2004:
PECOS4020 Research Methods, Thuesday December 2nd 2004, 4 hours

Answer both part 1 and part 2.

Choose three of the four following questions

Part 1 (60% of the total grade on the written exam)

Question 1: 
In the article “Coming to terms with the past: A framework for the study of Justice in the Transition to Democracy”, Jon Elster writes:

The dependent variables are the actions taken by the political forces in presence after the transition with respect to the prosecution, amnesty, restitution, rehabilitation, truthfinding and so on. The independent variables include the beliefs and values of the actors, as well as the constraints under which their policy choices are made

Give an account of this way of explaining human actions, and discuss what role peoples subjective conceptions of right or wrong may play in such explanations.
Title: Explanations

Question 2:
What according to E.C.Carr, is a “historical fact”? You may use examples for purposes of illustration.
Title: Historical facts

Question 3:
A distinction may be made between qualitative ( “interpreting content”) and quantitative (“counting content”) content analysis. Give an account of differences and similarities between the two. When would you choose a qualitative over a quantitative approach? Justify your answer.
Title: Content analysis

Question 4:
What characterizes a topical interview) Discuss strengths and weaknesses of this kind of interview for different research purposes.
Title: Topical interview

I

A student wants to study quantitatively what determines the lethality of wars, and have collected data on some civil wars. (Fictive) data on battle deaths are given in Table 1.

Table 1. Battle deaths in civil wars, 1981-2000

Conflict no.	Battle deaths (in thousands)
1	150.0
2	87.8
3	1.8
4	67.5
5	2.0
6	4.2
7	1.8
8	10.2
9	5.5
10	55.5
11	13.0
12	1.4
13	9.8
14	13.0
15	42.8
16	8.7
17	1.9
18	38.0
19	56.0
20	2.1
21	8.0
22	21.1

a) What is the student’s sample size? 

b) Calculate the median for the battle deaths data.  

c) Calculate the sample mean for the battle deaths data for the first ten observations. Formulas are given below.  

d) Calculate the sample variance and standard deviation for the battle deaths observations for the first ten observations. Formulas are given below.  

e) Explain what is meant by Type I and Type II errors, and define ‘level of significance’ in terms of this.  

The student wants to investigate the hypothesis that civil wars in the 1981–2000 period were less lethal than wars in the 1960–1980 period. An earlier (fictive) study has shown that the mean ln(battle deaths) for the 1961–1980 period was 4.6. The mean of ln(battle deaths) in the student’s sample is 2.4. You may assume that the standard deviation σ for ln(battle deaths) is 1.4.

f) Test the hypothesis that civil wars have less battle deaths than colonial wars, using the t-test. Specify the null hypothesis and the alternative hypothesis. Use the .05 level of significance, and state the conclusions from the test. The formulaes you need are given below. Critical values for the t distribution with 21 degrees of freedom is given in a table below.

Table 2 Critical values of the t distribution

<!--[if !supportMisalignedColumns]--> <!--[endif]-->

d.f.	Level of significance for one-tailed test
.10		.05		.025		.01		.005
Level of significance for two-tailed test
.20		.10		.05		.02		.01
21	1.323		1.721		2.080		2.518		2.831

II

Table 3 reports one set of results in Jacob Shamir and Khalil Shikaki’s article on “Determinants of Reconciliation and Compromise Among Israelis and Palestinians” in JPR no. 2, 2002.

Observational units are respondents in a survey.

The dependent variable is a “Political Compromise Scale” (PCS) which measures the repsondents’ attitudes to compromise. The scale ranges from 0 to 100, such that 100 is the most supportive for reconciliation with the other group, and 0 the least. Only the analysis for the Israelis is reported here.
 

a. Which variables are estimated to be significantly different from 0 at the .05 level? (use the z-statistic, which has a standard normal distribution).

b. Calculate the estimated score on the PCS for a female respondent, aged 50, with 5 as income, 15 as education, 1 as religiosity, and 0 for all the ethnicity variables.  

c. By how much will the PCS change if ‘Gender’ is changed from 1 to 0?

d. By how much will the PTS change if the Education variable is reduced by five units?

e. The Age variable ranges from 20 to 70, whereas the Gender variable can take the values 0 and 1. Which variable has the largest substantive effect on the PCS?

f. The ‘Ethnicity variable’ is a set of three dummy variables to represent the ethnic background of the respondent. The ‘Ashkenazi’ variable has the value 1 if the respondent is Ashkenazi, and the value 0 otherwise. The ‘Sephardi’ variable has the value 1 if the respondent is Sephardi, and the value 0 otherwise. The ‘Israeli’ variable has the value 1 if the respondent is 2^nd generation Israeli, and the value 0 otherwise. ‘Arab’ group is the reference category: If the respondent is Arab, all the three dummy variables have the value 0. Which ethnic group is most favorable to reconciliation (among the Israelis)?

g. Give your opinion (in words, exact calculations are not necessary for this sub-question) on the extent to which there is a difference between Ashkenazis, Sephardis, and Israelis in their opinions on reconciliation.

h. Use the results from this analysis to discuss problems with using linear regression for a dependent variable that ranges from a minimum to a maximum value.

Table 3: OLS regression, dependent variable: Political Compromise Scale

<!--[if !supportMisalignedColumns]--> <!--[endif]-->

Variable	Parameter estimate		Estimated standard error
Constant	87.42		5.97
Gender (male as reference)	1.03		1.81
Age	0.07		0.06
Income	1.18		0.80
Education	0.29		0.30
Religiosity	–7.89		0.96
Ethnicity
Arabs		Reference category
Ashkenazi	–32.56		3.00
Sephardi	–34.49		2.95
Israeli (2nd generation)	–29.01		3.45
N=525
R2=0.39

1:The natural log of population was taken due to skewed distribution

Table 4 Critical values of the standard normal distribution

Level of significance for one-tailed test
.10	.05	.025	.01	.005
Level of significance for two-tailed test
.20	.10	.05	.02	.01
1.328	1.645	1.960	2.326	2.576

 

 

2005:
PECOS4020 Research Methods, Thuesday 8 December 2005, 5 hours

Answer both part 1 and part 2. 

1. Explain what is meant by the following statement: “Correlation does not imply causality.” You may use examples for purposes of illustration.

 

2. What does it mean to ‘control for third variables’? How can one control for third-variables in small-N studies?

 

3. A particular text may be used as a ‘report’ or as a ‘remain’. Explain the difference between the two. You may use examples for purposes of illustration.

 

4. Give an account of Jon Elster’s understanding of what it means to explain human actions. Discuss what role people’s subjective conceptions of right and wrong may play in such explanations.

 

Part B (50%)

 

Answer both questions

 

1.

 

Throughout: Include your intermediate calculations in the answer.

 

 

Table 1. Democracy and Average Income

Country	Ln GDP per capita	Democracy index
Greece	9.5	0.98
Cyprus	9.6	0.96
Bulgaria	7.3	0.96
Moldova	6.5	0.77
Romania	7.3	0.90
Russia	7.8	0.73
Estonia	8.4	0.76
Latvia	7.9	0.77
Lithuania	7.6	0.94
Ukraine	6.8	0.85
Georgia	6.2	0.67
Azerbaijan	6.2	0.33

Table 1 reports ln GDP per capita and democracy index for a selection of East and Central European countries (ln is the natural logarithm log_2.7). The figures are for the year 2000.

 

a)      Calculate the median and the sample mean for the ln GDP per capita variable. Formulae are given below.

b)      Draw a scatterplot with ln GDP per capita along the X axis and the democracy index along the Y axis

c)      Calculate the covariance between ln GDP per capita and the democracy index using the formula given below. Mean for the democracy index is 0.80.

d)     Calculate the correlation coefficient between ln GDP per capita and democracy from the covariance. The formula is given below. Sample standard deviation for ln GDP per capita is 1.13. Sample standard deviation for democracy is 0.181.

e)      When fitting an OLS regression with the democracy index as the dependent variable and ln GDP per capita as the independent variable, we obtained the following results for the regression coefficient (b) and the constant (a):

(boks inn her)

Test the hypothesis that the level of democracy is increasing in ln GDP per capita. Specify the four steps in hypothesis testing outlined in Hinkle et al. Use significance level

f)       Explain what is meant by Type I and Type II errors and define ‘level of significance’ in terms of this.

g)      What is predicted score for the democracy index when ln GDP per capita is 8?

h)      Discuss in words what would happen to the correlation coefficient and to the estimated regression coefficient if Azerbaijan was removed from the dataset. Justify your prediction.

2.

 

The following facsimile is taken from Russett, Oneal & Cox, 2000: ‘Clash of Civilizations or Realism and Liberalism Déjà Vu? Some Evidence’, Journal of Peace Research 37(5). The table reports the results from estimating a logistic regression model on a data set where annual observations of pairs of countries is the unit of analysis. The occurrence of militarized interstate dispute is the dependent variable. The table reports the regression coefficient and its standard errors (in parantheses) for each of the independent variables. The independent variables in the model are:

 

• SPLIT           whether the two countries belong to two different ‘civilizations’
• CONTIG       whether the two countries are contiguous (neighbors)
• ALLY            whether the two countries are allied
• CAPRATIO   log ratio of the military capabilities of the stronger country
                         over the weaker
• DEM_L            democracy score of the least democratic country in the pair
• DEM_H            democracy score of the most democratic country in the pair
• DEPEND_L      trade dependence - the value of the bilateral trade divided by
                         GDP of the least trade-dependent country in the pair

(tabell inn her)

a)      Which factors are estimated to decrease the risk of militarized disputes?

b)      The article is written as a response to Samuel Huntington’s claim that civilizational differences increase the risk of violent conflict. Discuss the implications of the Russett, Oneal & Cox analysis for this claim.

c)      What is meant by ‘one-tailed tests’ in the note to the table? What are the implications for the reported significance levels of using one-tailed tests?

d)     The democracy variable ranges from -10 to 10, whereas the alliance variable can take the values 0 and 1. Which variable has the largest substantive effect on the log odds of disputes (i.e., which variable has the potential to make the largest change)?

e)      Why does the analysis support the dyadic democratic peace hypothesis and not the monadic one?

f)       In an additional analysis reported elsewhere in the article, Russett, Oneal & Cox replace the SPLIT variable with a dummy variable denoting whether one of the two countries in the pair belonged to the ‘Western’ civilization and the other to the ‘Orthodox’ civilization. This dummy variable is positive and significant. Discuss the extent to which this result supports Huntington’s Clash of Civilizations thesis.

Formulae:

(formler inn her)

---

Redaksjon: Redaksjon for informasjon om studietilbudet ved UiO
Dokument opprettet: 02.07.2007, endret: 10.12.2007Administrer dette dokumentet