Important class info follows:
January 14- We started talking about the research paper and the scientific behavioral era in political science. We talked about the party identification, issues, and candidates affecting the presidential vote model, and how it applied to explaining the outcome of the last two presidential elections. I urged students to think about their own desired model for their research paper, before the next class in a week. Review the sample student research paper I e-mailed you, and take a look at the variables you can choose from the Mississippi Poll.
January 21- We covered the 6 characteristics of the behavioral era, in weeks 1-2 of the class notes, listed on a link at the top of this webpage. This is important, and will be one question on the first test. We also discussed choosing your model for your research paper. Review the sample student papers on my class website at: https://sds17.pspa.msstate.edu/classes/analysis/analysis.html. Over the weekend, pick a subject that interests you, and choose a relevant variable which will be your Dependenbt Variable (at the extreme right of your model). Then, pick three variables that will be your Independent Variables, at the extreme left and in the middle of your model. On Monday during lab and class, we will disscuss your papers once again. Choose your four variables for your model from the webpage at: https://sds17.pspa.msstate.edu/poll/ContinuityGuide.html
January 26- Well, we are into remote operations today, so let's get started. First, I predicted that we would go into remote operations a couple of hours before MSU announced it. America is fortunate to be so wealthy that we can "take it easy", so to speak, rather than taking risks when facing possible crises. Our society has become more risk adverse. Ten years ago, for example, our high school cross country team would run 6 miles a day, 7 days a week, regardless of weather. So I have fond memories of running up and down hills in one hundred degree heat index weather. Then Covid came, and our high school track season was cancelled. Then, a couple of years ago, they would not even let us run in hot weather, so we would spend two or three weeks each summer running at 6 or 7 in the morning. More and more, we heard the phrase "out of an abundance of caution" in our national life. Indeed, Trump's plane had to turn back due to electrical problems, and they used that phrase. Another explanation for universities declaring remote operations (even if the bad weather does not greatly effect our area) is that many staff members and students drive from out of town to get to class or work, and we do not want them to be injured. Plus, we obviously wish to avoid lawsuits. So I predicted the remote operations ahead of time. Class is asked, "any down side to a concentration on caution?" Think about it. I also predicted that as someone gets older, they get more risk adverse. Why? They remember all of the bad experiences in their lives, they get more fearful and fragile. As such, I predicted that I would check myself into the Hilton Garden Inn for a week or two. My own experiences in Starkville include most of the city losing power for three days, and faculty moving their families into a Bowen Hall classroom; I had a fireplace back then, but my house was still so cold that winter that I got a bad cold. So when you have money, why not stay at a nice hotel with a restaurant and bar?
Today, we're going to cover the Theory Building section of Weeks 1-2 class notes. A theory seeks to explain and predict a political phenomenon, such as people's political attitudes or their vote decisions. It was discussed as the first characteristic of the Behavioral Era. A theory is so important that this subject will constitute the second question on our first essay exam. Examples of a theory will be each of your models for your research paper, as you are seeking to explain and predict a political attitude or phenomenon. There are 4 characteristics of a Good Theory. 1) It seeks to explain why something happens or why someone behaves as they do. In the satisfaction-dissatisfaction model of voting, we showed a pattern of public dissatisfaction often resulted in an incumbent candidate or party losing an election, while satisfaction often resulted in a candidate winning the election (like Reagan and Clinton's re-elections). Why? People have a general sense or how things are in the country, and they all know the name of one political figure- the President. They also know what political party the President is. Bad news for the party holding the presidency if things are believed to be bad, even if the President is not running, as Republicans found out in 2008 when Obama won after Bush's 2 terms. The University of Michigan model also provides a good explanation for presidential vote choices. There are big differences between Democrats and Republicans, and if a person is psychologically attached to a party, it makes sense for them to support candidates of that party. This party identification concept is called a long-term factor, as it exists over time. However, issues can also be important. An explanation for why Republican Goldwater lost in 1964 and why Democrat McGovern lost in 1972 is that they were too extreme, even for members of their party. And certainly too distant from the Independents' own views. The candidate factor can also explain election results. Carter in 1980 was seen as a failed leader because of all of the economic and foreign problems, so Reagan was able to win by asking, "Are you better off today than you were four years ago?" Heck, no, most answered. 2) Prediction is the second characteristic we look for in a good theory. Calling up the Mississippi Poll dataset from my I drive and Mississippi Poll folder, MississippiPollRecoded file. It shows you the many variables we have in past Mississippi Polls at the window top, and the 12,271 cases (people) at the left side. We run an Analyze, Descriptive, Frequencies for the variable Year, and it shows in each year how many people were polled from 1981 thru 2014. Now I do an Analyze, Descriptives, Crosstabs, Target List (dependent variable) is "elecpres", and column is Party. Cells is Column Percents. The predictive ability over this 30-year period shows that 89% of Strong Democrats voted for the Democratic presidential candidate, while 94% of Strong Republicans voted for the Republican candidate. Much better than a coin toss. Prediction was also pretty good for the weaker partisan categories. 71% of Independent Democrats voted for the Democrat, while 89% of Independent Republicans voted for the Republican. Again, much better than a 50-50 split. So the University of Michigan party identification model had great predictive value. What about the candidate or issues/performance part of the model? Now I'm going to do a Data, Select Cases, If condition is satisfied, if year = 2012. We then look at Mississippians rating of President Obama's job performance. Among the 65 Mississippians polled who rated Obama's job as Excellent, fully 100% voted for him. And among the 122 who rated his performance as Poor, a full 100% voted for Republican Romney. How's that for predictive ability! (90% of those rating him Good voted for him, and 66% of those rating his job as only Fair voted for his opponent.) Does that make sense, think about it? 3) Generalizability is the third characteristic of a Good Theory. When the University of Michigan scholars wrote the classic, The American Voter, it examined only the 1956 presidential election. But later scholars repeated their analyses for all other presidential elections to the present, and they also found the party-issues-candidate model upheld. Therefore, the theory was Generalizable, not confined to only one case (one office and year). I made the theory even more generalizable, as my PhD dissertation at Ohio State tested the model for U.S. House, Senate, and gubernatorial elections. Party id was a strong predictor of voting in these other offices. For the candidate factor, I included incumbency and name recognition, and both of those did indeed affect these non-presidential races. I also used ideological self-identification as the issues variable. Just before the turn of the century, I expanded the model to state legislatures and statewide non-gubernatorial offices, and showed how each southern state was becoming more Republican in party identification, thus eventually resulting in Republicans gaining control of nearly every southern state legislatures and state offices such as lieutenant governor, attorney general, auditor, etc. My Ohio State professor Brad Richardson applied the theory to other countries, and countries like France, Britain, Italy, Netherlands, and Japan were affected by this model. Therefore, this theory is Very Generalizable. 4) We look for Parsimony. If the first three characteristics of a Good Theory are met by two theories, we prefer the simpler theory. What is a simpler theory- the one with the fewest predictors (independent variables). The party-issues-candidate model is parsimonious, as it only has three predictors. The satisfaction-dissatisfaction model is parsimonious, as it only has 1 predictor (with 2 categories). A model with 10 demographic factors, 10 different specific issues, and 10 different candidate characteristics would have 30 predictors. It would be much less parsimonious.
Well, given how slow I speak, how I write on the board a lot, repeat myself, we have already run out of class time. We now move to the lab. The purpose of our lab today is for you to choose the four variables for your model. You will have three time periods- earliest, middle, and latest. The latest is your dependent variable. The earliest is your "outside" predictor, which comes first. The middle of your model is your "intervening" predictor, which goes between your outside and dependent variable. You can have one outside and two intervening predictors, or have two outside and one intervening predictors. The first thing you should do is choose a subject that interests you, which will be your dependent variable. One example is a student who may be interested in explaining why there is a gender gap in party identification. It is shown in Weeks 1-2 of the notes under the heading "YOUR RESEARCH PAPER." Therefore, Party Identification is the latest variable (at the right), and Sex is the earliest variable, at the extreme left. She hypothesized that women tended to be more Democratic than men because they tended to be more liberal in ideology. So Ideology was one intervening variable placed in the middle between sex and party id. She also hypothesized that women tend to have lower incomes than men, and since the New Deal the poor have tended to be more Democratic in party id than the rich. So Income was another intervening variable, placed in the middle between sex and party id. She ends up with 5 hypotheses with arrows drawn between pairs of variables: sex affects party, Ideology affects party, Income affects party; but also, sex affects ideology, and sex affects income. Also note how precisely she words her 5 hypotheses. Such as, women are more like than men to be liberal. More likely is the phrasing. Not all women are liberal, it's just that a higher percent of women are liberal compared to men who are liberal. When you link up pairs of variables in each of your hypotheses, you always phrase them that way- more likely. Liberals are more likely to be Democrats, compared to conservatives. Someone asks, can you word it with opposite categories. How about, men are more likely to be Republican, compared to women. That is correct also. But it does help to try to be consistent in the categories that you choose, as it is easier for a reader to follow them. In this example, she relies on the categories women-liberals-low income-Democratic. They all go together, so to speak. A reader can easily follow the logic. Now, you are ready to choose your four variables. Click on https://sds17.pspa.msstate.edu/poll/ContinuityGuide.html" on my website, or just type it into the web. It has all of the variables you can choose in the Mississippi Poll for your model. I gave you a shortened version of it as a handout the last class. Pick four relevant variables that are asked in the same years. Using 2010, 2012, and 2014 is easiest. But you can use other years, if a subject appears in only a few years. If you pick 4 variables that make sense for your model, and if all of them are asked in all three of these years (or even two years), then you are all set. I need to go around and approve each of your models before you leave the lab. Since we are not physically meeting, you can show me your model in class on Wednesday (or you can e-mail it to me before then). After your model is approved, you can start writing up the three sections of the first part of your paper. You begin with an Introduction, where you explain and defend why you chose your topic, and why it is important. For example, if your dependent variable is Desired State Spending on Health Care, you could cite statistics of Mississippi health woes and how much money is spent on Medicaid. You can already mention national studies on the problem or issue you are examining. Give some specifics, even early on. The next part of your paper is your Model and the 5 hypotheses. An example is in Weeks 1-2 of the notes, plus the sample student paper. The third part of the paper is the Methodology section (we skip the lit review for now). It has three paragraphs. The first one can be copied right from this section (YOUR RESEARCH PAPER), if you use the 2010-2012-2014 polls, since the information is the same. The second paragraph gives the exact question wording of your four variables. For that, you need to go to the Mississippi Poll Codebook, so click on https://sds17.pspa.msstate.edu/classes/methods/mpolcode.html. If you use any of the STATE SPENDING PRIORITIES, you should include the intro that are read to people: "Now I'm going to ask you about some issues facing state and local government in Mississippi. As you know, most of the money government spends comes from the taxes you and others pay. For each of the following, please tell me whether you think state and local government in Mississippi should be spending more, less, or about the same as now. How about ...?" For the ..., just insert the all caps title of the program that appears after the number. For example, if you're looking at spending preferences on poverty programs, that is number 39, and right after it in caps is "Programs for the Poor." You must have the exact question wording for your variables. The third paragraph is your best guess as to how the categories of each variable should be combined, so that you get a manageable number of categories. For age, you don't want 80 unique age categories. Maybe you'd combine them into 3 age groups. Same with income. Well, that is a lot of material that we've covered in the lab, so time to go.
January 28- we covered Ethics section of Weeks 1-2 notes. 4 components of Informed Consent is important to know, plus the difference between Anonymity and Confidentiality. We then reviewed from Monday the 4 components of a Good theory. We then had a short lab where we talked about the paper, and each student came up with their model having 4 variables, and the professor approved them. So, they can start on the first part of their paper.
February 2- We covered two important items that will be on the test: Units of Analysis, and Levels of Measurement. Review these notes, which are in Week 3 of the class notes. Most students turned in the first part of their papers, which include the Introduction, Model and Hypotheses, and Methods section. In Lab students worked on finishing these parts of their papers. Wednesday is the new deadline date for turning in your papers without grade penalties.
February 4- TBA.