GVPT 423
E. Uslaner Congressional Elections Spring 2002
x54151 Tydings 2126E TuTh 11-12:15
Undergraduate Assistant: Rebekah Mayer, rjmayer@wam.umd.edu
Course home page: http://www.bsos.umd.edu/gvpt/uslaner/gvpt423.html
In order to understand the literature on Congressional elections, it will be important to have some background in understanding statistics. I do not presume that you have any, so have appended to this syllabus a glossary of terms that you will come across in your readings.
Each student will also pick a Congressional campaign (either House or Senate) from the 2000 elections. Using the background gained in the course, students will write a paper of approximately 10 pages examining why your candidate won or lost. Your choice of a candidate must receive my approval--or your paper will not be accepted. You must select a race before Spring Break and hand in a short paper (one page will do) describing the race. You must hand in the paper by March 19. You may hand it in any time before that date. The paper itself is due on May 7. Each student will make a brief (no more than 10 minutes) presentation about his/her paper beginning in mid- April. Following each presentation, there will be about five minutes for other students to ask questions. The presentations are not meant to be summaries of final products. They are designed to give others an idea of the race you picked, the key issues, and to give you feedback about issues that others in the class believe to be interesting or important. Students who are working on the same race will make joint presentations.
Your paper should include the following:
(1) a brief description of the candidates; describe the incumbent's record and the challenger's qualifications. Is the challenger a "quality challenger"? Has the challenger run for the seat before? Did either candidate face a primary challenge? How might this affect the November race?
(2) a description of the district, including its demographics and electoral history. How long has the incumbent served? Has (s)he faced strong challengers in the past? Has the incumbent generally won by big or small margins? How did the Presidential candidates do in the district in 2000? Describe the two candidates' campaigns. Which candidate ran the stronger campaign? Did both candidates conduct negative campaigns--or did either?
(3) an analysis of campaign spending. How much did the incumbent and challenger spend? Was the challenger competitive in spending? Why or why not?
(4) an analysis of the issues in the race. What were the key issues? Did they benefit the incumbent or the challenger?
(5) an analysis of the incumbent's record. Was the incumbent a strong supporter or opponent of the President? On what committees did the incumbent sit? How might the incumbent's record affect the primary and general election voting?
You should use the descriptions of the campaigns in the Herrnson text as models for your paper. You can find information on the candidates and the campaigns from the following sources:
Politics in America 2002
Almanac of American Politics 2002
Lexis-Nexis Academic Universe, on the web through the library (using a campus connection) at http://www.lib.umd.edu/UMCP and go to data bases and then scroll down till you get to Academic Universe.
The Hotline: a great source for today’s politics at http://www.nationaljournal/policycentral
(through a university connection only)
Congressional Quarterly Weekly;
National Journal
and the many election web sites. The best are:
Web White and Blue also has links to almost every other election-oriented Web site you could imagine. And you can get complete data on campaign finance for any race at the Federal Election Commission’s web site, http://www.fec.gov.
You will also find that the national press often covers Congressional races. Check the New York Times and the Washington Post in our libraries. You might also want to check the local press in your district, although this will require a visit to the Library of Congress if Nexis doesn’t cover them or if the local paper doesn’t have an on-line source with a search engine. You should check to see if there is a web site for newspapers in the district. A convenient way to find this is to use Yahoo (http://www.yahoo.com) and then to go to “News and Media,” then click on “Newspapers” and then on “Browse by Regions.” Next go to “States” and choose your state and browse.
The course requirements include: (1) a mid-term examination, to be held in-class on March 12, covering material through Topic 6 (20 percent of your grade); (2) the ten page paper on a Congressional race (30 percent of your grade); (3) a take-home final examination of approximately 10 pages (40 percent of your grade); and class participation (10 percent of your grade). The one page paper required by March 19 will not receive a grade. It is required. If you do not hand in this one-page paper on March 19, I shall deduct a full grade from your final paper.
The term paper will be due IN CLASS (not later) on May 7, 1999. The final examination will be due either in my office (Tydings 2126E) or in the main GVPT office (Tydings 3140) by 10 a.m. on May 20. Papers that are late without a verified emergency will be downgraded one full grade for each day late starting at the time the papers are due (the end of class on May 7 for the term paper and 10 a.m. on May 20 for the exam). There are no exceptions to this policy.
ALSO I CANNOT ACCEPT PAPERS BY E-MAIL. IT SIMPLY WOULD TAKE ME TOO LONG TO DOWNLOAD EVERYONE’S PAPER. SO PAPERS MUST BE TURNED IN DURING CLASS IN HARD COPY.
We shall also have a class webboard maintained by Ms. Rebekah Mayer, the undergraduate TA for the class. The webboard will be a forum for discussion of issues related to Congressional elections. Participation in the webboard will count toward your overall participation grade. The listserv will also cover issues we discuss about Congressional elections but don’t have time to discuss in class. IT IS NOT A FORUM FOR YOUR GENERAL POLITICAL VIEWPOINTS, NOR IS IT A FORUM FOR COMMENTS (POSITIVE OR NEGATIVE) ABOUT THE COURSE, THE PROFESSOR, THE ASSISTANT, OR OTHER STUDENTS. THIS RESTRICTIONS ARE IMPORTANT TO MAINTAIN CIVILITY AND WILL BE STRICTLY ENFORCED. IF YOU HAVE QUESTIONS OR COMMENTS ABOUT THE COURSE, PLEASE DIRECT THEM TO PROFESSOR USLANER OR MS. MAYER. FINALLY, PLEASE LIMIT YOUR RESPONSES TO OTHERS. DON’T HOG THE LIST!
We shall also have at least two classes devoted to open discussions, linked to questions on the webboard and the video we shall see in class. The first will ask you whether you think incumbents have an unfair advantage and what might be done about it. The second will focus on the campaign video.
IF FOR ANY REASON YOU CANNOT TAKE AN EXAMINATION OR HAND IN AN ASSIGNMENT ON TIME, YOU NEED TO CONTACT ME BEFORE THE ASSIGNMENT IS TO BE HANDED IN. IF WE DON’T HEAR FROM YOU BEFORE THE DEADLINE, THERE WILL BE NO POSSIBILITY OF A MAKE-UP FOR EXAMS OR YOU WILL AUTOMATICALLY LOSE A FULL GRADE FOR EACH DAY LATE FOR PAPERS/TAKE-HOME EXAMS. WE WILL MAKE EXCEPTIONS FOR TRULY EXCEPTIONAL CIRCUMSTANCES, BUT YOU MUST BE ABLE TO DEMONSTRATE TO US THAT YOU WERE UNABLE TO CONTACT US. IF YOU DO MISS AN EXAMINATION, YOU NEED A STATEMENT FROM YOUR DOCTOR VERIFYING YOUR ILLNESS.
The mid-term examination will consist of essay questions drawn from the assigned reading (all of which will be available at Hornbake's Reserve Room) and the class lectures and discussions. All assigned reading is required and may be the subject of examination questions. The take-home final is attached to the end of the syllabus. It will not be due until the day of the regularly scheduled examination, but you may, of course, begin thinking about it at any time. All assignments are to be typed, double-spaced and with reasonable margins.
All written work must be your own. Copying the work of others, whether that of fellow students or anyone else, constitutes plagiarism. You need not copy a work in its entirety to plagiarize. Should you plagiarize, you will be reported to the appropriate authorities at the university and the case will be prosecuted. The penalties for plagiarism range from failure in the course to expulsion from the university. Since there has been a large number of plagiarisms, I must insist that everyone state the following on the term paper and the take-home final examination:
I have read the statement on plagiarism, understand it, and state on my honor that this paper is my own work.
Signed, (your name)
Copies of the university policy on plagiarism will be distributed in class. Should anyone have any questions, please feel free to consult me. If your paper or your take-home final examination does not contain the statement, you will automatically receive an F for the paper unless I determine that you have plagiarized. In that case, I shall refer the case to the College of Behavioral and Social Sciences for adjudication. Don’t ask for an exception: There won’t be any.
I expect you to take care with your writing. An excessive number of spelling and/or grammatical errors will lead to a reduction in your grade on both the simulation paper and essay questions on examinations. I also expect you to come to class--and to arrive on time. If you miss more than a few classes or come into class late, it will adversely affect your participation grade.
Incompletes will not be granted unless: (1) you die; (2) you have a baby; or (3) you can convince me that something terrible will happen to you if an incomplete is not granted. Failure to request an incomplete prior to April 16 will, except under the most unusual circumstances, eliminate the possibility of receiving an incomplete. Also, I try to be accessible to answer questions that you might have.
In the list of readings below, the following abbreviations will be used for journal citations:
APSR American Political Science Review
AJPS American Journal of Political Science
LSQ Legislative Studies Quarterly
JOP Journal of Politics
The topics below are listed roughly by week. The following books should be purchased at any of the local bookstores:
Gary C. Jacobson, The Politics of Congressional Elections, fifth edition
Richard F. Fenno, Jr., Home Style
Richard F. Fenno, Jr., Congress at the Grassroots
Sunil Ahuja and Robert Dewhirst, eds., The Roads to Congress 2000
And there is a packet at Bel-Jean’s available from the University Book Center. Articles marked with an asterisk (*) below are in this packet.
Topic/Date
1 (1/29) Introduction
Introduction to course; read Glossary of statistical terms in this syllabus.
2 (1/31-2/5) House Members in Their Constituencies
Fenno, Home Style, entire
3 (2/7-2/9) Why Incumbents Run So Well in House Elections
Jacobson, chs. 2, 3, 5.
Ahuja/Dewhirst, chs. 1, 2, 4, 5, 6
*George Serra and David Moon, “Casework, Issue Positions, and Voting in Congressional Elections,” JOP, v. 56 (February 1994), pp. 200-213.
*Diana Evans, “Johnson vs. Koskoff: The 1998 Campaign for the Connecticut 6th District”
4 (2/14-2/19) The Role of Money in Congressional Elections
Jacobson, ch. 4.
Ahuja/Dewhirst, chs. 3, 7, 8, 9.
*James Campbell et al., "Television Markets and Congressional Elections," LSQ, v. 9 (November 1984), pp. 665-678.
5 (2/21-2/26) Incumbent Vulnerability in Senate Elections
Ahuja/Dewhirst, chs. 10, 11, 12, 13
*Alan I. Abramowitz, "Explaining Senate Election Outcomes," APSR, v. 82 (June 1988), pp. 385-404.
*Gerald Wright and Michael Berkman, "Candidates and Policy in U.S. Senate Elections," APSR, v. 80 (June 1986), pp. 567-587.
*Gary C. Jacobson, "Strategic Politicians and the Dynamics of U.S. House Elections, 1946-86," APSR, v. 83 (September 1989), pp. 773-794.
*L. Sandy Maisel, Kara E. Falkenstein, and Alexander M. Quigley, “Senate Retirements and Progressive Ambition among House Members in 1996,” Congress and the Presidency, v. 24 (Autumn, 1997), pp. 131-148.
*Liza Mundy, “The Politician,” Washington Post Magazine (November 4, 2001)
2/28 Open discussion on incumbent advantages: Good or bad? What, if anything, should we do about it?
6 (3/5-3/7) Comparing Senate and House Elections
*Fenno, The United States Senate (in Bel-Jean packet and on reserve), entire
*Alan I. Abramowitz, “The End of the Democratic Era? 1994 and the Future of Congressional Elections Research,” Political Research Quarterly, v. 48 (December, 1995), 891-918.
*Earl Black, “The Newest Southern Politics,” JOP, v. 60 (August 1998), 591-612.
March 12 Mid-Term Examination
7 (3/14-3/19) Campaign Video and Discussion
*Matt Bai, “Running from Office: Why Max Kennedy’s Congressional Run Never Took Off”
*Alexander Bolton, “Endangered Rep. Kennedy Gets $90 million in pork,” The Hill (December 12, 2001)
NO CLASS MARCH 21
8 (4/9) The Nationalization of Congressional Elections: Retrospective Voting and Issues in Congressional Elections
Jacobson, chs. 6, 8.
*Uslaner and M. Margaret Conway, "The Responsible Congressional Electorate," APSR, v. 79 (September 1985), pp. 788-803.
*Donald Kinder and D.R. Kiewiet, "Economic Discontent and Political Behavior in the 1980 and 1982 Congressional Elections," AJPS, v. 23 (August 1979), pp. 495-527.
*John C. Green et al., “Faith and Election: The Christian Right in Congressional Campaigns, 1978-1988,” JOP, v. 55 (February 1993), pp. 80-91.
NO CLASS ON APRIL 4 OR APRIL 13 OR APRIL 18
9 (4/11-4/16) Changes in Congressional Representation
Fenno, Congress at the Grassroots, entire
NO CLASS APRIL 25
4/18, 4/23, Student Reports
5/2-5/14
GLOSSARY OF STATISTICAL TERMS
This brief glossary is arranged alphabetically! Within each entry, terms that are referenced elsewhere are underlined.
Before the glossary begins, here is an extraordinarily brief review of some issues in the analysis of electoral data.
Suppose you want to determine what leads some people to vote Democratic and other people to vote Republican. This is what you want to explain. It is called the dependent variable. The factors that explain the dependent variable are called independent variables. A likely factor in leading people to vote the way they do is ideology. People who are liberal are more likely to vote Democratic and people who are conservative are more likely to vote Republican. You then offer the following hypothesis:
The more liberal someone is, the more likely he/she is to vote Democratic.
For a sample survey of voters, you have measurements on how people voted (Democratic or Republican, excluding those who did not vote at all) and also how liberal your respondents are. So you want to know whether your hypothesis is supported by your data. To do this, you conduct a regression analysis. Here you predict the level of voting Democratic from the measure of liberalism. From your hypothesis, you expect that the two will be positively correlated. Higher levels of liberalism will lead to a greater probability of voting Democratic. On the other hand, liberalism will be negatively correlated with voting Republican. In a regression analysis, the regression coefficient or b tells you how much great the impact of voting Democratic each increase in liberalism is (see the discussion of b below). In other words, how big a push does liberalism give vote choice?
Even if one has a big impact on the other, this does not mean that your results are reliable. You need to consider two other measures to determine that. The first is the correlation, or r. The correlation tells you how well your prediction from the regression analysis fits the actual data. A correlation of either +1.00 or -1.00 is a perfect fit (see correlation or r below). The sign of the correlation depends upon whether the relationship is expected to be positive (liberalism and a Democratic vote) or negative (liberalism and a Republican vote). A zero correlation indicates that you have no predictive power at all. The second measure is the t-test. This test is based upon the assumption that your independent variable really has no impact on the dependent variable--that the regression coefficient you obtain really should be zero if you had data on all voters, not just your sample. You use the t-test to see if this assumption is realistic or if your results really are strong enough so that they cannot be attributed to "chance." In other words, if you conducted many, many surveys, you would still get strong results.
Finally, you surely realize that there is more to voting decisions than whether voters are liberal or conservative. If we include other possible factors that might influence vote choice (e.g., party identification, the state of the economy, etc.), then we have many predictors and a situation of multiple regression. The measure of correlation now becomes not r, but R (see below).
Why do we use regression analysis and how do we interpret it? Let’s look at an example. What determines campaign spending by challengers? I estimated a simple model using data from Senators serving in 1977-78 who ran for reelection in either 1978, 1980, or 1982. There is no particular reason for choosing these Senators (or even the Senate rather than the House). I happened to have the data on hand for my own research. Three factors that could affect how much money challengers spend are: (1) how much incumbents spend, (2) whether the challenger has previously held prominent elective office (either a statewide office or as a member of the House of Representatives), and (3) whether the incumbent faced a divisive primary. We would hypothesize that:
(1) Challengers will spend more money when they face incumbents who also spend a lot of money. The more money incumbents raise, the greater the pressure on challengers to raise money. So high incumbent expenditures lead to high challenger spending. We thus expect a positive (+) relationship and a positive coefficient for the regression analysis.
(2) Challengers who have more experience--and more name recognition--will find it easier to raise (and thus spend) more money. So a higher quality challenger will lead to greater challenger expenditures. We thus expect a positive (+) relationship and a positive coefficient for the regression analysis.
(3) Challengers will raise (and spend) more money when the incumbents they face have had difficult primary races. A close primary will send a message to campaign contributors that the incumbent is vulnerable. So contributors might act strategically and be more willing to give money to the challenger. Here’s the logic of what we expect: The lower the incumbent’s primary vote percentage, the more money challengers should raise (and spend). The higher the incumbent’s primary vote percentage, the less money a challenger should raise (and spend). We thus expect a negative (-) relationship between incumbent primary vote share and challenger spending and a negative coefficient for the regression analysis.
Note: We measure both challenger and incumbent expenditures in thousands of dollars. We measure challenger quality by what we call a “dummy” variable. It’s a “dummy” because it can take only two values: 1 for quality challengers and 0 for non-quality challengers. And we measure incumbent primary vote share by the percentage of the vote the sitting Senator received in the last primary. Also, every regression analysis includes a “constant” term. You don’t need to worry about it, but everyone reports them. (If you have a good memory from your high school algebra--you did take it, didn’t you--you might recall that the formula for the equation for a straight line is Y = a + bX, where b is the “slope” and a is the “intercept,” the point where the line crosses the Y axis. Regression analysis is exactly the same--for just one independent variable, you get a straight line.)
So what do we have? Let’s see.
Dependent Variable: Challenger Expenditures
Independent Estimated Standard t-
Variable Coefficient Error Statistic
Constant .023 .040 .567
Incumbent
Expenditures .295 .060 4.934
Challenger
Quality 620.800 .015 4.046
Incumbent
Primary Votes -.268 4.501 -.059
Number of Observations 74
R-squared .396
What does this mean? The interpretation of a regression coefficient is the change in the value of the dependent variable (challenger expenditures) for every unit change in each independent variable. OK, so what does this mean?
Let’s start one variable at a time. For every $1 an incumbent spends, the challenger spends .295 (the regression coefficient for incumbent expenditures). If the incumbent spent $1000, the challenger would spend $295. If the incumbent spent $1 million, the challenger would spend $295,000. If the incumbent spent nothing, the challenger would spend nothing. Note that the regression coefficient, as expected, is positive. What else can we say?
The second column gives us the “standard error,” which in essence tells us how reliable our measurement is. Suppose we had data on elections for other Senators at other (maybe more recent) time periods. Would we get the same results or is this sample a bit odd in some way? A high standard error tells us that our results are suspect--and may not be statistically significant. We take our regression coefficient, divide it by the standard error, and get the t-statistic (or t-ratio) in the third column? So what do we do with a t ratio? We look up a table for t to get the probability that our regression coefficient really is zero, not the .295 that we estimated. We find that the probability is less than .0001 (p < .0001), so we conclude that the effect of incumbent spending is real and did not occur by chance.
When we go to our next independent variable, challenger quality, we find that a quality challenger will spend $620,800 more than a candidate who has not held elective office before. The coefficient of .062 tells you that the challenger will spend $621 for every $1000 the incumbent spends. If an incumbent spends $1 million, a quality challenger will spend $620,800. But this only holds for quality challengers. The “unit change” in the independent variable means going from zero (not quality) to one (quality). Poor quality challengers spend nothing additional, so good quality challengers get quite a financial boost. The relationship is real. The t ratio tells us that the probability that real coefficient is zero is also less than .0001. And, as expected, the relationship is positive (better challengers spend more money, not less).
Finally, what about the effect of a divisive primary for the incumbent? As expected, the relationship is negative: The greater the percentage an incumbent receives in the primary, the less money a challenger spends. But the impact is very small: Each percentage point the incumbent receives in the primary costs a challenger just $268.00. Even if an incumbent gets 100 percent of the primary votes, say by running unopposed, the challenger can spend just $26,800 more than if the incumbent got no votes at all (and, of course, was defeated). So even though the coefficient is negative, that doesn’t mean that the challenger gets much of an advantage from a tight incumbent primary. In fact, the very small t ratio (-.059, negative because the regression coefficient is negative), suggests that the regression coefficient isn’t at all different from zero (no effect at all). The coefficient is not significantly different from zero. How can we tell? Social scientists traditionally use a “rule of thumb” that the significance level has to be less than .05. This means that we need (in most cases) a t ratio that is either greater than 1.645 or less than -1.645 (if we expect a negative coefficient). Clearly, .059 is much less than 1.645. So we can say that incumbent’s primary vote shares have no real effect on challenger spending.
Overall, how much does our regression tell us. For this we use the R2 statistic. R2 varies between zero and one. A value of zero means that we could do as well by guessing that all challengers spend $830,500, the mean amount that challengers raised in these three years. A value of one means that we can predict challenger expenditures perfectly from our set of independent variables. (We rarely get a value of zero and never get perfect predictions). The R2 we get is .396, which means that we do reasonably well, accounting for about 40% of the spread between the low figure of zero and the high figure of $4,113,000. We still have 60% of the spread that we haven’t accounted for, but we are doing far better than chance.
One additional note: Sometimes you will see references to probit or logit analyses. Researchers use probit or logit when the dependent variable is a dummy variable, such as voting Democratic or Republican or voting for the incumbent or the challenger. You almost certainly don’t care about why you can’t (or shouldn’t) use regression analysis in this case, but note two things. Probit or logit (which are almost identical to each other) are like regession. But (second) you can’t interpret their coefficients in the same way. In fact, you can’t interpret them at all. Sometimes researchers will give you something useful such as the “effect,” which is the difference in the probability (say, of voting for the Democrat or the Republican) based upon the values of each independent variable. So, if the Democrat spends $1 million rather than nothing, does this increase--and by how much--the probability that a vote will cast a Democratic ballot.
The glossary contains terms relating to regression and correlation, but also some other terms as well (mean, median, standard deviation, variance, probit) that you will confront in the readings.
a The intercept in a regression equation, Y = a + bX+ e. The intercept tells you the mean of the dependent vari-able when all of the other variables equal zero.
b The regression coefficient in a regression equation, Y = a + bX+ e. b tells you the impact of each independent variable upon the dependent variable. For example, suppose the dependent variable is vote choice (0 = Republican, 1 = Democrat) and the independent variables are: (1) whether you have met the Democratic candidate (0 = no, 1 = yes); (2) whether you have met the Republican candidate (0 = no, 1 = yes); and (3) whether the Democratic candidate is the incumbent (0 = no, 1 = yes). If the regression equation is:
Y = .45 + .10b1 -.15b2 + .32b3 + e,
where b1 = meet Democratic candidate
b2 = meet Republican candidate
b3 = party of incumbent,
then: .45 (the intercept, or a) is the probability that a voter will cast a Democratic ballot if he/she has met neither the Democratic nor the Republican candidate and the Democratic candidate is not the incumbent. If the voter has met the Democratic candidate, he/she is--other things being equal--10 percent more likely to vote Democratic. If the voter has met the Republican candidate, he/she is 15 percent less likely to vote Democratic (15 percent more likely to vote Republican), other things being equal. And if the Democratic candidate is the incumbent, the voter is 32 percent more likely to vote Democratic, other things being equal.
beta What is called the "standardized" regression coefficient. In the example of b above, we compared the interpretations of these coefficients for three independent variables. All three variables had the same possible values (0 or 1). But suppose we added campaign expenditures for the Democratic candidate to the equation. This variable will range from some small positive value (generally no less than a few thousand dollars) to some very large value (in the millions of dollars). The resulting b4 coefficient cannot be directly compared to the other coefficients because the variables are measured on different scales. In particular, the coefficient for expenditures cannot be interpreted as a probability (since the independent variable does not range from zero to one). In order to compare coefficients, we divide each coefficient by the standard deviation for the respective independent variable. Thus, we divide the coefficient for meeting the Democratic candidate by the standard deviation of meeting the Democratic candidate, etc. This allows us to compare the magnitude of each coefficient, so that we can state that the impact of one variable is greater than that of another.
correlation see r
dependent
variable What we seek to explain or predict. For example, we want to predict whether someone will vote Democratic or Republican for Congress; this variable depends upon the values of other variables--campaign spending, who is the incumbent, etc., the independent variables. We denote the dependent variable by Y.
e The error term in a regression equation. What is left to predict in the dependent variable that the independent variables don't account for. Also called residuals.
independent
variable The variables we use to explain or predict the dependent variable. We denote them by X's. They do not variables depend upon other variables.
mean The average value, as, for example, in a ballplayer's batting average.
median A measure that complements the mean. half the cases in a sample lie above the median, half below. For example, for the numbers [0, 1, 2, 88, 99], the median is 2, while the mean is 38. The mean, the median, and the mode are identical if the data correspond to a normal distribution.
mode A measure that complements the mean. for a sample of data cases, the mode is the category with the largest number of entries. For example, for the numbers [0, 0, 0, 1, 1, 1, 1] for seven voters (where 0 indicates voting for the Republican and 1 indicates voting for the Democrats), the mode is 1. For data that are normally distributed, the mean, median, and mode are identical.
normal
distribution
A concept that relates to the shape of the data. A normal distribution is the familiar bell-shaped curve, bution with the most extreme values having the fewest cases. Some instructors (not this one) use this distribution to determine course grades, with a lot of C's, fewer B's and D's (in the same amount), and even fewer A's and F's (also in the same amount). For a normal distribution, the mean, median, and mode are identical.
probability The likelihood that something occurs. Must be between zero (no chance at all that something happens) and one (something occurs with certainty). The probability that you obtain a head when you flip a fair coin is .5.
probit A variation on regression analysis. You don't need to know more about this for this course. The only thing that is important is that probit coefficients have no ready interpretation. That is why we generally derive estimates of the probability a particular result occurs.
r, R, R² r is the correlation coefficient. It ranges between +1 and -1. A positive value of r indicates that as the values of the dependent variable increases, so do those for the independent variables. A negative value indicates that as the dependent variable increases, the independent variable's values decrease. For example, if the dependent variable is voting turnout (0 = abstain, 1 = vote) and the independent variable is interest in politics (0 = not interested, higher values show increasing levels of interest), we would expect people with greater interest would vote more often. Thus we expect a positive value of r. On the other hand, with the same dependent variable (turnout), our independent variable were now a measure of distrust in government (0 = feel government can be trusted, 1 = government can not be trusted), we would expect higher turnout among citizens who feel that government can be trusted. Thus, we would expect a negative correlation between turnout and distrust. If a correlation equals +1, this means that we can predict the dependent variable perfectly from the independent variable. If a correlation equals -1, this means that we can also predict the dependent variable perfectly from the independent variable, but that the relationship is negative. A correlation of zero indicates that there is no relationship between the two variables, as we might expect, for example between shoe size and voting turnout. The correlation coefficient indicates whether there is a strong relationship between the dependent and independent variables in a regression analysis. The higher the value of r (either positive or negative), the better the regression analysis does in predicting Y.
R is the multiple correlation coefficient used in
regression analysis when there is more than one independent variable. R only varies between zero and one; it cannot take on negative values.
R² is the usual measure for determining how well a regression analysis performs. It is the square of R. Any dependent variable, Y, can be expressed in terms of the range of values it can take on (turnout, for example, is zero for abstain and 1 for vote in our example). We can "normalize" the total range--or variance--of the dependent variable to 1. Then, the value of R square is the proportion of that range that the independent variables predict. This measure ranges from zero (we have failed to predict anything) to one (perfect prediction).
regression See b above. What regression does is to explain or predict a dependent variable, Y, from independent variables, or X's through an equation such as:
Y = a + bX + e,
where a is the intercept and the b's are regression coefficients and e is the error terms or residuals. We judge how well the equation predicts Y by the correlation or R.
X Used to signify the independent variable(s) in a regression analysis.
Y Used to signify the dependent variable in a regression analysis.
TAKE-HOME FINAL EXAMINATION
The consulting firm of Victory, Inc. has learned that you have taken a course in Congressional Elections at the University of Maryland and has engaged you to write a report in which you recommend a strategy for the five-term Representative Harvey Cliffhanger, a Democrat from the Fourth Congressional District of the state of Transylvania. Cliffhanger has to make a decision whether to seek an seventh term in the House of Representatives or to run for the Senate against the three-term Republican incumbent Frank N. Stein. You are to prepare this memorandum, of approximately 10 pages (typewritten and double-spaced), based upon the knowledge you have gained in this course. Victory, Inc. does not want a "game plan" gas to how to run a campaign. As a six-term veteran, Cliffhanger has plenty of knowledge on that score. Rather, the firm wants you to relate what broader knowledge you have gained in this course, drawing on the lectures and in particular on the reading. Indeed, your compensation for this assignment (reflected in your grade for the final examination) will bear a direct relationship to the way in which you integrate the course readings. There is no such thing as a satisfactory answer that does not dwell heavily on course readings. An essay that does not have substantial references to the course reading will receive a grade no higher than a C.
The report will be delivered either to the Government and Politics office (Tydings 3140) or to your professor's office (Tydings 2126E) no later than 10 a.m., Monday, May 20, 2001. Absolutely no papers will be accepted after 10 a.m. on the dot without a certified illness or family emergency. DON'T EVEN THINK OF ASKING FOR AN EXTENSION--EVEN OF FIVE MINUTES FOR ANY OTHER REASON.
Every student must include the following statement on a separate page:
I have read and understand the University's policy on plagiarism and realize that the minimum penalty for submitting work that is not entirely my own is failure in the course. I certify that I have not committed plagiarism on this assignment.
Signed and dated,
(your signature) _________________________
REMEMBER THAT YOU MUST SIGN THE PLAGIARISM STATEMENT AND YOU MUST INCLUDE IT WHEN YOU SUBMIT THE EXAMINATION. ALSO, REMEMBER THAT YOUR SPELLING AND GRAMMAR WILL AFFECT YOUR GRADE. IF YOU DON’T HAVE THE PLAGIARISM STATEMENT, YOU WILL RECEIVE AN F FOR THE ASSIGNMENT. THERE WILL BE NO EXCEPTIONS.
The strategic situation is as follows: Cliffhanger has won victories against Republican opponents by modest margins in most of his seven election campaigns since he won a traditionally Republican seat in 1990. His average margin of victory has been 57.8 percent, but this masks considerable variations in his contests. His electoral history is as follows:
1990 53.0
1992 59.3
1994 51.1
1996 52.3
1998 68.5
2000 63.9
Cliffhanger’s vote share increased in 1992 when Bill Clinton and the entire state Democratic party scored impressive victories in Transylvania. Voters punished George Bush and the Republicans for a depressed economy. But the anti-Clinton backlash of 1994 cost the states’s two ranking Democrats, Governor Paul Dracula and Senator Victor Fang, their seats and almost led to Cliffhanger’s defeat. He didn’t fare much better when Clinton carried his state two years later. But Cliffhanger came back strong in 1998 to defeat Republican activist Lois Blood by a wide margin. Blood is a small businesswoman who is a strong fiscal conservative but a social liberal who supports abortion and gay rights.
In 1990 and 1994, State Senator Roger Wilco, one of the most vocal conservatives in the state legislature, ran against Cliffhanger –almost beating him in 1994. Wilco passed up the opportunity for a third try in 1996. Instead, he lost a close Republican primary for Governor. In 2000, Cliffhanger’s vote margin slipped just a bit as Al Gore won 60 percent of the state’s votes, with a similar share in the 4th Congressional District. The Republicans recruited Amityville Mayor Michael Horror to run against Cliffhanger. He beat Blood by 55% to 45% in the Republican primary, but spent $750,000 to score this victory and was badly outspent by Cliffhanger in what was generally a good year for Democrats in Transylvania. They won a Senate seat (handily), kept the Governorship (barely), but lost control of one house of the state legislature.
To date only Blood has announced against Cliffhanger, but Republican leaders are trying to convince a reluctant Wilco to make a third stab at the seat. Wilco is currently serving as Deputy to Homeland Security Secretary Tom Ridge. On the other hand, Democratic leaders worry that if Cliffhanger does not make the race, Wilco certainly will and the seat will change parties. With only six seats separating the Democrats and Republicans in the House, Democratic House leaders worry that their best candidates may be tempted to run for the Senate, where they would be more likely to be in the majority (as the Democrats are now) rather than in the minority. Cliffhanger has said that he would like very much to be in the majority (some pundits believe he really wants to be a candidate for Vice President).
There have even been rumors that Cliffhanger might be offered the position of Majority Whip if the Democrats retake the House (Whip Nancy Pelosi would move up to Majority Leader and Minority Leader Dick Gephardt would move up to Speaker). Yet, despite these entreaties, Cliffhanger believes that his 2000 opponent might be more formidable this time because of the very high level of popularity for President Bush. Already, Blood is preparing for a more aggressive campaign by raising money early through her new fund-raising organization, the Blood Bank. Wilco’s organization from 1996 is still active and some pundits believe that if he saw 2002 as a good Republican year in Transylvania, he would get into the House race as preparation for another run for Governor in 2004.
Senator Stein has been a Washington fixture since 1984 and has won handily in two of his three races. In his first contest, he replaced a retiring Republican and carried 52 percent of the vote, even as the Democrats captured the state legislature for the first time since 1936. Six years later, as the Republicans swept Transylvania, he took 65 percent of the vote, winning all but five of the state's 25 counties. In 1996 Stein captured 57 percent of the vote to win a third term. Stein appears to be well entrenched, but some people in the state do not think him to be invincible.
Stein will be tied up in Washington for much of 2002. He has become the Republicans’ chief spokesman in the Senate on homeland security–and he is also Ranking Minority Member of the Senate Appropriations Committee. Yet, Stein has taken considerable flack in Washington for his willingness to negotiate with Democrats. He was the only Senator from his party to vote for President Clinton’s budget in 1993 and was threatened with the loss of his position as Chair of the Appropriations Committee when he opposed the Balanced Budget Amendment to the Constitution in 1995. He voted against conviction of President Clinton, saying that the Senate trial “is an incredible waste of time when we have important legislative business to do.”
While he has voted against the administration 85% of the time, some party colleagues do not find him confrontational enough. He once called former House Speaker Newt Gingrich a “legislative terrorist.” As a result, Stein will be challenged in the Republican primary. A well-known former astronaut, Harvey Wallbanger, served two terms as Governor from 1981 to 1989 and was Secretary of the Air Force from 1990 to 1993 before returning to the state as President of Modern Dynamics, a defense contractor that is the state's largest employer, has announced he intends to challenge Stein in the Republican primary. Wallbanger will run as a "real Republican," arguing that Stein too often votes with the Democrats in the Senate. “In these days when America is under threat from abroad, we need someone in the Senate who understands the military challenge that confronts us,” Wallbanger argues. Neither Stein nor Cliffhanger served in the military and they just don’t understand the threat from abroad, he charges.
Cliffhanger worries that if he does not run for the Senate this year, he never will get the chance since the other Senate seat was captured in 2000 by his 38-year-old Democratic colleague, Martin Safe. He recognizes that the Senate race will be difficult compared to running for the House again--especially if there is another Republican tide in the nation (and in Transylvania) as the President’s popularity continues to be very strong.
In analyzing Cliffhanger's options, consider the various aspects of the course relating to both House and Senate elections. You MAY NOT assume that anything of a scandalous nature (or the like, such as a candidate dying or getting a divorce in the middle of the campaign). Good luck (to both you and Cliffhanger).