in the seven states cited above, and the gender breakdown from CNN exit polls from the 2004 elections for those same states.

	SUSA poll	CNN 2004
	(2/26-28/08)	exit poll
	gender	gender
	breakdown	breakdown
	Male	Female	Male	Female
CO	48%	52%	44%	56%
IA	50%	50%	46%	54%
ND	50%	50%	49%	51%
NH	49%	51%	51%	49%
NV	51%	49%	48%	52%
OR	50%	50%	45%	55%
WA	50%	50%	45%	55%

To illustrate how weighed data gives a more accurate picture, we’ll examine the McCain v Clinton contest in Colorado, where the gender breakdown was 48% male, 52% female.

 COLORADO:  MCCAIN V CLINTON
	Unweighted Data (averaging male and female)
col		A	B	C	D	E
					Average	Error
line		All			Male &	From
		Voters	Male	Female	Female	Average
1	McCain	48.0%	60.0%	37.0%	+48.5%	-00.5%
2	Clinton	42.0%	26.0%	57.0%	+41.5%	+00.5%
3	Margin	-6.0%	-34.0%	+20.0%	-07.0%	+01.0%
	Weighed Data (adding male plus female)
					Male	Error
		All			Plus	From
		Voters	Male	Female	Female	Adding
4	McCain	48.0%	28.8%	19.2%	+48.0%	-00.0%
5	Clinton	42.0%	12.5%	29.6%	+42.1%	-00.1%
6	Margin	-6.0%	-16.3%	+10.4%	-05.9%	-00.1%
	Error from averaging unweighed data versus 
adding weighed data
			Unweighed data	Weighed data
			Average	Error	Male	Error
		All	Male &	From	Plus	From
		Voters	Female	Average	Female	Adding
7	McCain	48.0%	+48.5%	-00.5%	+48.0%	-00.0%
8	Clinton	42.0%	+41.5%	+00.5%	+42.1%	-00.1%
9	Margin	-6.0%	-07.0%	+01.0%	-05.9%	-00.1%

As the above chart shows, a failure to weigh the data properly results in a distortion of the actual results. Simply averaging the male and female data results in a margin that is one point greater (Row 9, Cols B&C) than was actually reported (Row 9, Col A). Whereas when the data is weighed, the difference is only one tenth of one percent (Row 9, Col D&E) which rounds to the margin that was reported.

There is, however, a noticeable flaw in the weighed data, because the male vote adds up to only 41.3% instead of 48%, and the female vote adds up to only 48.8% instead of 52%. This is because 14% of males, and 6% of females, were “undecided” in the McCain v Clinton matchup. When those percentages are weighed, you get an additional 6.7% on the mal side, and and additional 3.1% on the female side. By adding the “undecideds” in each gender category, you get the appropriate gender breakdown (males 41.3% + 6.7% = 48.0%, females 48.8% + 3.1% = 51.9%, which rounds to 52%)

In Colorado, because of how “undecided” voters behaved the distortion had only minor impact on the outcomes reported when the comparison was made between McCain v Clinton and McCain v Obama.

LINE 7 FROM COLORADO STATE BY STATE
COMPARISON ABOVE
	Clinton margin minus Obama margin
7	Change	-15%	-13.4%	-01.0%	+12.4%

The overall change in the margin, based on unweighed SUSA data, was -15%, while the weighed data, adding the shifts in both the male (-13.4%) and female (-1.0%) vote, showed a change of -14.4%. (Undecided voters overall declined by only 1% from McCain v Clinton to McCain v Obama, with the percentage of undecided male voters declining from 14% to 10%, and the percentage of undecided female voters increasing from 6% to 9%. As a result, the male decrease/female increased tended to balance themselves out in Colorado.)

However, this is not always the case. In Iowa, undecideds overall dropped from 12% to 9% when McCain v Clinton was compared to McCain v Obama, and both male and female undecideds being lower in the McCain v Obama contest. As a result, the final numbers are not nearly as close.

LINE 7 FROM IOWA STATE BY STATE
COMPARISON ABOVE
	Clinton margin minus Obama margin
7	Change	-14%	-06.5%	-06.0%	+00.5%

As a result, while the unweighed SUSA data shows a change in the margin of –14% of voters, the combined shift in the male(-6.5%) and female (-6.0%) votes add up to only -12.5%. Again, this is due to a change in the undecided vote.

Because this study is designed to contrast male and female voting patterns in the aggregate, and measure the change in how each gender splits their votes when the gender of the candidate opposing McCain changes and how the gender gap widens when a male candidate is substituted for a female candidate, these apparent discrepancies remain in the tables.

They also remain there as a reminder that this is simply polling data that has a margin of error of plus/minus 4%, and that no percentages can be considered a precise reflection of reality. Instead, this data is being presented because of the overall trends that it represents—-trends that appear in the national averages examined in Part I, that are confirmed in the regional averages examined in Part 2, and show themselves at work in the data from the individual states examined in Part 3.

APPENDIX 2 – LIMITATIONS OF THE DATA

The SUSA national poll represents a snapshot in time. The data was collected on February 26 through 28, 2008, at a point where (according to the national trends lines from pollster.com) Obama was at his peak when matched against McCain, and Clinton was at her lowest point when matched against McCain. (Since that time, the trend lines show that Obama has fallen relative to McCain, while Clinton has risen relative to McCain) This was before the Texas and Ohio elections on March 4 took place, and in a media environment where most pundits were predicting a victory for Obama in Texas, and a close race, if not an upset for Obama in Ohio. And of course, the data was collected before the recent controversy surrounding Obama’s association with Dr. Jeremiah Wright, and the well-received speech on race relations given by Obama in response to that controversy. Subsequent polls have showed significant shifts in voter sentiment since the time the SUSA poll data was gathered.

As with all such general election polling data this far from November, the predictive value of the SUSA poll is practically nil. This is especially true because it is a poll of registered voters, not “likely” voters. But the timing of the poll itself makes it practically ideal for analyzing trends concerning the electorate as a whole, especially when it concerns gender issues. While Hillary Clinton has always been defined as a woman first (wife, mother, and a symbol of gender equality) and a “politician” second, at the time the poll was taken everyone knew who Barack Obama was, but he remained essentially undefined in the minds of the most voters, and despite his race was as close to a “generic” male candidate as we are likely to see. (As will be shown in Part 4, Obama’s race has been a factor in many states, but the data shows that both gender and race are factors in most of these states.)

Finally, it should be noted that the margin of error in each state is around plus or minus four points – and that the margin of error is higher within demographic categories because the sample size of each category is smaller. Thus, data from individual states should be considered far from authoritative, and this may (or many not) be especially true in states that seem to be anomalies.

What makes the SUSA data so valuable, however, is that it offers an “apples to apples” comparison – all the data was collected at the same time, using the same questions, in the same order, for all 50 states, in a way that each state had approximately the same margin of error. This is entirely different from most “national” polls, which simply look at a national aggregate of voter preferences, and individual state polls that are taken during different time periods by different polling firms, in which the questions asked often vary in both content and order.