Chapter 4

Chapter 4. Voting Behavior and Survey Research

Last updated 9-29-2011

Public opinion pollsters are people who count the grains of sand in your bird cage and then try to tell you how much sand there is on the beach. Fred Allen

OUTLINE

Introduction

I. Problem Selection‑‑Voting Behavior

II. Theory Formulation

A. Partisanship

B. Political Involvement and Efficacy

C. Demographic Factors

D. Issues

1. Character

2. Valence Issues and Retrospective Voting

3. Ideological

a. Economic

b. Regulation

c. Social Welfare

d. Social/Moral

e. Civil Rights

f. Foreign Policy

III. Hypotheses

IV. Operationalization

V. Data Gathering‑‑Survey Research Methods

A. Sampling

TEXT

Introduction

The following module serves four purposes. First, it is a review of the steps of the scientific research process. It will take you through the steps that any political scientist follows in doing virtually any research project. You will note that the first seven major headings in this module are the steps of the scientific research process.

Second, it will familiarize you with what is involved in the planning, collection, and analysis of public opinion survey data. You should learn to be a smarter consumer of surveys as a result. All of you will be confronted with survey and polling information. You need to know how to critically evaluate it.

Third, you will become familiar with some of the findings of political science in the area of voting behavior. This will give you an idea of some of the things that political scientists study. Voting behavior is one of the most active areas of empirical research, in part because voting data are so readily available. The theories you will read about are good examples of empirical theories.

Finally, at the end of the module you have a little written exercise that will provide you with hands‑on experience in manipulating and interpreting data in order to test some very simple hypotheses. The exercise will give you an idea of what is involved in hypothesis testing. It will also show you how all the steps of the scientific research process fit together.

I. Problem Selection

The first step of the scientific research process is selecting a problem to study. Thus, this is the first major heading in this module. The problem area we will study is voting behavior, a major area of concern for many political scientists. Political scientists study voting behavior for a number of reasons. As you think about these reasons, you can see how values enter into the problem selection process.

First, it can be a good way to make a living. Enough people care enough about why citizens vote the way they do‑‑or don't vote at all‑‑that they are willing to pay political scientists as consultants to improve their chances of winning. However, once you get into this kind of work, you have really left the field of political science and entered the closely related field of campaign consulting. According to the American Association of Political Consultants, about 5,000 people make at least part of their living doing political consulting as of late 1992. That number has certainly grown even more over the last decade or so. Even though political consulting is really separate from political science, consultants study and apply voting behavior studies. Incidentally, other than through political science, people often get into the field of campaign consulting in another way. Many business marketing people engage in political consulting work in addition to their year-round business accounts. Some maintain that selling candidates is really no different than selling laundry detergent.

Second, in so far as being scientific goes and measuring things easily and accurately, we can do few things better than measuring votes. If you remember empirical theory, almost more than anything else, those who would wish to be "scientists" want to be precise. I know you think this is not a very good reason‑‑and I agree‑‑some of the most important things are not easy to measure precisely. (Do you remember the "second face of power?") But nevertheless, this is a reason why political science focuses so heavily on voting. To claim precision and expertise in an area that people care about elevates the status of political science as a scientific profession. If you remember generic politics, we could see this as political science negotiating itself into more valued categories. An element of Machiavelli is also here. The profession is attempting to enhance its public reputation.

Third, elections can tell us much about the current events and recent history, because elections reflect the tensions and concerns that shape any society, especially a republic. So by studying elections, we are studying and coming to better understand ourselves and others, depending on which elections we study. For example, in looking back at the 1992 presidential election, we must examine a wide variety of issues and questions if we want to go beyond cliches. Among them are: considering the American economy in a world context, looking back at the Vietnam War and how people balanced feelings of national obligation with concerns about the morality and wisdom of the war, discussing how government and private action and inaction affects the family, deciding what constitutes a family, and thinking about what role the national government can and should play in improving education and health care.

The presidential election in 1996 also said a lot about current events—how we had been moving toward lower deficits, the importance of a tax cut when Americans were living in an era of economic growth, welfare reform and who gets credit for its passage, and the relative importance of social issues to economic issues, the fading of the World War II generation, and other things.

Many of these issues from '92 and '96 remained issues in 2000. And in 2004 many of these issues were revisited: the human and financial cost of foreign wars, the reappearing deficits, health care and drug costs which had been soaring out of sight, and the fairness of our tax system when the differences in income between the wealthy and the middle and working classes had been growing. But above all 2004 was about the war in Iraq, and whether the nation should turn out of office a“wartime President.” Because a majority of voters saw the war at that point in time as an extension of the war on terrorism, they chose to back the incumbent.

But by 2006 the war had dragged on and with no end in sight, and the public then saw the war as no longer connected with the war on terror. And the party of the incumbent president lost their majority in both the House and Senate. If the war was not resolved, it would be the centerpiece of the 2008 election, unless some other crisis overshadowed it.

As you now know, the war was not resolved, but combat deaths in Iraq did fall, and another crisis that directly affected more families hit the nation. The near collapse of the economy was set off by the home mortgage crisis that felled banks and Wall Street financial firms that were heavily invested in the “derivitives,” which was the name for complex packages of mortgages and insurance on these mortgages and other things put together and sold as stocks. Those were the current events that help explain why Republicans lost more seats in both houses of Congress and lost the White House as well.

As of this writing two months before the 2010 congressional elections, the painfully slow recovery from the "Great Recession" will be the central explanation for significant to massive losses by Democrats on both houses of Congress. Increasing deaths in the war in Afghanistan, which the nation felt had dragged on too long, would not help the party in power. General dissatisfaction with all those in power also led to the defeat of incumbent Republicans in primaries by relatively more ideologically extreme candidates. Some of these candidates seemed unlikely to move to the center for the general election, as candidates usually do. This gave some hope to Democrats that moderate voters in the general election might save some Democrats from defeat. But would moderates come to the polls? Turnout is a key in all elections.

Fourth, elections are one very important dimension of democracy as we know it. To understand the meaning of democracy and the health of democracy, one must necessarily consider how elections operate and how and why individuals participate. Indeed, many of the changes and conflicts in the American political system have involved how our elections are run: women's suffrage, voting rights for African-Americans, the eighteen year old vote, the laws regulating the role of interest groups in elections, and the openness of political parties in nominating candidates.

Imagining democracy without some kind of elections would be hard. Indeed, the holding of elections with competing parties and choices among candidates in the Soviet Union and their satellites was a defining event in their move to democratize. Studying elections is an important part of comparative politics, an area we shall examine later in the course. However, you should keep in mind that elections certainly do not guarantee either a democracy or stability in any democracy. Again, think about all the instability within the former Soviet Union. Think of how the military reacted in East Timor when the side they supported lost the election on the question of independence in 1999. Will an election stop the bloodshed in Iraq, Afghanistan, or Israel and Palestine? A willingness to lose power is a prerequisite for both successful elections and democracy.

Finally, elections are quite a bit of fun to talk about and think about because they do have a strong element of competition in them. As you know from the Money Game simulation, Americans love competition. The media focus quite heavily on this aspect of elections‑‑many would argue too heavily. This kind of coverage is called "horse race coverage." Regardless of whether coverage should focus more on issues, elections do make exciting and entertaining human drama. Two sides are in conflict, each has ups and downs, strategy is important, and unlike other dramas or sporting events, the audience gets to choose the winner in the end (well, at least most of the time – in 2000 the Supreme Court decided the winner). Indeed, elections may be the most democratic of all sports! Nothing about this is new. People regarded elections as "contests" long before CBS News was telling us who was ahead. Many sporting and dramatic heroes go on to success in politics after they retire from their first occupation (e.g. Bill Bradley, Jack Kemp, Arnold Schwarzenegger, and of course Ronald Reagan).

II. Theory Formulation

This is the step in scientific research where you go to the library and find out what is already known about the problem area. What theories already exist about this problem? What facts are known? What hypotheses have been tested? What relationships are known to exist?

Here is where you should be creative. Think of logical relationships you can test yourself. Can you improve on existing theories by thinking of the circumstances where they should or should not work?

At this point, let's assume that we did go to the library and used the Social Science Index, DISCUS (which allows electronic searches of periodicals and journals), and the electronic card catalog (USCAN) to find books about voting behavior. Perhaps you even used the Internet. What follows are some of the things you would learn about what has been established by political scientists in the area of voting behavior. We'll stick to the fundamentals to keep it simple.

A. Partisanship

Partisanship refers to the degree to which a person identifies with a particular political party. (Often political scientists just call this "party id.") As you know, or should know (do you remember "critical election” theory?), people usually pick up their identities from their parents. The process is much the same way that they learn their religious identification, although party identity is not usually as important.

These transmissions across generations are less than perfect. As each generation passes, the identities usually get weaker. This weakening process continues until a critical election comes along and people adopt new identities of their own. Alternatively, they could have their old identities reinforced. White working class southerners had their Democratic identity reinforced by Roosevelt's New Deal program in the 1930s. Blacks became Republicans during the Civil War. Their identity as Republicans eroded across the generations. In a series of three steps, they became Democrats (the 1930s when FDR's New Deal helped them economically, 1960 when John Kennedy expressed support for Martin Luther King, who was imprisoned in Alabama, and finally in 1964 when Lyndon Johnson campaigned on a strong civil rights platform and the Republican candidate Barry Goldwater campaigned on his opposition to the Civil Rights Act of 1964). Although the shift of blacks in 1964 did not constitute a critical election because it did not create a new majority, it was a significant realignment that later led many anti‑civil rights whites to move to the Republican Party. Had more whites moved to the Republicans and stayed with them, we might have had a critical realignment. (We should note that some political scientists argued in the 1980s that a big enough white shift has happened gradually in what they call a "rolling realignment" to create a new Republican majority, at least at the presidential level. That idea is now in need of some revision.)

Because we haven't had a clear critical election in a long time, people's identities are weaker now than they used to be. Relatively more people tend to call themselves "independents." This means not identifying with any political party. The table below shows how party identification has changed in the U.S. over the last several decades. You should note that despite change, the picture remains relatively stable over time. The changes take place relatively slowly and in small increments over time. Party identification is a relatively stable variable in American politics.

Patterns of Partisan Identification In the U.S.*

Year % Democrat % Republican % Independent

strong weak lean strong weak lean

1952 22 25 10 14 13 7 6

1956 21 23 6 14 15 8 9

1960 20 25 6 14 16 7 10

1964 27 25 9 14 11 6 8

1968 20 25 10 15 10 9 11

1972 15 26 11 13 10 11 13

1976 15 25 12 14 9 10 15

1980 18 23 11 14 9 10 13

1984 17 20 11 15 12 12 11

1988 18 18 12 14 14 13 11

1992 17 18 14 15 11 13 12

1994 14 22 12 18 11 10 13

1996 14 20 12 11 17 9 16

1998 13 21 12 8 17 9 17

2002 15 19 10 11 16 7 19

2004 17 15 17 16 13 11 10

2006 15 16 12 7 14 11 22

2008 D: 42 % R: 32% Indep/other: 26%

2009 D: 36% R: 34% Indep/other: 31%

2010 D: 35% R: 34% Indep/other: 31%

Or you can look at a cool line graph that averages numbers for most all national polls on party id for the last several years at:

http://www.pollster.com/polls/us/party-id.php

* Sources: National Election Studies, Inter-Consortium for Political and Social Research, General Social Surveys, National Opinion Research Center, Rasmussen (2008-2010)

Questions: Generally speaking, do you think of yourself as a Republican, a Democrat, an independent, or what." If Republican or Democrat, "Would you call yourself a strong (R or D) or a not very strong (R or D)?" If independent, "Do you think of yourself as closer to the Republican or Democratic Party?" (leaners)

Note: The %'s do not add to 100% because of identifiers

with third parties and "don't knows."

As you can see, up until 2002 the biggest changes were fewer strong identifiers in either party, especially Democrats and more independents. The proportion of Democrats had declined from over half the electorate to well under half. The largest overall increase was "independents," who tripled since the early 1950s.

However, in 2004 Democrats were up to nearly half the electorate and the percentage of pure independents was nearly cut in half. Two counter trends seemed to be happening. The percentages of the strong identifiers increased for both parties, but the percentage of leaners also increased. These weak leaners replaced the independents. Democrats may have benefited from dissatisfaction with the Bush administration, but the Bush’s campaign theme of reelecting a wartime president who stood strong against terrorism at least temporarily kept most Republicans together and attracted enough independents and leaners to win the election.

In 2006 the impact of an unpopular Republican president was taking its toll. Democrats remained the same, but new voters and some Republicans were defecting to the ranks of the independents. The stage was being set for 2008.

I have some data for 2008 which shows that the trend toward a Democratic id continued (though it is not NES data). Adding leaners to strong and moderate Democratic identifiers comprised 51% of all adults. Republicans plus leaners were 39%, and pure independents were at 10%. Continued dissatisfaction with the war and with a declining economy greatly helped Democrats. Other polls show the Democrats with a significant lead in late 2008 -- Rasmussen showing a 10 point lead (42 to 32). But since then Democrats have slipped significantly in identification. Some has gone to Republicans to the point where in late 2010 the difference is only slight between the two parties. But more has gone to neither party as citizens have disaffection with both parties.

We would conclude from all this that the national electorate is divided very roughly into thirds among the two major parties and the "no party" identification of "independent." What this means is that on a national level, most elections are up for grabs today—that short term issues decide elections. No party has a decisive advantage, as the Democrats once did for a long time after the Great Depression.

For the purposes of our discussion here, we are limiting ourselves to talking about the Republicans and the Democrats, although you should know that there are other parties in existence. However, these other parties, often called third parties, are relatively insignificant in determining the outcomes of elections. Therefore, they get less attention in voting behavior research than the two major parties. They rarely get more than 5% of the vote. So even if we were to focus on them in a public opinion poll, we would have to draw a huge sample to get a significant number of people who identify with third parties.

We should note that these third parties are important in other ways. They express ideas and proposals that the major parties may pick up later. They can occasionally punish major parties by providing a place to go and protest for those who are angry with one of the major parties. In 1992, many business oriented Republicans and independents who might normally lean to the Republican side and who were all angry with President Bush found a way too punish him by supporting Ross Perot. Many of them came back to Bush when Perot temporarily dropped out. But a good number went to Clinton. Some went ahead and voted for Perot. It all made reelection quite difficult for the incumbent Bush. Looking back further in time to 1968, the many southern Democrats who flocked to George Wallace's American Independent Party almost surely contributed to the defeat of the Democrats (led by Vice President Hubert Humphrey) that year--though he probably would have lost regardless of Wallace, just as Bush probably would have lost without Perot. Perot was less of a factor in 1996. However, looking at the election of 2000, third party candidacy of Pat Buchanan had minimal effect on George W. Bush as he drew very few votes in states that Bush lost. But on the other side, Ralph Nader, the consumer activist who was nominated by the environmentalist oriented Green Party clearly cost Al Gore the election in Florida and consequently in the nation. It is not how many votes, but where the votes are that counts in the electoral college system. If you remember, rules are never neutral!

HOW MUCH does party affect the way people make voting choices? Party identification is the best single predictor we have for how people will vote. That was true 50 years ago when relatively more people had stronger party identifications. It is still true today. Quite simply, if you wanted to predict how a person was going to vote and could only ask them one question (other than asking them "for whom are you going to vote"), your best bet would be to ask them their party identification. This method does not work all the time, but it will predict correctly most of the time.

Under what circumstances does party identification correctly predict the vote? Well, that depends on several things, especially on which party an individual identifies with. As a rule, Republicans are more likely to stick with their party than are Democrats. Republicans are generally more loyal. Something in the range of 80 to 90% of Republicans vote for their party's candidates, while the figure for Democrats is usually between 70 and 80%. This means that Republicans can successfully compete even though they are the smaller party.

How can we explain this difference in loyalty? It is due partly to the fact that Democrats have long been the larger party (only slighter today). Being larger they encompass a wider variety of groups and interests, and therefore satisfying all the groups in their party is harder for Democrats than for the Republicans.

This rule has predictable exceptions. When a party is unable to unite itself after a nomination battle because of some highly salient issue or personal difference, they are likely to suffer more defections in the general election. The Republicans found themselves in this unhappy situation in 1992. The party split badly over the abortion issue and over how tolerant and open they would be toward those with AIDS. The same problem plagued the GOP in 1996 and was a major concern for the Bob Dole forces as they came to the Republican Convention in San Diego August right after the Olympics. The Democrats were united as no one challenged Clinton for the nomination. The enabled Clinton to spend his primary monies attacking Dole and the Republicans while Dole used all his money in just winning the nomination. George W. Bush had that advantage in 2004. He was able to spend over $50 million in primary money on television ads attacking John Kerry in July and August before he accepted the nomination on Thursday, September 2, 2004. And that does not count all that he was able to spend in the months prior to that while Kerry battled opponents for his party's nomination.

Precisely HOW does party identification affect the way people vote? Political scientists have done an immense number of studies on this seemingly simple question. Study will continue as long as political scientists and political parties exist. Let me offer you a couple of generalizations that are pretty much agreed upon by the research that has been done. First, having a strong party identification simplifies the problem of making a voting choice for people. All you need to know is the party of the candidates and you can then make your choice based on party. That does not require much effort. Gathering other information‑‑like issue information and information on character and competency‑‑is both time consuming and difficult. Having a favorite party is something like having a favorite football team. The more you love Clemson, the less you care who the opposition is. You simply pull for Clemson. Political scientists have a fancy way of saying this. They say that having a strong party identification "lowers the information costs" for voters. That is, you don't have to figure out very many things to decide who to support.

Second, party identification acts as a psychological filter for gathering and evaluating information. You tend to believe the ads run by your party's candidates and disbelieve or not even pay attention to information that is contrary to your identification. This also makes life simpler for us. The filter enables us to ignore a great deal of potentially confusing information‑‑nobody likes to be confused. Think about it! Most people turn off the tv or turn down the volume or leave the room when a commercial comes on that they don't like. Incidentally, this idea is borrowed from a psychology theory called "dissonance theory." Dissonance refers to the discomfort we feel when confronted with contradictions.

Finally, party helps us to decide even when we do look at other things. Some fairly recent research has concluded that most people use a rather simple decision rule in deciding how to vote. They add up their likes for each of the alternative candidates and choose the candidate who comes out ahead. If it comes out even, then they use party identification as a kind of tie-breaker. This works even when party identification is weak. Given that we live in a period in the U.S. when party identification is weak for many people, even this tie-breaking role for party identification is pretty significant.

B. Political Involvement and Efficacy

Political involvement is the degree to which a person is involved in political activities. It includes a wide variety of things all the way from reading and talking about politics through voting to contacting government officials and getting involved in political campaigns.

Political efficacy refers to the degree to which a person feels he or she can be effective in political matters. Do you think public officials care about what you think? Can your vote make a difference? Is politics understandable to you? If you would answer "yes" to these questions, then you have a high sense of political effectiveness. (Incidentally, a number of other variables are closely related to effectiveness that political scientists also talk about, e.g. alienation and cynicism. In these two examples, you might think of these as opposites to effectiveness.)

How do these two variables relate to each other? In short, the higher a person rates on efficacy (or effectiveness), the more likely she or he is to be politically involved, including voting.

C. Demographic Factors

A large number of demographic factors also affect how many people vote (called turnout), the direction of party identification, and therefore voting choice. By demographic factors, we simply mean the social, economic, and physical categories into which you fit. Voting is positively associated with higher education, higher income, higher social class, middle age, living in a given community for a longer period of time, being married, having children, and being non-southern. The same kinds of factors are also positively associated with a Republican identification. Those who are Catholic or who are a member of some minority group (Hispanics, Jews, and Blacks) are more likely to be Democrats in party identification. Women are also slightly more likely to be Democrats (in a March 1987 Gallup poll, 43% of women identified as Democrats versus 36% of the men; in 1992 Clinton won 5% more of the women's votes (46%) than men's votes (41%)). Going into the 1996 election some polls showed Dole running 30 points behind Clinton among female voters--a potential disaster for Bob Dole. The final difference was not that wide, but women did clearly elect Bill Clinton in 1996. These gaps remained in 2000 when Gore won seven percentage points more among women than men, and eleven percentage points more than Bush among women.

In 2004 a gender gap remained, but it was much smaller. The soccer moms had become more concerned with security. Bush got about 53% of the male vote and 48% of the female vote. Kerry got about 51% of the female vote, so while Bush was behind Kerry among females, the margin was not very large (3 percentage points). Of course the flip side is that Kerry was behind Bush among males (an 8 percentage point gap), and that gap was larger, more than enough to offset Kerry's female votes. In 2008 a gap remained as women worried about the economy and grew weary of the wars.

You should keep in mind that all of these trends are merely that‑‑statistical trends, not hard and fast rules that apply to each individual. Even among African Americans where the trend is strongest, not all are Democrats--7% consider themselves Republicans and 16% are independent.

Given that many of the same factors are associated with both turnout and a Republican identification, we can conclude that in addition to being more loyal, Republicans are more likely to vote than are Democrats. This is a second reason why Republicans can successfully compete with the larger party.

D. Issues

People are often told they should vote on the basis of issues. Indeed, saying you "vote on the issues" or "vote the person" is one of the accepted signs of being a middle class, educated, and politically sophisticated person. Very few people say they vote on the basis of party. Despite what they say, many more do vote their party than don't. Voting the party works indirectly when you account for the fact that party acts as a psychological filter in processing information about candidates and issues.

We could divide most issues in a campaign into one of three general types: character or personality issues, valence issues, and ideological issues. Party plays an indirect role on all three, but on the latter two, party plays an especially significant role, so we must necessarily discuss party in talking about these kinds of issues.

1. Character

Character issues center around the personal qualities of the individual candidates‑‑and they have been a very hot topic in both political science and the popular press for a good while now. A well known political scientist, James David Barber, predicted back in 1971 that President Richard Nixon had a badly flawed political character that would ultimately lead both he and his party to political ruin. Of course, you know from history that his prediction turned out to come true. Barber has gotten a great deal of attention from political reporters ever since. They take seriously his exhortation that they should look more into candidates' characters in reporting on campaigns. The new importance placed on character by both the press and the public contributed to the downfall of two Democratic presidential primary candidates (Gary Hart and Joe Biden) in the summer and fall of 1987. It badly scarred Bill Clinton in the primaries and general election of 1992 as reporters probed and re-probed the basis for the image he had as a "slick Willie" who tried to be all things to all people. Character issues continued to hound Clinton after he went into office right on up to and beyond the 1996 election. Some of that rubbed off on Al Gore in 2000, and may have been enough to cost him the election.

The emphasis on character has been around a long time. In one of the dirtiest campaigns of all time, Andrew Jackson was accused of being a bigamist in the 1828 election when opponents raised questions about the legality of his wife's divorce to her former husband. Jackson won, but his wife died shortly thereafter. Jackson blamed his political opponents for her death. He carried his anger with him to his own grave. Candidates have long worried about image and spent millions to build and mold the desired image.

Character played a major role in the 2004 election. Both candidates and their supporters made multiple attacks on each other, though the Republicans were far more effective in throwing mud. Kerry supporters (though not Kerry himself) charged that Bush used family influence to get in the Air National Guard ahead of others so as to avoid service in Vietnam. And then after getting in, failed to perform his duties. On the other hand, Bush supporters (though not Bush himself), charged that Kerry exaggerated his exploits that had led to his purple hearts and a silver star and then turned against his comrades in arms when he returned to the states and joined the Vietnam Veterans Against the War. Bush charged that Kerry had weak character in that he changed his positions and had seemingly inconsistent votes. Bush’s running mate, Dick Cheney, went further in charging that Kerry was so weak that his election would encourage more terrorist attacks. For his part, Kerry stuck closer to issue questions charging that Bush made wrong choices on issues. The closest that Kerry came to a direct personal attack was in his acceptance speech at the Democratic National Convention in which he implied that Bush was not very bright, saying that strength and intelligence are not mutually exclusive.

The role the press plays in covering character is somewhat new. Character judgments used to be made by party leaders and political opponents. Today with the rise of the primary system for choosing candidates, the media make these judgments as they mercilessly dig and probe every facet of the candidates' lives. Perhaps, as James David Barber argues, this new role for the media is healthy. We as voters are somewhat more protected from the manipulation of professional image-makers. As long as the media are plural, we have some protection from being manipulated by them. We need not look far to find alternative perspectives and points of view among the many media.

But the fact is that most people do not look, and the media has done a poor job in pointing out unfair and untrue attacks. It simply takes too long for the fast paced story that audiences seek out. It does not sell. So what we saw in 2004 was simply reporting what either side was saying with little critical analysis. In that contest the side that throws the most mud wins. By 2008 new websites do allow inquisitive voters to sort out truth from untruth (“politifact” and “factcheck”), but only a tiny percentage will do this. The fault lies with us. As long the mud moves us and as long as we consume simplistic stories that do not tax our minds too much, that is what we will get.

2. Valence Issues and Retrospective Voting

Valence issues are those issues that involve questions of who can do the better job. They usually involve goals on which everyone agrees like peace and prosperity. Most all people want low inflation, low unemployment, low interest rates, high productivity, increases in real wages, peace at home and abroad, and a government that can deal effectively and efficiently with whatever problems arise. The question is which candidates and which political party can best provide these desirable things. You can see why candidates are so interested in an image of strong leadership.

Generally speaking, voters tend to look at how things are currently going and blame or give credit to the party that holds the White House. When voters make this calculation and base their vote on which party is to blame or reward, it is called retrospective voting. Voters typically focus on whatever valence issues seem most important at the time of the election and then reward or punish depending on who is in power and how well or badly they have done.

Nevertheless, each party has a general image on being better or worse on different kinds of valence issues. Since World War II, most voters viewed the Republicans as being able to do a better job in defense and foreign affairs. They seemed to hold the office in times of peace. The Democrats were popularly viewed as doing a better job in running the economy. This image dates all the way back to the association of Republican Herbert Hoover with the Great Depression.

These popular images began to change in the 1970s. Nixon prolonged the Vietnam War, but then was given credit for finally bringing the troops home. Democrat Jimmy Carter occupied the White House during difficult economic times when inflation ran high and interest rates soared while the economy slumped. This unusual condition was called "stagflation." While his defenders blamed the Arab oil embargo, voters blamed him. Ronald Reagan ran his 1980 campaign challenging the incumbent on the valence issue of who best could handle the economy. He talked about a "misery index," which combined inflation and unemployment rates. He asked voters: "Are you better off today than you were four years ago?" Voters said "no" and then voted "no" on Carter's reelection.

Though voters doubted whether the Democrats could do a good job on the economy any more, the Republicans had yet to prove they could perform. That is why the Republicans suffered a harsh defeat in congressional elections in 1982, two years after Ronald Reagan was first elected. We were in the midst of a very severe recession. However, the economy rebounded and Reagan and the Republicans in Congress did much better in 1984. Voters began viewing Republicans as the party that could do the best job in handling the economy. The cliché that people "vote their pocketbooks" is usually not far from the mark.

In 1988 the economy was in pretty good shape after eight years of growth. The only real blip on the horizon was a huge and quickly mounting national debt. But that didn't bother too many people so long as they had jobs and money in their pocket. But the growing national debt did create some uncertainty. All other issues aside, the edge the Republicans had in their perceived ability to handle the economy and the close association that Bush had with the economic boom years of most of Reagan's presidency created the critical base for Bush's win. But it was a very narrow win in partisan terms. The Democrats picked up a few seats in both the House and the Senate.

In 1992 the Republicans lost their edge on the valence issue of the economy. This is the main reason why Bush had such a difficult reelection campaign. If he was to win, he had to turn voter's attention to other kinds of issues. That is hard to do when people are afraid for their economic futures. The Republicans did have a clear edge on the valence issue of keeping the nation strong militarily. But without the Soviet empire to frighten Americans, this was no longer a great issue. Moreover, except in times of great crisis, foreign policy had always played second fiddle to economics. Polls in the summer of 1996 indicated that Clinton and the Democrats were still seen as better being able to handle the economy, by about a 10 point gap. Having a relatively healthy economy was a major plus for Clinton going into the fall campaign. That advantage held up through the November 1996 election, and was probably the single most decisive factor in the contest, despite the Dole campaign's vigorous efforts to make questions about Clinton's character the major issue in the campaign.

In the 2000 campaign, valence issues cut both ways in the minds of voters. The economy helped the Democrats a little, but not as much as 1996 because there were indications that it was beginning to slow down. On the other hand, who could do a better job in preserving and protecting the nation's morals really helped the Republicans, thanks in large part to Clinton's incredible poor personal judgment and behavior. In effect, a character issue about the lame-duck president became a valence issue that hurt that president's party.

In 2004 the election, like almost all elections in which an incumbent is running, was also about valence issues and retrospective voting. How good a job did the incumbent do with the economy, health care, taxes, the budget, foreign policy, protecting the nation from terrorism? On many of these questions Americans had grave doubts about the incumbent. Bush and his team knew that. So they had to follow the same strategy that incumbent Jimmy Carter tried in 1980, to make the voters afraid of the challenger and use character issues to trump valence issues. Carter failed at that, but in 2004 it appeared that the Bush team is did a good job in making the voters feel that Kerry was an unacceptable alternative. They successfully transformed the election from valence issues to character issues with Bush as the strong character and Kerry as the weak character. This worked because they were also able to convince voters that the major issue in the election was security from terrorism and that the war in Iraq was part of the war on terrorism. Kerry was never able to reassure voters about his character and get voters back to Bush's performance in handling the Iraq War separate from terrorism or the economy, which was beginning to weaken.

And 2006? Valence issues of the handling of the war and the economy clearly helped Democrats take narrow control of both the House and Senate as voters retrospectively punished the Republicans. As for 2008, the housing slump, financial market crisis and record deficits and continued quagmire in Iraq doomed the Republicans. Failure to move the economy out of the Great Recession as fast as citizens wanted was certain to hurt Democrats in 2010.

3. Ideological

As you know, ideology is a very complicated concept. Most citizens don't know much about politics and certainly don't know much about ideology. A vast number of voting studies have attempted to use the concept of ideology to explain voters' behavior‑‑and most have not done very well. Citizen ignorance is a major confounding problem. About a third of all citizens simply don't use the terms and refuse to identify themselves. Almost another third use definitions that have little to do with the philosophical issue positions that political observers associate with the terms.

When asked about their ideological positions, more than half of all citizens identify themselves as either moderate or slightly liberal or conservative. Moderates are the largest group (see table below). Therefore, you should not be surprised that most politicians want to be seen as moderates, and go to great lengths to promote that image. One notable historical exception to this rule was the 1964 Republican Presidential candidate Barry Goldwater, who proudly stated in many speeches that "extremism in defense of liberty is no vice." He didn't even come close to being President Goldwater. Attacking the other side as "extreme _____" (you fill in the blank) is standard fare for speech writers.

The Ideological Identification of Voters*

Ext Lib Lib Sl Lib Mod Sl Con Con Ext Con don’t know

1996 2% 8% 11% 24% 16% 16% 3% 23%

2002 2% 14% 11% 27% 15% 25% 5% 1%

2006 2% 9% 8% 25% 12% 17% 3% 24%

* National Election Study exit poll, sampling error, +/- 2%.

Remember, these are voters, not just citizens, only half of whom vote.

Despite the fact that most voters don't understand ideology, we can use it as an organizing concept to compare the tendencies of the two major political parties on a variety of issues. In general, we can safely generalize that both parties tend to be moderate in their positions because that is where most voters are and that is how they manage to remain dominant over third parties. You can see their moderation better by comparing their positions to those of parties in European parliamentary democracies where three or more significant parties compete with each other.

However, despite this similarity, Democrats are usually somewhat to the left or liberal end on many key issues and Republicans are somewhat to the right or the conservative end. George Wallace was wrong when he said "there's not a dime's worth of difference" between them. Significant differences do exist, differences that reflect the moderate differences that exist among most Americans. Let us briefly examine six issue subject areas that can be seen in ideological terms and note the partisan trends in each.

a. Economic issues

In general, Democrats tend to worry first about unemployment while Republicans tend to worry more about inflation. This in part reflects the interests of groups that tend to support the two parties. Blue-collar workers tend to have less savings and more debts. So inflation may even help them so long as wages keep up. Blue-collar workers are usually the first to be laid off when there is an economic downturn. So they tend to favor policies that stimulate the economy even at the risk of running inflation.

Republicans, on the other hand, are more willing to slow down the economy to keep inflation under control even if that means laying off a lot of workers (who tend to be Democrats). Please remember I am merely talking about tendencies. These are not hard and fast rules. There are exceptions. If the economic slow down becomes too bad or too long, as it did prior to the 1992 election, Republicans do worry. Low inflation and low interest rates don't help bankers when people don't have money to put in banks and don't want to borrow money either. Moreover, the slump that extended into the 1992 election was different from previous economic recessions. Many white-collar workers lost their jobs as companies across the nation downsized their professional staffs as well as their production force. This created fear among many upper middle class voters who had long supported the Republican Party.

Therefore, in 1992 both parties were talking about what government could do to stimulate the economy. The normal conservative Republican position of minimal government involvement was mixed with talk about new job training programs. Of course, the Democrats wanted to stimulate more by spending more, though Clinton preferred to use the term "invest" rather than spend. How each party would fund these proposals reflected the economic interests of their core supporters. Clinton and the Democrats proposed to raise taxes on those who had over $100,000 annual incomes‑‑hurting mainly Republicans. Bush and the Republicans promised to stimulate by reducing capital gains tax rates‑‑helping mainly Republicans. We can see these tendencies reflected in the voting choices of various family income groups in the 1992 presidential election.

Family Income and 1992 Voting Choice

<$15,000 $15-30 $30-50 $50-75 >$75,000

Clinton 59% 45% 41% 40% 36%

Bush 23% 35% 38% 42% 48%

In 1996 the two parties divided along similar lines. Dole proposed massive tax cuts that would have been especially beneficial to the wealthy and Clinton argued against these cuts on several grounds, including preserving social programs like Social Security and Medicare. Given that the economy was in strong shape, the need for economic stimulation was much less convincing to voters. Clinton did a little better among all groups.

Family Income and 1996 Voting Choice (NES Exit Poll)

< $22,000 $22-50,000 >$50,000

Clinton 69% 54% 44%

Dole 24% 40% 49%

Economic issues, such as who benefited from the tax cuts George W. Bush was able to get passed, did play a major role in the 2004 election, but these issues were secondary to security issues. Nevertheless, the electorate again divided along income lines.

b. Regulation

The appropriate role of government in the economic marketplace is fairly closely related to economic differences between the two parties. Back in the 1930s, we could say that the Republicans wanted minimal government involvement. Government was to act only as a neutral umpire, making sure contracts were enforced and perhaps preventing restrictive monopolies. Government was to do little else. However, today the difference is much more a matter of degree. Both parties support a great deal of regulation of everything from food packaging and the effectiveness and safety of drugs to the safety of airplanes. What we are left with is that the Democrats want more environmental regulation and more work place and product safety regulation while the Republicans also want some of this but not quite so much. Indeed, both parties recognize that the public has grown accustomed to expecting the government to perform this kind of role.

c. Social Welfare Issues

The Democrats are distinctly different than Republicans on the appropriate role of government in helping people who have problems. This difference is also related to and consistent with the regulation question. Democrats are willing to spend more on public health care, school nutrition programs, food stamps, job training programs, education and student loans, and so on. Republicans, realizing how popular these programs are, for the most part do not want to kill them entirely, but merely want to spend less, feeling that more of this should be the responsibility of the individual.

Once again, these differences reflect the interests of those who tend to support the two parties. Democrats, thinking about their demographic characteristics, are more likely to need such programs. Republicans need them less. A central piece of Clinton’s budget and tax cut negotiations with the Republican Congress in the summer of 1997 was restoring some of the cuts in social programs that Clinton had allowed in passing welfare reform just before the 1996 election. In debating what to do about the budget “surplus” in 1999, Clinton wanted to spend more money on education and bringing down the national debt while the Republicans in Congress wanted tax cuts. (Please note that I put the term surplus in quotation marks because many observers argued that there was no real surplus—but that is another story that we do not have time to go into now. Moreover, the surplus is now clearly gone, no matter how we measure it.)

d. Social/Moral Issues

Social and moral issues create a different kind of dimension. It is also relatively new. On it the parties do something of a flip‑flop on the appropriate regulatory role of government. I want to greatly qualify my generalizations here, because supporters of each party have many internal conflicts within their membership on these issues.

In this area, which is sometimes not separated from social/moral issues, the parties take their more familiar and usual stands. Democrats are more in favor of government action to protect civil rights and to ensure equality while Republicans want to minimize government action.

Again, we can see this as more of a matter of self-interest of the coalitions that make up the parties than any real strong sense of ideological purity. Minorities make up a much larger portion of the Democrats, so the party favors programs that would help minorities. Consequently, Democrats tend to favor affirmative action programs, legislation to protect the rights of gays, and programs that increase the rights of women in such areas as employment and sexual harassment. Republicans want to minimize government help of minorities, from whom they get little political support, unless they see some political benefit, of course.

After the 1990 census Republicans around the nation formed coalitions of convenience with Black Caucuses in state legislatures to create majority black representational districts. This helped African-Americans win office, but it also helped Republicans in surrounding districts that lost African-American voters. The courts began disallowing this kind of racial gerrymandering in 1993. But they have allowed political gerrymandering. Today we are in the process of redrawing these districts. Whichever party controls state legislatures has the power to redraw district lines. Republicans controlled more legislatures after the 2000 census, and they took full advantage. If the Democrats keep and expand the gains they were making after 2006, they could have a major impact on drawing district lines after the 2010 census.

f. Foreign Affairs

This is a complex area. The history of the last 20 years has largely reversed the traditional party differences. If you go back to WWII, the Republicans were that party that wanted to steer clear of foreign involvement. They could be described as isolationist. The Democrats tended to be more supportive of intervention abroad. Remember who led the nation into WWI and WWII. Both Woodrow Wilson and Franklin Roosevelt were Democrats. However, you would be wrong to think of these as "Democratic wars," as people sometimes think. Once declared, large majorities in both parties supported the efforts. The distinction was that the Republicans were more reluctant to get involved in the first place.

Vietnam was different. In John Kennedy's inaugural address in 1960, the new president asked us to fight any fight any place in the world to promote democracy. Later he told us that Vietnam was the place where we had to fight next. However, what started as a Democratic war ended as a Republican war led by Republican Richard Nixon. Vietnam was much more divisive than any other war in the twentieth century. Vietnam divided the nation along many lines, including party lines after 1968. In that year Democratic candidate Hubert H. Humphrey broke with sitting Democratic President Johnson over the war. This was reinforced in 1972 when Democratic candidate George McGovern ran on the promise to pull out of Vietnam.

Partisans on both sides have tended to remain where the Vietnam War ended. Until recently, Republicans tended to worry more about the Soviet threat while Democrats, who were certainly concerned, were relatively less so‑‑and were more fearful of getting needlessly involved in wars in the name of fighting Soviet expansionism. Again, we are talking about a matter of degree. As Reagan showed, even conservative Republicans were capable of negotiating arms reduction treaties and would like to reduce tensions. In fact, one could argue that their more aggressive public posture made it easier for Republicans to get these treaties through congress and accepted by the public. But even Republicans were hesitant to get involved in long term military struggles like the Bosnian/Serbian/Croatian struggle in the former Yugoslavia where quick and clear victories seemed unlikely.

For a brief period after the Soviet threat was gone, both parties wanted to cut military expenditures. It was just that the Republicans wanted to cut the military budget more slowly than the Democrats.

September 11, 2001 changed that dramatically. There was a new enemy to combat, international terrorism. Both parties agreed to support new expenditures on initial parts of the war. Two massive and long term high cost invasions and nation-building efforts were led by a president who had campaigned against sending American troops in nation-building activities. Deep divisions have been developing over cost and the wisdom of going it alone, as President Bush chose to do.

Thinking about all three general types of issues we have covered here (character, valence, and ideological), which is the most important in determining voters' behavior as they retrospectively pass judgment? Research suggests that the relative importance depends to a great extent on the political environment of a particular election. Following the tragic fiasco of Watergate, "character issues" were clearly the most important in voters' minds. Jimmy Carter's pollsters knew that electing a President who "would not lie" to the American people was an effective campaign theme that year. It would be in tune with voter concerns. The strong personal morality of Carter was a major factor in his election. However effective or ineffective a president he was, Carter has shown since leaving office that a strong sense of personal morality was no mere campaign facade.

Following economic failures and seeming disarray in Carter's White House, people began to focus more on valence issues in 1980. Carter and the Democrats weren't doing a good job as seen through voters' pocket books. Voters gave him an early retirement. They also retired a significant number of Democratic congresspersons. The embassy hostage affair in Iran and the failed rescue mission didn't help Carter's leadership image very much either.

In 1992 a combination of issues again determined the outcome. Valence issues clearly favored the Democrats. Voters no longer saw the Republicans as the party best able to lead the nation to economic prosperity. Foreign affairs, where George Bush and the Republicans had the clear advantage, were no longer important. Ironically, Bush's success hurt him. His emphasis on foreign affairs was perceived as one more sign that he did not care enough about the economic problems that troubled many Americans. On ideological issues, Clinton had the benefit of being seen as even more moderate than Bush. Bush had to court the social conservatives in the Republican camp to even build a base from which to begin his campaign. The entire Republican effort turned on creating enough doubts in voters' minds about Democrat Clinton's character. Bush could only win if he were seen as the safe, even if dismal, alternative to a clearly untrustworthy "slick Willie" Clinton. The off‑and‑on‑again Perot candidacy complicated things for both parties. At first Perot helped Clinton by his attacks on Bush's economic policies. The more Perot talked about the deficit, the more people began to question Bush as President. But many of Perot's supporters did not trust Clinton personally. When Perot dropped out, the Clinton campaign feared that many would go back to Bush as the lesser of two evils. When Perot came back into the race, Perot's own character began to come into doubt. Despite this, Perot finally won nearly one in five votes. Exit polls show that Perot took about the same support from both Bush and Clinton. So Clinton probably would have won even if Perot had not reentered the race. On the basis of valence issues and the popular feeling that it was a time for change, the election should not have even been close. The Bush campaign (and Clinton's weakness) deserves some credit for making the election at least somewhat competitive.

1996 turned more simply on valence issues. A strong economy gave the incumbent president a nearly insurmountable advantage and voters rewarded him. Hints of scandal in campaign finance and continued doubts about Clinton’s character kept the election from being a complete rout. Ironically, the economy also helped the majority of Republicans who were incumbents in Congress. It was clearly an incumbent's election that favored the status quo.

As noted above, different valence issues both helped and hurt both sides in the extremely close 2000 election. Character issues ultimately helped Republicans get enough popular votes, even thought they got half a million less than the Democrats, to eek out a very controversial win in the electoral college.

What all this suggests is that valence usually have the edge over character issues. However, character issues can be pushed to the forefront if valence issues are indecisive or if voters have major character questions about a candidate. Ideological issues are only important if one of the candidates is clearly out of the moderate mainstream.

In 2004, the issues of security and the economy were at the forefront. The outcome was determined by which of these two short term issues voters felt to be more important. The economy was a short term issue that favored Kerry and security favored Bush. Bush portrayed himself as a “wartime” president who was strong in doing whatever was necessary to fight terrorism. That strategy worked. Exit polls showed that a majority of voters felt that the war in Iraq was part of the war on terrorism, and these voters strongly favored Bush. Polls in 2006 showed that citizens no longer felt that way, so the strategy worked for Bush in 2004 did not work for his party in 2006. The problem for Republican presidential candidate McCain 2008 was that a “stick it out and win” position in Iraq was no longer popular. A vast majority of Americans felt that the war has been badly handled and was not going very well at all. McCain claimed with some justification that the “surge” had worked to reduce causalities. But all that was overshadowed by the valence issue of the economy with what is now being called the Great Recession of 2008. How well Obama and the Democratic majorities in both houses of Congress handled the Great Recession would be the key issue in 2010 and most probably the key in deciding whether or not to give Obama a second term in 2012.

III. Hypotheses

For the sake of keeping this exercise simple and so that we can manipulate the data by hand and see what is going on, we shall test some simple hypotheses using only a few interviews or cases, those of your fellow classmates. We shall focus on seven variables: your party identification, the identifications of your parents, political ideology, a valence issue about the economy, a character issue concerning trust, and voting choice. From the existing theory you have just read we can generate several testable relationships among these variables.

Do you remember what a hypothesis is from the previous module? If you don't, it was defined as a specific statement about a relationship that logically flows from theory that can be tested by empirical means.

Let's list the variables, giving them symbols and then see what hypotheses we can generate from what you already have studied. If past theory is correct, it should have held up for most of you in the 2004 election.

Variables: v1 ‑‑ your party identification (pid)

v2 ‑‑ your mother's party identification (mpid)

v3 ‑‑ your father's party identification (fpid)

v4 ‑‑ your ideology (ideo)

v5 ‑‑ change in family economic situation (econ)

v6 ‑‑ which of the candidates is more trustworthy (char)

v7 ‑‑ which candidate you support in 2008 (vote)

Now let's try a few hypotheses. We'll do a couple together as a group.

You know from existing theory that party id is supposed to be the best single predictor of how one votes. From this alone, we can state a hypothesis we can test. We'll call it H1.

H1: An individual’s party identification usually matches their voting choice.

In arrow diagram form, this appears as follows:

pid (v1) ‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑> vote (v7)

Let's try one more simple one. We know that party id is supposed to be the best single predictor. We know that ideology is also a predictor, but we know that it is usually not as good for a variety of reasons. (Do you remember what they are?) We also know that the Democratic candidates and their supporters are generally more liberal while the Republican candidates and their supporters are generally more conservative. So if all this is correct, then we would expect the following hypothesis to be true.

H2: Those who identify as conservatives were more likely to vote for conservative candidates like McCain while those who are more liberal were more likely to vote for relatively more liberal candidates like Obama.

Again, in arrow diagram form, H2 appears as follows:

ideo (v4) ‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑> vote (v7)

Remember that these hypotheses should be guided by theory. At a minimum the researcher should have some good logical reasoning as to why a relationship should hold true. We may want to retest old relationships at new points in time. In any case, the point is that the researcher has a good reason to examine a relationship based on theory and what she observes in the world around her. She does not spend valuable time and money to look at something merely because she thinks it might be interesting or because she might uncover something new.

Let's make a couple of observations about these two hypotheses that serve as a review of earlier material. First, what roles do the variables play? In H1 the independent variable is party id and the dependent variable is vote. In H2 the independent variable is ideology and the dependent variable is vote. Do you remember that?

Second, what kind of relationships are we talking about? Are they causal, sufficient, necessary, or contributing conditions? Both hypotheses describe weak causal relationships in the sense that both of the independent variables do help determine the value of the dependent variable. But knowing party id is certainly not sufficient to completely determine voting choice. Nor is ideology. We also cannot reasonably argue that one must necessarily be a Democrat to support Obama, or be a conservative to support McCain. Thus, we would be wise to conclude that we are describing conditions that contribute to the likelihood of making a voting choice. We are talking about "contributing conditions."

IV. Operationalization

If we want to test these hypotheses and measure the other variables we have listed in order to test other hypotheses, we need to transform each variable into a concrete measure. Because we are interested in opinions and individual actions, we need measures that can be applied to individual people. This suggests that our best measures are questions that can be asked to people about their attitudes and voting actions. Therefore, for each of our variables, we must develop a questionnaire item that can be administered in a public opinion interview. This very important step of going from the concept of a variable to the exact measuring instrument that will be used is the process of operationalization.

Many people fail in their research at this step. They often create measures that are either "unreliable" or "invalid." Do you remember what these terms mean? Again, these are terms that you saw in the previous module. "Unreliable" means that the measure is so inexact that it does not give us consistent results. Unreliable survey questions typically are vague and subject to multiple interpretations. When the respondent answers them, they may give you different answers depending on exactly how they interpreted the question. An invalid question may give you consistent results, but it measures the wrong thing. For example, if you measured party id by asking people how they voted in 1996, you would get consistent results, but the results would misclassify a large number of people as Republicans. A number of Democrats voted for Dole because they disliked Clinton on personal morality issues, not because they were Republicans.

If you remember, one of the best way to avoid problems of reliability and validity is to use measures that have stood the test of time. We are lucky in that all of the variables we are using have been operationalized many times by other researchers. So we can use questions that are now considered to be fairly standard. The operationalized variables or questions will be as follows.

1. Do you generally consider yourself a Democrat, a Republican, or an independent? If independent, do you lean toward the Democrats or the Republicans? [CHECK THE MOST APPROPRIATE BLANK.]

__ 1) Dem __ 2) ind __ 3) Rep

2. Which of the following best describes your mother's party identification?

__ 1) Dem __ 2) ind __ 3) Rep

3. Which of the following best describes your father's party identification?

__ 1) Dem __ 2) ind __ 3) Rep

4. Which of the following best describes your political philosophy?

__ 1) liberal __ 2) middle of the road __ 3) conservative

5. How did your family’s economic situation change between 2004 and 2008?

__ 1) better __ 2) no change or don't know __ 3) worse

6. Thinking about character, in which of the two major party candidates did you have the more trust?

__ 1) Obama __ 2) neither/no diff/don't know __ 3) McCain

7. Who did you support in the 2008 election? (Note: the way this question is usually asked before an election is "If the election were held today, for whom would you vote?") If you don’t know,” leave this blank, but try and choose from these two if at all possible. With so few people in our “sample,” any “missing data” will really have an impact on the numbers in our tables.

__ 1) Obama __ 2) McCain

Before we go on, let's review a bit more. Do you remember what "levels of measurement" are from the last module? What levels of measurement are each of these operationalized variables?

If you remember, there are three levels: nominal, ordinal, and interval. We can eliminate the most exact level, the internal level, very quickly for all of the variables. We aren't making exact measurements of amount. None of the questions involve units of measurement.

In question 7, we seem to be just measuring categories. No particular order matters. We are not measuring more or less of anything. These are all nominal level measures.

However, in the first three questions, we are measuring strength at least partially: Democrats and Republicans are more partisan than "independent," which means having no partisan feelings in either direction. It would seem to belong in the middle. Therefore, these questions are ordinal level measures, even though they are fairly weak. We could make them stronger by measuring whether the Democratic and Republican identifiers are weak or moderate or strong identifiers.

The question measuring change in family economic condition also is ordinal in that it has a natural order. "No change" rather clearly belongs in the middle, though we can arbitrarily put either worse or better at the low end of this simple scale. A similar argument can be made on question 6, the character question.

What about the ideology question, number 4? Again, like political party, order counts. We could again add adjectives like weak or moderately or strong to produce a more precise measurement. But given the small sample we will be using, we need to keep things simple. There is no point is having a lot of extra categories when only one or two people may fit them.

V. Data Gathering ‑‑ Survey Research Methods

For the sake of simplicity, we will pretend that the students in this class are a good sample. It is not really a good sample. Using classmates creates two major problems. They are not representative of any other group and they are too few to make a prediction with any statistical certainty. So we'll just have to pretend.

About the only thing we can say is that the class can be used to test hypotheses that in theory should apply to all groups of citizens. So if most people get their party id's from their parents, then that should also be true of college students in your class. If that turns out not to be true, we still can't say for sure that the theory is wrong. With a small group, we could get the wrong results as a matter of bad luck rather easily. We'll soon see.

To give you an analogy, you can get four heads in a row when flipping a coin rather easily. If you do, that does not prove the coin is "unfair." (The chances of four heads are .0625.) However, if you get 10 or 15 in a row, you'd be wise to take a good look at the weighting of that coin. If almost nobody in the class has the party id of their parents, we might want to take a closer look to see if we can figure what is getting in the way. Negative results in even a small unrepresentative sample can suggest further research using larger samples. That's why social scientists often do things called case studies or in‑depth small group studies. These studies allow scientists to look closely enough to see if the results suggest relationships that can be tested later on larger groups.

If we were actually doing a serious survey, our sample would have to be much larger and we would have to utilize scientific selection procedures. The discussion that follows should give you some insight in what would actually be involved in a survey that is designed to represent national opinion.

Designing, performing, and interpreting the results of a survey or opinion poll is a most complex and difficult task. You could spend a lifetime just studying the different ways to draw samples. We teach whole courses on survey research and really don't get much beyond scratching the surface. However, despite this, I think I can give you some valuable and useful information in a brief discussion. You don't really need to know all the things you would have to know to perform a real poll, because most of you will never have to do this. If you do, you'll need more training. However, you will be consuming polling information for the rest of your life. Almost any newspaper contains some polling results that someone thinks are significant. So what I want to give you is enough to be able to evaluate the significance and limitations of polls that you see. In other words, I want to make you a little more informed consumer of polls.

Evaluating polls involves knowing the right questions to ask. What questions? You should ask about the possible sources of error that could result from the way the sample was drawn, from the size of the sample, from refusals, or from bias in the interviewing techniques or the question wording. Let us briefly talk about each of these things.

A. Sampling

A census is when you ask everybody in a given population about something in which you are interested. We don't do this very often, because it is very expensive, very time consuming, and very hard to do, especially if the population is very large. The U.S. Constitution requires that the government do a national census every 10 years. We are still counting the data from the 2000 census. The next one will be in 2010.

In a survey, we don't look at the whole population. Instead, we look at a sample of the population and use what we find out about them to generalize to the entire population. By population, we simple mean whatever group it is we are interested in. It could be all Aiken City voters, Aiken County voters, property owners, S.C. citizens over the age of 18, all cities of a certain size, all nations, all states, and so on. Each member of the population, like each state if the population is all states, is called the unit of analysis or the case.

Sampling simply refers to the way in which you choose this smaller group of cases from the larger population. Ideally we want to choose the sample so that the sample is representative of the larger population. The problem is exactly how to do this. Here is where the laws of probability come into play. If we choose units from the population so that every possible unit in the population has an equal chance of being chosen, we are likely to get a sample that looks pretty much like the population in virtually any characteristic or opinion you can think of. You have no guarantee. If you flip a fair coin, you are likely to get about as many heads as tails. This kind of sample‑‑where every case has an equal chance of being chosen‑‑is called a simple random sample.

If you think about the simple random sample, you can intuitively see how it works. Suppose you have a big shopping bag of red and green jelly beans. You want to know what the % of each color is without handling them all. What you might do is mix them up well (so that each has an equal chance of being chosen) and pick out a hundred without looking. If you did a good job at this and if you are not unlucky ("fortuna" plays a role here as well), you will have about the same proportion of red and green ones in your sample of 100 as are in the whole bag. To be more precise, the laws of probability tell us that you will be within 10% of the true percentages on one side or the other nineteen times out of every twenty. And it does not matter much whether there were 10,000 or 10,000,000 jelly beans in the bag. Your results are just as good regardless!

This 10% error in either direction is called sampling error. It is calculated from the laws of probability and can be controlled by choosing a larger or smaller sample.

Simple random samples of different sizes yield different expected sampling errors. You might find the table below useful. But remember, a one in twenty chance exists that the error could be worse, and the technique only works to the extent that the sample fits the essential characteristic of a simple random sample (that every unit in the population has an equal chance of being chosen).

Sample Size Expected Sampling Error

100 + or ‑ 10%

150 + or ‑ 8%

400 + or ‑ 5%

1,100 + or ‑ 3%

2,500 + or ‑ 2%

10,000 + or ‑ 1%

Of course, people are not jelly beans, and people are not in bags or always on lists from which you can randomly choose. That's why we usually can't use simple random sampling. It isn't practical.

People are more complicated (kind of like multi‑colored jelly beans whose colors are always changing). They are much harder measure. But nevertheless, all of the complex sampling techniques that pollsters use are pretty much aiming at getting as close as we can to this idea of a simple random sample‑‑making it so that each person in the population of interest is equally likely to be chosen. If each person is equally likely to be chosen, then the likelihood of choosing each opinion should be the same as the proportion of the people who hold that opinion.

Let me describe a couple commonly used techniques so that you can see how sampling works. Many telephone surveys use a sampling technique called random digit dialing. The idea here is that you dial different sets of digits drawn at random so that all telephone numbers are equally likely to be dialed. You also try to do most of your calling at times when most people are likely to be at home. You may make a few calls in the morning or afternoons, but not many. However, you call each number you originally selected several times at different times of the day to give those who are not home an equal chance of actually being interviewed.

Cluster sampling is a commonly used technique in planning door-to-door surveys that cover populations over a wide geographic area. Suppose we needed to do a survey of all people in the state of South Carolina. It would be too costly to use census maps and randomly choose a single person in a single house scattered all over the state. You might have to travel 50 or 60 miles between each interview! So instead, you proceed in several stages. For a state level survey, you might choose about 10 small geographical areas of the state at random. That's the first groups of clusters. Next you choose about 3 or 4 neighborhoods within each area at random. Then you perform 4 interviews in each of those neighborhoods in houses chosen at random with individuals chosen at random within each house. This four-stage process (areas/neighborhoods/houses/respondents) would give you about 160 interviews. The exact numbers at each stage depends on the total number of units you want in your sample and how many different kinds of areas you need to cover (e.g. urban, suburban, and rural).

In mail surveys where you have lists of people and addresses from which to choose, systematic sampling is an excellent technique to employ. Here you simply go through the list and choose every nth person so that each person again has an equal chance of being chosen. For example, suppose you have a list of 1,000 people and want a sample of 200. You would go through the list and choose every 5th person starting with one of the first 5 in some random way.

B. Bias

Bias is the kind of error we don't like because we never can be sure how big it is and that makes bias harder to control. All we can do here is be as careful as possible and keep bias to a minimum‑‑unless we want to bias the poll to create propaganda. Candidates sometimes want to so this to stimulate campaign contributions from groups that are looking to influence the likely winner.

Several possible sources of bias exist in a poll. Let's look at each one and discuss what we can do about it. As a consumer of polls, you should be able to spot examples of these biases.

1. Sample Bias

The further you get from the ideal simple random sample, the more likely that certain kinds of people and opinions are likely to be chosen out of their true proportions. In other words, sample bias is when certain kinds of people are more likely to be selected than others and when that choice has little to do with their actual frequency in the population.

A common example of this is the "straw polls" that tv networks or newspapers run. In effect, by volunteering to call in or send in a "ballot," people are asked to choose themselves as part of the so‑called "sample." Here you are likely to get too many intense and extreme opinions. In addition, the results are highly subject to artificial manipulation. "Man on the street" interviews where reporters go out and talk to those whom they can conveniently find are called "accidental" or "convenience samples." They are likely to be badly biased in some way. These things should not even be called polls. But unfortunately, they are and they tend to give more scientific polls a bad name.

2. Response Rates

Even if you choose your sample well, people may not always cooperate with you. They may hang up, not send the questionnaire back in the mail, or even slam the door in your face. This is not only psychologically discomforting (few of us like rejection), it creates unknown bias in the sample you get. Therefore, pollsters are very concerned with getting as high a response rate as possible (or to put it the other way, a low "non‑response rate").

How good is good? Most surveys shoot for 70% or more by using call-backs for those who don't answer the phone the first time, multiple mailings, and trying to give those who are contacted as much incentive as possible to participate. As a rule of thumb, bias becomes a pretty large dark cloud over the results anytime the response rate begins to get as low as 50%.

3. Question Bias

Question bias is more controllable. Even though we can't make people respond, we can write the questions in any way we want. As we noted above in laying out the questions we will use in our little pretend poll, this is an area in which people often make mistakes, either unknowingly or purposefully. Let me offer some rules of thumb to consider in either writing questions or in evaluating questions used by others.

a. Beware of value laden terms and titles in a question. Answers may have more to so with people's reaction to the term than to the content of the question itself. For example, anytime a President's name is in a question, you pick up feeling about the President as much as about the content of the actual question. For example, pollsters found in 1994 that President Clinton's unpopularity hurt any health care proposal that had his name attached to it. Take his name off and more people would approve of the proposal.

b. Beware of long and complicated questions. They are probably unreliable in that people are likely to interpret them in many different ways‑‑thereby yielding inconsistent results. This is the complaint that students often have in answering long test questions--if the question is unclear, then how can it fairly measure knowledge?

c. Beware of questions that ask for subjective self‑evaluations. People want to be seen in the best light possible. Objective questions are far better. "Did you go to church last week" is better than "are you generally a faithful member of a church."

d. Beware of how the order of questions can affect answers. A good example here occurred in the 1984 presidential election in which Reagan did relatively better than Mondale in polls in which the "if the election were held today who would you choose" question was at the beginning of the poll rather than at the end. The explanation was that people were less likely to pick Reagan after they had been asked about a number of issues questions that reminded them of the problems of his presidency. (Herbert Asher, Polling and the Public, Washington, D.C.: Congressional Quarterly Press, 1988, pp. 47‑8.)

4. Interviewer Bias

The more the person who is doing the interview interacts with the person being interviewed, the more likely that this interaction will affect the answers that are given. When you hire people to do the interviews for you, as most organizations do, you can never be sure what they do‑‑unless you go with them. However, on telephone interviews, you can use special phone banks that allow you to monitor at random the phone calls to make sure interviewers are following proper procedures. In mail surveys, of course, this is not a problem. Probably the best advice is to make sure you have spent adequate time training interviewers and testing their skills. You also need to pay them enough to get quality people doing the work.

As a consumer of polls, about the best way to judge is by the reputation of the polling organization. You can generally trust Gallup, Harris, and the polling done by major newspapers, tv networks, wire services, and universities (like USC Aiken!).

C. "Door‑step" Opinions‑‑The Intensity and Non‑opinion Problem

Suppose a pollster comes to your front door and asks you about health care reform, which is certainly an important topic. But suppose you have not given the subject much thought, so you give an answer that might well change if you did get some information and did give it some thought. Your opinion gets recorded with other opinions that may be just as lightly held and a report gets published concluding that Americans feel by a 2 to 1 margin that we should scrap our entire health care system.

The question that must be considered is whether these are real opinions or whether they were artificially created by the mere asking of the question. If you and a lot of other people who were asked gave thoughtless answers, what we have is really the measurement of non‑opinions, or what are called door‑step opinions. At best, the opinions you get when you ask people about things that they haven't thought about very much are very low in intensity. These opinions are subject to dramatic changes if the question is asked slightly differently or asked in a different context. To put it another way, these kind of questions are unreliable. They yield inconsistent responses. The answers you get are close to meaningless.

So what can you do? The usual practice is to ask a filter question that attempts to eliminate all those who don't have opinions before you ask the actual question. You might say something like: "Many people worry about our present health care system. Have you had a chance to think about it enough to form an opinion on the subject?" Then if they say "yes," you go on and ask their opinion. You may also wish to give them a choice of answers that allows them to give the intensity of their opinion, ranging from strongly something on one side to strongly something else on the opposite side.

Here is a good rule of thumb. Anytime a poll shows that EVERYONE asked had an opinion on almost anything, you can bet that the results are polluted with thoughtless "door‑step" opinions. Somewhere between 10 and 30 or 40% should have "no opinions." The more complex the subject area, the higher the percentage of "no opinions" should exist. If not, take the results with a big grain of salt.

D. Costs, Compromises, and Tradeoffs

Even though polls are cheaper than doing a census, polls are expensive. Costs range from a low end of $7 to $10 per interview for a short mail survey, $10‑$20 each for telephone interviews, up to $50 or more an interview for a door-to-door survey. Some kinds of surveys can also be done more quickly. Completing a mail survey in under 2 to 3 months is virtually impossible. A telephone survey can be done in 24 hours if you have the necessary equipment and staff. You also can't ask about everything you would like to. You can only hold people's attention for so long. The more questions you ask, the more likely people are to get tired and give you thoughtless responses or terminate the interview.

What all this means is that the perfect survey does not exist. Every decision involves compromises and tradeoffs depending on the time and money you have, the nature of the population you are trying to survey, and the information you are trying to gather. In short, here too Machiavelli's advice is worth remembering: do what is necessary employing all the skill and experience you can muster along with getting good help where you need it.

VI. Analysis

A. Coding

Coding is the process of transforming the answers respondents give on the questionnaires (raw data) to codes (either numerical or alphabetic or both) on special sheets called "coding sheets" so that they can be more easily loaded into the computer for manipulation and summarization (statistical analysis). That's a mouthful. Read it again. In this process the researcher must make sure that all answers are consistently and accurately coded. Sometimes that's called "verification." The coding process also condenses the data so that it can be more quickly loaded into the computer.

We are going to code the interviews from the class for a slightly different purpose. We want to make it easier for us to manipulate the data by hand. On a long interview, each interview usually has its own coding sheet. However, because ours is so short, we will place all the data on a single sheet. After you do this task by hand, you will appreciate how much time and trouble a computer can be in helping the researcher statistically analyze long questionnaires (that is, with many variables, at least 20‑25) that were administered to large samples (typically around 1,100 in national surveys).

B. Hypothesis Testing

Anytime you are testing for a relationship between two variables that each have only a few possible values, you should set up something called a "crosstabulation" or a "two way frequency distribution." This is what we shall do together for each of our three simple hypotheses. Later I'll be asking you to do this as a written exercise for some hypothesis of your own with some real data from a national survey. Here are the steps you should follow.

1. Place the codes and associated labels for the independent variable on top of the table (so that each possible code forms a column, for example, 1. Democrat 2. independent and so on).

2. Place the codes and associated labels for the dependent variable down the left hand side of the table (so that each possible code forms a row). In Hypothesis 1, the dependent variable is voting choice. The first row would be "1. Obama" and the second row would be labeled "2. McCain."

Putting steps 1 and 2 together for Hypothesis 1, we would get a table that looks like follows, BEFORE we put any data into it.

Party Identification

‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑‑-----‑‑-----------------‑‑‑------------------

1. Dem 2. ind 3. Rep

Vote |‑‑‑‑----‑-----‑‑|‑‑‑‑‑‑‑-----‑‑‑‑‑‑|‑‑‑-----------‑-‑‑‑|

1. Obama

|‑‑‑‑----‑-----‑‑|‑‑‑‑‑‑‑-----‑‑‑‑‑‑|‑‑‑-----------‑-‑‑‑|

2. McCain

|‑‑‑‑----‑-----‑‑|‑‑‑‑‑‑‑-----‑‑‑‑‑‑|‑‑‑-----------‑-‑‑‑|

3. Use the coding sheet to count the frequencies that go in each box or "cell" (the proper term) that is formed by the codes of the rows and columns. For example, count the number of Democrats who say they would vote for Obama. Put that number in parentheses in the cell in the first row and the first column.

4. Add the frequencies for each cell down each column to get the "column frequencies." Place this number in parentheses at the bottom of each column.

5. Compute the "column percentages" for each cell by dividing the cell frequency by the column frequency for that column and record these percentages in the appropriate cells. For example, if 8 Democrats voted for Obama out of a total of 10 people in that column‑‑the Democratic column‑‑then the column percentage for the cell of Democrats who voted for Obama would be 8 divided by 10 or 80%. See how simple this is!

6. Examine these computed column percentages across each row and look for trends. Ask yourself the following question as you move across each row. As we shift from Democrats through independents to Republicans, what happens to the %'s? Are any trends (shifts in the %'s) obvious that are consistent with those that are predicted by the hypotheses. If not, how is the pattern different. Look at each row.

VII. THEORY REFORMULATION

Theory reformulation is where you take what you have found and compare it with what we know from existing theory. Do your results confirm existing knowledge? If not, how do your results change or modify it? What implications do your findings have for political behavior? What further research does your research suggest should be done?

VIII. Data Manipulation and Hypothesis Testing

This is a two-part exercise. First, we will test the hypotheses we laid out earlier in this module. We'll do this as a class. Then I will ask you to think of your own hypotheses using some of the other questions (or variables) in the questionnaire. We may do this in class or for homework.

Before we can do anything, we need data. Fill out the questionnaire at the end of this module. If you do this before class, it will save time in class. PLEASE do it NOW. You will need to print this up and bring it to class.

We will pretend that the class comprises a reasonable sample of students at the school. As you should know, it does not. It is a "convenience" sample. That means we can't use laws of probability to draw statistical inferences about anything else. We don't know how badly it is biased. However, if we can test relationships that are supposed to generally hold for all citizens, they should be true of this group as well. If not, we might wish to ask what about this group makes the group different than other more representative samples of citizens. With that huge caveat in mind, let us proceed.

After we have filled out all the questionnaires (the data gathering stage), we need to code the data. What we will do is read out our answers around the room and each of us will enter the codes for the answers on the coding sheet that is also at the end of this module.

Next, we must set up the crosstabulation tables to test each hypothesis. The first one, H1, is already shown to you back under the "Analysis" section of this module. All you have to do is to follow the six steps in that section. We'll go through them together for the first table.

After we have done that, I want you to try setting up the appropriate tables for H2 and compute the appropriate numbers, and interpret what the table tells you.

In class I will show you how to use MicroCase, a powerful but easy to use statistical package. We will use some national survey data as well as perhaps some survey data from Aiken County and go through all the steps of the scientific research process. We will do this in class as a group in class rather than as a graded paper. If it were a paper you had to do, here would be the steps you would follow in writing the paper – on outline, if you will, for the report you would have written.

First, you would have to think of a significant problem that you could explore. Write a paragraph explaining why this is significant problem. You may have to work a little backwards here, first finding variables that might explain why people vote the way they do or even why they vote at all. Label this paragraph “Problem Statement.”

Second, the “Theory” paragraph, you would briefly discuss why you think what the two variables might be related. You could use any of the information you have learned in previous modules or in this module for each step in the research process. Alternatively, you could come up with a common sense theory, so long as it is based on some reasonable sounding argument. You could look back at section II where we discussed what we know about voting behavior.

Third, under “Hypotheses,” you would need to state the hypothesis, just as we did with H2 and H2. Also you would include an arrow diagram showing the independent variable pointing to the dependent variable.

The fourth section should be labeled “Operationalization.” It should show exactly how the variables were measured. In this case, all you would need to do is give the question wording that was used in the two questions in the NES survey.

The fifth section is data gathering. Because this would be what is called secondary analysis, meaning that we did not gather the data ourselves, all you would have to say is that the data came from the National Election Survey and was made available to us through MicroCase. See the statement in the sample paper below.

Sixth, the “Analysis” section would have a copy of the crosstabulation that you used Microcase to produce. You probably would just copy it by hand and then type it up so that it is completely labeled and laid out like we did on the board in class. Under the table, you would need to show how to read it. After doing this, you would conclude whether or not the table supports the original hypothesis.

Finally, in the “Theory Reformulation” section, you would discuss what this finding means in terms of your original theory (section two above). Does it support the theory? If so, is it strong or weak support? What does this suggest about more research that might be done? If it does not support the original theory, what do you think may be wrong? How might your theory be altered? And of course, this means that someone else can then do some more research on this new idea. That is how knowledge is built in science. The process is ongoing.

Even though you are not doing this paper, these are the steps you would follow if you were and in fact they are the steps that all scientific research generally follows.

APLS 110 Student Questionnaire ID # ___ ___ (assigned in class)

1. Do you generally consider yourself a Democrat, a Republican, or an independent? If independent, do you lean toward the Democrats or the Republicans? [CHECK THE MOST APPROPRIATE BLANK.]

__ 1) Dem __ 2) ind __ 3) Rep

2. Which of the following best describes your mother's party identification?

__ 1) Dem __ 2) ind __ 3) Rep

3. Which of the following best describes your father's party identification?

__ 1) Dem __ 2) ind __ 3) Rep

4. Which of the following best describes your political philosophy?

__ 1) liberal __ 2) middle of the road __ 3) conservative

5. How did your family’s economic situation change between 2004 and 2008?

__ 1) better __ 2) no change or don't know __ 3) worse

6. Thinking about character, in which of the two major party candidates did you have the more trust?

__ 1) Obama __ 2) neither/no diff/don't know __ 3) McCain

7. Whom did you support in the 2008 election? __ 1) Obama __ 2) McCain

CODING SHEET: APLS 110 Student Survey

pid mid fid ideo econ char vote

ID # V1 V2 V3 V4 V5 V6 V7

01 ______ ______ ______ ______ ______ ______ ______

02 ______ ______ ______ ______ ______ ______ ______

03 ______ ______ ______ ______ ______ ______ ______

04 ______ ______ ______ ______ ______ ______ ______

05 ______ ______ ______ ______ ______ ______ ______

06 ______ ______ ______ ______ ______ ______ ______

07 ______ ______ ______ ______ ______ ______ ______

08 ______ ______ ______ ______ ______ ______ ______

09 ______ ______ ______ ______ ______ ______ ______

10 ______ ______ ______ ______ ______ ______ ______

11 ______ ______ ______ ______ ______ ______ ______

12 ______ ______ ______ ______ ______ ______ ______

13 ______ ______ ______ ______ ______ ______ ______

14 ______ ______ ______ ______ ______ ______ ______

15 ______ ______ ______ ______ ______ ______ ______

16 ______ ______ ______ ______ ______ ______ ______

17 ______ ______ ______ ______ ______ ______ ______

18 ______ ______ ______ ______ ______ ______ ______

19 ______ ______ ______ ______ ______ ______ ______

20 ______ ______ ______ ______ ______ ______ ______

21 ______ ______ ______ ______ ______ ______ ______

22 ______ ______ ______ ______ ______ ______ ______

23 ______ ______ ______ ______ ______ ______ ______

24 ______ ______ ______ ______ ______ ______ ______

25 ______ ______ ______ ______ ______ ______ ______

26 ______ ______ ______ ______ ______ ______ ______

27 ______ ______ ______ ______ ______ ______ ______

28 ______ ______ ______ ______ ______ ______ ______

29 ______ ______ ______ ______ ______ ______ ______

30 ______ ______ ______ ______ ______ ______ ______

31 ______ ______ ______ ______ ______ ______ ______

32 ______ ______ ______ ______ ______ ______ ______

33 ______ ______ ______ ______ ______ ______ ______

34 ______ ______ ______ ______ ______ ______ ______

35 ______ ______ ______ ______ ______ ______ ______

36 ______ ______ ______ ______ ______ ______ ______

37 ______ ______ ______ ______ ______ ______ ______

38 ______ ______ ______ ______ ______ ______ ______

39 ______ ______ ______ ______ ______ ______ ______

40 ______ ______ ______ ______ ______ ______ ______

41 ______ ______ ______ ______ ______ ______ ______

42 ______ ______ ______ ______ ______ ______ ______

43 ______ ______ ______ ______ ______ ______ ______

44 ______ ______ ______ ______ ______ ______ ______

45 ______ ______ ______ ______ ______ ______ ______

46 ______ ______ ______ ______ ______ ______ ______

47 ______ ______ ______ ______ ______ ______ ______

48 ______ ______ ______ ______ ______ ______ ______

49 ______ ______ ______ ______ ______ ______ ______

50 ______ ______ ______ ______ ______ ______ ______

KEY TERMS AND IDEAS

Why we study voting behavior

partisanship

independents

trends in partisanship

third parties

how party id affects voting

choice

political involvement

political efficacy

turnout

demographic factors

why people are more likely to

vote

character issues

valence issues

retrospective voting

citizen knowledge about ideology

ideology & economic issues

ideology & regulation

ideology & social welfare

ideology & social/moral issues

ideology & civil rights

ideology & foreign affairs

case or in-depth small group

studies

census

survey

population

unit of analysis or case

sampling

simple random sample

sampling error

random digit dialing

cluster sampling

systematic sampling

sample bias

straw polls

accidental or convenience

samples

response rate

call-backs

question bias

interviewer bias

door-step opinions

filter questions

tradeoffs in types of surveys

coding

crosstabulations