Political Science Methods
Political Science Methods POLS 2220
Popular in Course
Popular in Political Science
This 53 page Class Notes was uploaded by Clark Grady on Sunday October 25, 2015. The Class Notes belongs to POLS 2220 at University of North Carolina - Charlotte taught by Staff in Fall. Since its upload, it has received 37 views. For similar materials see /class/228971/pols-2220-university-of-north-carolina-charlotte in Political Science at University of North Carolina - Charlotte.
Reviews for Political Science Methods
Report this Material
What is Karma?
Karma is the currency of StudySoup.
Date Created: 10/25/15
Doing a Research Paper 272008 The Most Impt Part of this Course is the Research Paper Intuitiver it would seem that you start off with a research question you want to answer You build a causal model that you think describes how the process of interest works Derive the hypotheses from the model You then figure out what data you need to answer it 272008 Pts Behind in latest poll Decline in poll over past 5 days Days until the election of negative comments per of mentions in the evening minute of news speech Incumbent Female 272008 When does a candidate go nega ve Hypotheses 1 2 3 4 5 6 272008 The steeper a candidates decline in the polls the greater the ratio of negative comments The greater the percentage distance between the candidate and the leader the grater the ratio of negative comments The lower the percentage of mentions in the press the greater the ratio of negative comments The fewer the number of days until the election the greater the ratio of negative comments Incumbents have a lower negative ratio than challengers Women have a lower negative ratio than men Reality It s easier when we are first learning to start with the data You have data on the following variables of college credits Sex of the student Birth order of student firstoldest middle youngest Grade in POLS 2220 High School GPA What are your possible research questions What is your model 272008 5 Female First Grade in HS GPA POLS Middle 2220 of College Credits 272008 How to Proceed with your Project 1 Pick a dataset from the 8 drive that has variables that interest you For example if you are interested in comparative politics you might choose the Peru dataset If you are interested in legal issues you might choose the judicial dataset 2 Go to the codebook of the dataset and find several variables that you might want to explore For example you might be interested in explaining why countries vary in their support for civil liberties or what factors influence a Supreme Courtjustice s vote 3 From the codebook identify several variables that might be important in explaining the variation in your dependent variables 272008 7 4 272008 How to Proceed with your Project 2 Go to Google Scholar or other search tool discussed by the librarian in lab httpscholarqooqlecom Search for books and articles that relate to the question you want to answer If there is one pick a recent book on the topic and the most recent scholarly article Check out the book and print the article Research Design The Many Pitfalls of Doing Research Four Types of Studies Exploration We are new to a topic and are trying to find out how to study it What factors might be important in explaining what we want to explain eg early research concerning what Ale was and the sources of infection Description We are interested in describing a phenomenon without necessarily looking cause and effect what where when how eg Where is the spread of HIV greatest and what is the yearly infection rate Explanation HOW and WHY eg Understand how epidemics spread Prediction Policy related research can we change what is happening Can we prevent or slow an epidemic Exploratory No Child Left Behind Act allows students in schools not making AYP to transfer but almost no one transfers to better schools Why First we reviewed the existing literature concerning how parents make choices among schools and how schools respond to these choices Second we reviewed CMS documents amp news accounts of the choice program Was CMS focusing on the target families in an appropriate manner Third we held focus groups Parents Why CMS parents did or didn t choose What school factors important Teachers Did they help or discourage choice Why Descriptive Examined 100000 records of CMS students and compared the characteristics of students who transferred and those who did not Examined the patterns of transfers Did whites transfer to predominantly white schools and blacks to predominantly black schools How important was distance to different groups How important were school test scores to different groups Were students who transferred low average or high performing SES of families who transferred and the SED of families that did not Explanation Using hypotheses developed in exploratory and descriptive studies we developed hypotheses and tested with Survey questions in the Urban Institute Mecklenburg County Survey probability sample Interviews with parents at the CMS Choice Fair and discuss why they are looking for a different school purposive sample CMS student record data on who chose to transfer and how this affected their future test scores Policyrelevant Research Research shows choice helped learning Choice probably should be encouraged Focus groups and quantitative analyses suggested that parents did not believe they could choose well and minority parents distrusted CMS CMS has a credibility problem Focus groups with teachers and interviews with principals showed that CMS discouraged parents from choosing an alternative school Quantitative analyses indicated that most choosers were higher performing students Attempted Policy Solutions Design a computerassisted decision system to help low income parents and less well educated parents choose a better school for their child Bring the decision system to parents neighborhood Offer cash incentives CCTs to participate But parents rarely participated despite 7 innan vng Types of Explanations Nomothetic the goals is to find the variables that explain and predict social phenomena large of cases Usually quantitative research Supreme Court justices votes Justice s political party Political party of the appointing president Previously a prosecutor Prior record as a judge Illegal Political Activity Peru Was the individual a student Level of personal efficacy Past political activity Married or single Domestic terrorism in a country Ethnic amp religious heterogeneity Democratic government How long an independent nation Region of the world Idiographic Explanations Exhaustiver list all the causes of a condition or an event trying to make sense of what happened Why did the USSR collapse Why did Romney s campaign fail Why did a particular family choose a particular school Usually qualitative research lndepth interviews Document analysis Participant observation Snowball sampling May not be representative of general phenomena Easily biased by preexisting ideas and beliefs I believe it therefore I see it eg Margaret Mead amp sexual practices in the South Pacific We rarely see idiographic studies where results deviated from expectations Conditions for Causality Association between X amp Y When X occurs Y is more likely Hours studied amp grade on midterm Persons who selfidentify as liberals amp willingness to eliminate the Bush tax cuts Teacher s SAT score amp amount students learn Nonspurious association Do higher SAT scoring teachers teach at higher SES schools Time Order X occurs before Y Rationale there is a reason why X influences Y 10 Unit of Analysis The unit of analysis is the objects we are studying Individuals hours a student studied 9 grade in class Classes give a common exam to different sections of 2220 and grade the teachers Schools different universities and earnings after college Political parties importance of ideology to recruitment Regions of US or regions of the world region and political party iden ca on Countries military budget amp likelihood of going to war 11 Ecological Fallacy When we attribute to a member of a unit the collective attributes of the unit Census tracts with higher of minorities have lower literacy rates Can we conclude that minorities have lower literacy rates Students at high schools with higher average family incomes have higher SAT scores Can we conclude that higher SES students have higher SAT scores California congressional districts with a higher incomes California had higher vote for Obama Can we conclude that the wealthy people in those districts voted for Obama 12 Reductionism Single cause explanation of a complex phenomenon Marx Economic determinism Money decides politics Everything in America comes down to race Men are from Mars Women are from Venus Incorrect example in Babbie reading sociobiology and evolutionary psychology are genetic determinism Sociologists don t like EP so they misunderstand it Babbie believed it therefore he saw it 13 Time Problem for Research Problem Much of what we study deals with change but we have limited time and money Example My tolerance study in high schools I compared 9th grade and 12th grade students at hi h schools and assumed that the di ferences I observed between 9th amp 12th graders measured in the same year were caused by attending a public or a fundamentalist high school for 4 years But it may have been the case that the two cohorts were different when they entered This could have occurred because of a particular event such as 911 or Columbine H8 or a change in admission criteria to fundamentalist schools 14 Crosssectional Analyses Tolerance study was the result spunous Peru study on political participation Political efficacy and participation Which is the cause and which the effect Terrorism in countries Do failed states have more terrorist groups or do terrorist groups create state failure Do wealthy countries protect human rights or does the protection of human rights create economic growth Does Buddhist religion in a country lead to higher economic growth 5 Longitudinal Study Idiographic study of the rise of the Republican Party in the South Civil rights demonstrations Goldwater advocates states rights to decide racial issues such as separate but equal Nixon s southern strategy appeal to whites racism and identify Democratic party with integration Culture wars Bush and wedge issues such as gay rights creationism traditional roles for women abortion Problem Difficult to falsify We evaluate by comparing to alternative explanations Other Longitudinal Analyses Do high school students know more in 2007 than high school students knew in 1967 Examine SAT scores over time General knowledge tests over time Cohort Analyses follow different groups of people through time Example compare political attitudes of students who graduated in 65 75 85 95 2005 and how these change as they the people in each cohort grow older Panel studies study exactly the same people over time Problem of attrition Our study of SmartChoice users Lowincome families move frequently 17 Our Method Sometimes Determines our Findings Floyd Hunter Asked various people in Atlanta Who is influential in Atlanta Assumption Only a few people are really influential and they are influential across issues Robert Dahl Took the issues with the most press coverage in New Haven and then studied who was influential on each issue Found different people Bachrach amp Baratz Dahl did not look at agenda control or nondecisions Segregation women s rights socialism 8 Can you prove anything with statistics You can mislead people with statistics Most rich people are Protestants But most Americans are Protestants The difference between more and more likely The Tennessee class size experiment proved small class sizes improve learning in elementary school But all the difference in learning occurred in kindergarten amp the 1St grade After that class size did not matter The average American income has never been higher But adjusted for inflation the median income is lower today than in 2000 19 You can mislead people by asking loaded questions Should women be allowed to kill their children by having a partial birth abortion Should the government force a young girl to carry a fetus to term if her pregnancy was caused by rape or incest Should there be racial quotas in hiring Do you believe a humans descended from monkeys Should a member of a religious cult be elected President of the United States 20 But Most of the time we can t reject the null hypothesis Most new drugs never make it even to animal testing Even fewer get to the market Almost all policy experiments show no statistically significant impacts Although almost all published studies show statistical significance most studies are never published Brownfield reclamation does not increase property values around them to a statistically significant degree after controls Brownfield cleanups did not reduce health risks in the areas Brownfields cleanups did not reduce 21 crime in nearby neighborhoods Understanding Research Design Helps Us Make Better Decisions Example Most childbirths in the US that have severe complications occur in hospitals Why should we not conclude that women would be better off in birthing centers Most children in US are born in hospitals so the question is Is the percentage of complications greater in hospitals Remember column totals 100 o Are women who have highrisk pregnancies more likely to choose hospitals If so should we exclude those from our data Are there ethnic and SES differences in who uses birthing centers Correct question When comparing pregnancies with similar risk levels are there higher rates of morbidity and mortality in hospitals or in birthing 22 centers Statistics for Survey Research Key Terms Quantitative research Parameter Qualitative research Sample statistic Random sample Sampling bias EPSEM Response bias Nonrandom samples Random sampling error Convenience sample Standard error SHOWba sample Deviation from mean Quota sample z score Key lnformants t score Why is survey research important Anyone interested in politics economics or sociology wants to know about people s behavior expectations beliefs and attitudes Survey research helps us find out these things Importance of random samples amp probability theory Nonrandom samples Sampling Procedures 1 Nonrandom All members of the population do not have an equal chance at being in the sample Quota early Gallup studies Sample on demographic characteristics 40 white females 10 black females etc Convenience Accessibility not a problem if trait is universal but if it were universal then you probably would not be studying it Snowball The first interviewees name others to interview Purposive Sample Purposive orjudgmental Samples You sample the most visible or most easily identified members of a sample For example you want to sample left and right wing student activists You might go to group meetings of The Green Party and the Young Americans for Freedom You believe judge these individuals are reasonably representative of other students in left and rightwing organizations Purposive samples are as good as the judgments about how representative a particular individual is of what you are studying Sampling Procedures 2 Random EPSEM Equal probability of selection methods Strictly random rarely done Stratified random eg stratify by state school type city Key advantage is that you can use probability theory and statistics to tell you the likely size of your sampling error Probability Theory tells us how frequently a distribution could have occurred by chance one stan dev s 34 25 9539 3925 Normal curve is everywhere steps bird feeder height GPA rainfall overtime Key Terms 1 Parameter characteristic of the entire population Sample statistic an estimate of the population parameter u population mean 6 population standard deviation 62 population variance X sample mean s sample standard deviation s2 sample variance Sampling bias occurs when the sampling technique leads to an unrepresentative sample Random Sampling Error also called standard error of the Random sampling QSthal39lsat led the standard error is a function of the variation in the population 6 and sample size n on As 6 increases sampling error increases As n increases sampling error decreases Public school students mean 163 n999 std dev474 474N999 15 PriV sch students mean 167 std dev440 N 318 RSE 440318 25 95 con dent that the population mean for pub sch students is between 1615 and 1645 Private school mean between 1645 and 1695 I lUVV LU UdlUUldLC SUIII Ul SqulCS variance 02 standard deviation 0 amp Laborer Wage RS FDeviation from p Deviation SQ 1 1O 6 36 2 12 4 16 3 14 2 4 4 16 O O 5 18 2 4 6 2O 4 16 7 22 6 36 16 02112716 O39 164 112 Random Sampling Error 47151 Sum of squares Sample Size and RSE Sampling error on As sample size increases error decreases but not to the same extent eg o 4 Sample size 25 45 8 Sample size 100 410 4 In this case quadrupling the sample size only outs sampling error by half This explains why pollsters often seem to have very small samples Note The RSE does not include a term for the population size If the variation in the samples is the same for NC as for the entire nation then the same sample size is equally valid for NC and the nation Central Limit Theorem Central limit theorem with repeated samples of the population with replacement the mean of the means will approximate the population mean From the next slide I will take the mean of 5 samples of random numbers from O to 99 The population mean 495 Sample means are 514 495 515 484 459 mean of these ve means is 4936 25 51 1 1 83 77 55 89 1 41 81 96 95 63 28 38 4 76 8 10 77 514 495 25 72 98 99 14 68 54 17 86 0 64 25 6 88 38 54 84 17 47 44 516 484 58 45 30 39 14 44 77 45 70 37 459 Deviation from the mean and Z Deviation from mea icore XI u individual case value population mean Z Score X mc Deviation from the mean divided by the standard deviation Z score of 1 96 is 5 of the distribution Z score of 196 means 975 of cases above the individual score Wage Earner OOONCDU lFOON AA 0 n11 Wage Deviation from mean 8 5 3 2 SQ Deviation from mean 64 25 64 Sum of squares 2045 Z Score 186 116 070 046 012 0 012 046 070 116 186 02 1859 0 431 Percentiles Using the Zscore table in your book figure out the percentile ranking for each laborer p 119 Wage Wage Z Earner Score Percentile 1 2 186 3 2 5 116 12 3 7 070 22 4 8 046 32 5 9 5 012 45 6 10 0 50 7 105 012 55 8 12 046 68 9 13 070 78 10 15 116 88 11 18 186 97 Student s tdistribution We almost never know the value of o To allow us to substitute 8 for 0 we make a correction based on research by a graduate student who worked for Guinness ale As the sample size decreases the correction is larger For large samples over 120 there is little difference between Z and t Standard error This is a measure of how closely the sample mean estimates the population mean If we know n and 0 then we can estimate this As we rarely know 0 we use 3 and the tdistribution to make the correction Let s say that we take a random sample of 100 students and nd that the mearrGPA 280 with s 5 Standard error 5 1100 05 Using tdistribution we nd that it is likely that 95 of the samples are likely to fall between 28 198705 or between 27 29 This is what it means when the margin of error in a poll is i 3 Summary Random sampling takes advantage of very powerful statistical tools and probability theory Studies based on nonrandom samples generally should be doubted with respect to their generality to other members of the population Comparative Politics Dataset Codebook Variable Label Level country Name of country Nominal aids AIDS cases per 100000 people 1997 Scale CiVlib Freedom House civil liberties score Scale C mPUISe Compulsory voting used Ordinal democ ls country democratic Ordinal dem th Percent of other countries in the region that are democracies Scale dem th3 of democracies in region 3 cats Ordinal diVorce divorces as of marriages 1996 Scale econdeVs Econ devel 3 cats Ordinal ethhet Index gIi ti 39 39 highmore39 Scale ethhet3 Ethnolinguistic heterogeneity 3 cats Ordinal fhosreV Freedom House democ rating reversed 20032004 Ordinal gdp1000 GDP per cap 1000 Scale decaPZ GDP percapita 2 categories Ordinal Gini Gini coefficient Scale hdi2001 Human development index 200 Scale hLQdP High gdppcap dummy Ordinal indy Year of independence Scale interaCt regimehigdp Scale legdom Legis dominance of largest party Ordinal natSize Size of country by population 3categories Ordinal 0 Did ratio of fuel exports to total exports exceed 50 Ordinal Open Imports and exports as a share of GDP Scale Party Number of parties 3 categories Ordinal Pdwomos of women in lower house of parliament 2003 Scale POMS Political rights score Ordinal P P2002 Total population 2002 Scale PLSYS PR electoral system Nominal regime ls Country democracy or dictatorship Nominal region Region of the world Scale relcat Type of secular or religious society Ordinal rural Rural population of total population Scale seats Percent of seats in lower legis hse held by largest party Scale smoking Cigarette consumption per adult 1970 72100 1990 92 Scale typerel Religious culture Nominal unions union density 1995 Scale urban Urban population of total Scale VLre39 religion 39very39 important Scale VOteVaP Turnout vap 1990s Scale W mYL2 Women39s suffrage beforeafter 1920 Scale WVkJeSt Elections Day of work or rest Nominal POIitYIV Polity Democracy score Scale CiViIVVar Civil war occuring in country Nominal IntVVar Country involved in international war Nominal FD39 Foreign direct investment as a percentage of gross domestic product Scale Aid Foreign aid as a percentage of gross doemstic product PhYSInt Physical integrity measure of human rights Military Total military personnel Scale Ordinal Scale Variables in the working file Variable Values Value Label civlib 1 Least free 7 Most free compulse 0 no 1 yes democ 0 no 1 yes dem0th3 1 quotless than 30 percentquot 2 quot30 percent to 60 percentquot 3 quotMore than 60 percentquot econdev3 1 quotLeastquot 2 quotMiddlequot 3 quotMostquot ethhet3 1 quotLoxw 2 quotMediumquot 3 quotHighquot fh03l39eV 10 quotLeast Democraticquot 70 quotMost Democraticquot gdpcap2 1 Low 2 High higdp 0 Low 1 High legdom 1 Low dominance 5 High dominance natSiZe 1 Small Less than 1 million 2 Medium 129 million 3 Large more than 30 million oil 0 No 1 Yes party 0 None 1 One party 2 More than one party polrts 1 Fewest rights 7 Most rights prsys 0 No 1 Yes regime 0 Dictatorship 1 Democracy region 1 SubSaharan Africa 2 South Asia re l cat typerel womyr2 wrkrest PolitylV CivilWar ntWar Physlnt O O OLOOOI U IJgt0I 0I LOOOI UIJgtC0 o 00 OOO O East Asia South East Asia Pacific lslandsOceana Middle EastNorth Africa Latin America Caribbean Eastern Europe Industrialized Most secular Moderate Most Religious Roman Catholic Protestant Orthodox Jewish Muslim Hindu Eastern Other Missing 1920 or before After 1920 Day of work Day of rest Least Democratic Most Democratic No Yes No Yes Fewest rights Most rights