|Faculté||Faculté des sciences de la société|
|Département||Département de science politique et relations internationales|
|Cours||Introduction to the methods of political science|
- 1 Two Ways to Collect Information
- 2 Poll survey
- 3 Types of data collection through interrogation
- 4 Substantive problems of information gathering through interrogation
- 5 Four aspects of the polling survey
- 5.1 Types of questions
- 5.2 Formulation of questions
- 5.3 Batteries of questions
- 5.4 Methods of administering the questionnaire
- 6 Phases preceding data collection
- 7 Sampling
- 8 Sampling errors
- 9 Qualitative techniques
- 9.1 Types of Qualitative Research
- 9.2 Participant Observation
- 9.3 Qualitative interview
- 9.4 Reading - Documents
- 10 Annexes
- 11 References
Two Ways to Collect Information[edit | edit source]
There are two main ways of collecting data in the social sciences:
- Observation: directly observable behaviours such as tension. It is a relatively rare way of gathering information.
- Questioning: The choice depends on several factors, i. e. the choices made during the research question, which depends on the question of the research question. It is the choice of how to collect the data; for example, if you are more interested in behaviours than opinions or values that are not observable.
Poll survey[edit | edit source]
Most political science research involves questioning. The preferred way in the positivist and post-positivist paradigm is through sample surveys.
If you are interested in studying electoral behaviour, you go through data that has been collected by someone else through a sample survey or questionnaire.
Corbetta argues that the sample survey is a way of collecting information by interviewing individuals who are the objects of the research. Individuals belong to a representative sample through a standardised interrogation procedure to investigate relationships between variables.
The sample survey is based on questioning, asking people questions; it is the most commonly used and crucial instrument.
A sample survey is therefore:
- collect information by questioning people: asking and formulating questions.
- individuals who are the subject of the research: this makes it possible to make a distinction, because there are other methods of data collection that involve questioning and/or the individuals being questioned are not the individuals or subjects that we wish to study and in particular "key information" interviews, i. e. the questioning of people who can give us information about the subject under study.
- representative sample: we question part of the population, if we are interested in the electoral behaviour of the Swiss, we are not going to question the seven million Swiss, that's why we only question part of the population. However, one can fall into selection errors and especially sampling errors. Therefore, the sample survey is conducted by interviewing a representative sample of the population to be studied.
- Standardized procedure of the way of asking the question: there is a standardization of the procedure, because in quantitative research we want to standardize the information, because we want to compare the answers between them, that's why we ask the same questions to everyone and in the same way. This aims to be able to process the answers through the tools offered by statistics. The choices depend on the perspective from which you choose, the choice between a quantitative and qualitative approach is dictated not only by the research question, but also by the preferences and the tradition in which you choose. In other words, standardization allows statistical processing and comparison.
- study the relationships between variables: we put the concepts in relation, then we have the operationalization and through this procedure we arrive at variables that allow us to study relationships between variables. This distinguishes "sample survey" from "survey". A survey is not intended to test hypotheses, whereas a sample survey can be used to test hypotheses. The survey is mainly descriptive. We want to empirically test hypotheses, not just explore and describe.
Types of data collection through interrogation[edit | edit source]
Answers: standardised/free[edit | edit source]
There is a way to classify the different types of data collection through the query. The typology is based on the degree of standardization of questions, but also of answers. We can say that when we proceed by interrogation in social science research, we can ask questions that are standardized, that is to say, the same questions for everyone.
Different questions may also be asked depending on the subject matter. For example, men and women could be asked different questions if they want to look at different roles in the family or society.
It is also possible to consider a type of question or ask different people different questions because you are not interested in standardization.
With regard to standardisation, the formalisation and sequence of questions is the same.
Questions: Standardized/Free[edit | edit source]
There is also a distinction between standardized or free questions. We can make sure that the answers are the same for everyone, or we can give everyone a degree of freedom in the answers.
The standardised response presents ways of answering the questions, we ask to choose between different ways of answering them.
Typology[edit | edit source]
- in the questionnaire: questions and answers are standardized.
- free interview: no standardisation of answers or questions.
- structured interview: we have the same types of questions but everyone can answer as they please. The questions are standardised, but we don't give the answers, this obviously has advantages and disadvantages.
These are three types of procedures to test hypotheses through questioning.
Substantive problems of information gathering through interrogation[edit | edit source]
There are a number of substantive issues that need to be addressed.
[edit | edit source]
Standardization stems from the positivist paradigm, one is in an objectivist epistemological approach, one tries to standardize. This has a certain number of contraindications that give rise to problems: when we standardize we win on the one hand and lose on the other with a depersonalization between the investigator and the respondent.
For those who fall within this paradigm, it is preferable that there be differentiation, as the researcher must be separated from the subject.
This objective of trying to come closer to a neutrality of the instrument in order to obtain a comparability of people raises a number of problems, we are in the framework of epistemological reflections.
Reliability of verbal behaviour[edit | edit source]
If behaviours are observed, it is certain that they have been observed with an idea of the validity of what is observed. An entire methodological apparatus must be implemented to reduce the sources of errors related to the interviewees:
- social desirability of responses: individuals tend to respond in a way that offers a positive self-image or in a way that does not offend the sensitivity of the person administering the question.
- absence of opinion: linked to the question of "pseudo-opinions". When we ask individuals questions, if we ask the same question to two individuals with one who has already reflected on the question and formed an opinion and the other person is confronted for the first time with the question, in this case we form an opinion on the moment that may not be one; therefore the absence of opinion is a fundamental problem in the questioning process. If we are in more qualitative approaches, we have time to interact with the person allowing him/her to form an opinion and avoiding the problem of "pseudo-opinion".
Four aspects of the polling survey[edit | edit source]
Types of questions[edit | edit source]
Substance[edit | edit source]
There is an aspect related to the substance of the content of the question and its form.
- Socio-demographic data: there are different aspects addressed, including the socio-demographic characteristics of respondents. It's something you should never forget. Individual characteristics affect other behaviours.
- attitudes: this is an acquired predisposition towards a political issue. This makes it possible, for example, to study the values.
- behaviours: we can observe them, but if we cannot observe them, we must question the individuals. Research is oriented towards people's behaviour, such as asking questions about the participation of individuals. These data are more controllable than "attitudes" because they are less subject to variation and easier to observe. A distinction is made between "factual questions" that deal with facts and "questions on motivations". The idea is that, if possible, it is preferable to ask factual and behavioural questions.
Shape[edit | edit source]
Open questions[edit | edit source]
Respondents are given the freedom to respond; they are less standardized in the response.
Closed Questions[edit | edit source]
This method is more widely used in sample surveys (questionnaire) because it is possible to compare the responses since they are already standardised.
Mixed Questions[edit | edit source]
There is an "other" style answer item and at the same time we have the question closed with the answer modalities presented to the interviewee. In this case, we make the instrument more flexible; we have both standardisation and we avoid forgetting to offer certain response modalities, which is a fatal error, because it cannot be repaired. On the other hand, in case of doubt, this allows an answer to be left open.
Question « cafeteria »[edit | edit source]
Refers to closed questions with a longer list of answers, on the other hand, the response modalities can be single or multiple.
Formulation of questions[edit | edit source]
There are no good and bad questions. However there are a number of tips and tricks to make them more relevant. We try to reduce a number of errors through different recipes.
The manner in which the question is posed affects the answer. In the United States was asked the question on freedom of speech in these terms: do you think that the United States should allow public statements against democracy? Individuals answered 75% "no". Another question was asked: Do you think that the United States should ban public statements that go against democracy? Individuals answered 54% "yes".
The results are quite different while in the first question - people agree not to let people speak against democracy; for the same question asked differently individuals answer "yes" to 54%.
Another question is: if a situation like that of Vietnam were to develop in another part of the world, do you think that the United States should send or not send troops? 18% of people agreed The same question was asked to another group, it is another experimental design by splitting the sample in another way: the United States should intervene or not by sending troops to stop a takeover of the Communists? In this case, 33% said "yes". So the way of asking the question is crucial and changes the answer; in other words, the way questions are asked can create bias.
- Simplicity of the language: we build a question that we will give to the interviewees, we must use a simple language that adapts to the person or group interviewed.
- length of questions: try to ask short questions.
- number of alternative answers: in the context of cafeteria questions one should try not to give too long a list of answers, in other words to limit the number of answers.
- expressions in jargon: when studying subcultures, avoid using their language.
- Ambiguous definitions: avoid them like, for example, "do you have a stable job? »; above all, what is stable work? we must be as specific as possible in formulating the question.
- words with strong negative connotations: if we do a research on the relations between parents and children and the sanctions applied to them, an ill-posed question would be "hit your children".
- syntactically complex questions: avoid negative questions, especially those involving double negation; for example "do you agree or disagree with the following statement: is it not true that workers are in such a bad situation as the unions say today? ". It is best to ask the questions in an affirmative form.
- questions with non-unique answer: "Are your parents religious? »; the problem is that we have two parents and there is a distinction between mother and father.
- non-discriminatory questions: avoid questions that do not make differences between topics, because a question that then leads to an almost unanimous answer no longer gives rise to a variable, but to a constant. In order to work in quantitative research, one needs variations in subjects. For example, "Which of the following groups do you trust most? Everyone trusts relatives; we must expand the possibilities of response.
- In our country 700 priests have declared that the Bible is the message of the poor and exploited therefore of those working in factories, actively participating in trade unions and political organizations in order to obtain more social rights, do you think priests are right? ». There are too many tendentious terms.
- presumptive behaviours: avoid using certain behaviours, such as "for which party did you vote in the last election? ", first of all it is necessary to ask the questioned individual can vote. This type of question is also called "filter questions".
- focus over time: in general, avoid asking vague questions such as "how often do you read newspapers per month? ». it is always difficult to remember and set a benchmark. In general we find a standard question that allows us to specify the context and to fix a temporal reference allowing the compatibility of the answers.
- Abstract questions:"Do you think the death penalty should be applied to particularly serious crimes?" The concept of "serious crime" varies according to the individual and the groups of individuals interviewed. Depending on whether the question is formulated in abstract or concrete terms, the most valid question will be asked by concretizing the examples.
- behaviours and attitudes: it is always better to focus on behaviours than on attitudes that are more difficult to determine. For example, rather than saying "are you interested in politics", it sounds like "do you read political news in the newspapers?". However, there is no unanimity. Behaviours provide a factual benchmark that provides an easier, valid and comparable answer.
- social desirability of responses: there is systematic overestimation.
- Embarrassing questions: in sociology within the framework of sexological studies in particular, certain questions can be embarrassing.
- absence of opinion and "I don't know": in the context of pseudo-opinions, an individual answers because he was forced to answer something. The interviewee should be given the opportunity to say that he or she does not have an opinion. A large proportion of the sample surveyed may not have an opinion. That is a very problematic point.
- Attitudinal intensity: Often, a measure of the intensity of attitudes is missing in relation to specific issues, so scales are constructed to give a measure of the intensity of the response.
- acquiescence: tendency of respondents to say "yes". Individuals tend to say "yes", depending on how the question is asked, different answers may be given.
- memory effect: the memory is not precise, refers to the question of giving temporal and factual reference points. If possible, answer items should be given to fix the respondent's memory.
- question sequence: the question sequence can also influence the answers. Errors can creep into answering. Consideration should be given to the dynamics of the interviewer-interviewer relationship, the respondent's fatigue that allows for relevance of the response, the interview sequence in terms of knowing or asking the most important questions, the "contamination effect", because the answer to a question can be influenced by an immediate preceding answer. The answer to a question does not depend on the question itself, but on the questions asked previously.
Batteries of questions[edit | edit source]
We ask a question with different answer items:
- save space and time.
- facilitate understanding of the response mechanism.
- improve the validity of the response since there is a single given framework for a cycle of sub-questions.
- allow the construction of synthetic indices.
When we do a standardized questionnaire, we have to spend a lot of time formulating the right questions that will allow us to get the right data.
Methods of administering the questionnaire[edit | edit source]
There are three main ways of conducting a sample survey:
- face-to-face: copresence
- telephone interview
- self-administered questionnaire
- online questionnaire
Face-to-face Interview (CAPI : Computer-Assisted Personnal Interviewing /PAPAI : Paper And Pencil Interviewing)[edit | edit source]
Computer-Assisted Telephone Interviewing[edit | edit source]
Telephone surveys are a method of rapid data collection, which is why they are the most popular method.
Computer-Assisted Web Interviewing[edit | edit source]
Note: the most widespread form is the mail questionnaire. An examination is a self-administered questionnaire with a restitution constraint.
Online Questionnaire[edit | edit source]
This is a modality that is becoming increasingly important, but is criticized because there is a problem of coverage linked to a problem of representativeness of the sample.
The Internet penetration rate is so high in the West, especially in the West, which justifies the fact that the population is sufficiently well covered.
On the other hand, other types of questionnaire administration have as much bias as online questionnaires.
Phases preceding data collection[edit | edit source]
- Exploratory Study
This is a very important step in the quantitative approach and within the framework of a sample survey, if only to properly define the response modalities to be included in the questionnaire.
It is never easy to define the essential questions, which is why it is necessary to conduct qualitative interviews in order to identify people's possible answers, and this is all the more important if the subject has not been explored before.
It is the fact of testing the questionnaire before using it in research. The sample should be similar to the study population.
- Preparation and supervision of interviewers
Preparing interviewers requires good preparation and supervision.
- Initial contact
There is a need to clarify the survey representative, clearly define the objectives of the interview, ensure anonymity, stress the importance of collaboration, but also send out several reminders to ensure that the respondent responds reliably and reliably in order to reduce the rate of non-response.
- Graphic form of the questionnaire
It is necessary to have a careful presentation in order to maintain the concentration and interest of the person responding.
Sampling[edit | edit source]
Sampling is done by sampling rather than studying an entire population as a whole, which makes it possible to make inferences about the entire population.
Definition[edit | edit source]
According to Corbetta, sampling is a procedure by which a set of units that constitute the subject of the study are extracted from a reduced number of cases called "sample" selected with or according to criteria that allow the generalization to the whole population of the results obtained on the sample.
The sample survey is conducted on a population. A number of people are selected from a population that they wish to study. A number of criteria are therefore selected. To do this, it is necessary to be able to infer the results.
Basically, the sampling technique has only advantages. In the sampling, we can extract people who allow us to estimate the difference between what we find in the sample and what we would find in the population.
Main types of samples[edit | edit source]
Sample types are derived from the way the sample is extracted from the population. The method of calculating and estimating the margin of error applies only to probability samples.
Different types of extraction from a "mother population" are different; the great distinction is between probabilistic and non-probabilistic samples.
Probabilistic / simple / random / systematic / stratified probability sample[edit | edit source]
It is a sample where each unit is extracted with a known probability, we know the probability that each individual in the population has to be selected. The probability is different from 0.
This has a major advantage since the margin of error can be calculated. If we infer and generalize the results found in the sample of a population, there is a certain margin of error, with the probability sample we can calculate the margin of error. Because it is difficult to meet all of the criteria, non-likely sampling is sometimes chosen.
- simple random
All population units have the same probability of being included in the sample. In other words, each person has the same probability of falling into the sample. From the complete list, we extract by drawing lots; we base ourselves on the law of large numbers and we can statistically demonstrate that if we extract randomly these people will be representative of the population.
The list is incomplete, we systematically draw all the individuals; for example, on a list of individuals we draw all ten individuals.
The idea is the same as simple random sampling, but we first subdivide the population into different strata and then choose according to what we want to study (example: social classes). With simple samples within each stratum, the margin of error doing this is minimized in comparison to other ways of shooting in general, so the population is more homogenized.
However, one can make sampling errors or have a problem of non-response.
There are sources of coverage, sampling and non-response errors; however, we are lucky to be able to calculate one of the three if we did not perform a non-probability sampling: the parameter that we estimate in the population is the sum of the estimate plus the sampling error . The estimated parameter for the population includes a level of confidence on the basis of which the value can be estimated. For example, if we look at the number of smokers in the Swiss population and we find in a sample of 2000 people 30% of Swiss who smoke, we can say that 30% of the Swiss population is a smoker, but with a certain level of trust.
Going further, we can use the confidence interval, we can say that the 30% found in the sample are representative of the population according to a margin of error if we want to infer it in the population; we cannot say that 30% of the Swiss population, but we can say that between 30 - 33 and 27% smoke; there is an interval in which we will find the true value of the population from the value observed in the population.
There is a notion related to the interval and a notion of confidence in this interval. If the average income of the population of Geneva is 5000.- we can say that in the population it will be between 4500 and 5500 with a probability of 95% that the parameter will be in this interval. From an estimate made in the sample, we will find the value in the population with a certain margin of error and a level of confidence.
Non-probability samples[edit | edit source]
The probability of each individual falling into the sample is not known, so the margin of error cannot be calculated.
- by quota: same logic as for stratification, the difference is that in quotas we do not choose subjects randomly.
- factor drawing
- telephone survey
- convenience survey
Sampling errors[edit | edit source]
Sampling error for an average[edit | edit source]
The sampling error depends mainly on three factors:
- N, Size: The larger the sample, the smaller the margin of error; inversely proportional.
- S, Standard deviation (central trend as a function of heterogeneity): dispersion of the parameter to be estimated; the higher the standard deviation, the greater the margin of error, however, this is not very important if there are many samples. We speak of parameter variability, in other words it is an indicator of the homogeneity or heterogeneity of the sample. What is important is that the margin of error is proportional to the heterogeneity of the sample, but above all it is inversely proportional to the sample size, the larger the sample, the smaller the margin of error.
- Z, Confidence level of the estimate: it is a confidence interval, one cannot know the true value, but it is possible to approach it. In other words, it is the degree of certainty one accepts when making an inference. Thus this represents the confidence interval, it can be shown that for a certain confidence level, for example for 95% the confidence is 1.96, for a confidence threshold of 99%, z=2.58. In the case of a 95% confidence threshold, this means that we have 5% to make a mistake.
- f, applies in the case of a very small population, it is a correction factor between sample size and population, it is the division of the sample size by population size. If the fraction approaches zero, the factor is negligible. More conveniently, it can be said that in any survey the sample is always much smaller than the population, whereas samples rarely exceed 10,000 people, which means that this factor can usually be overlooked. On the other hand, if the difference between the sample and the population is very small, then no sampling is done and everyone is interviewed.
Therefore, the margin of error depends:
- sample size.
- the degree of heterogeneity of the parameter.
- the level of confidence we are willing to concede in statistical inference.
Sampling error for a proportion[edit | edit source]
P and Q are proportions in the parameter sample we want to explain.
Basically, in order to calculate the margin of error, particular account must be taken of the sample size. This is something that may seem surprising, since the size of the "mother population" does not appear in the calculation of the sampling error formula. In other words, a sample of 1000 persons out of a population of 50,000 gives rise to the same margin of error as a sample of 1,000 persons out of a population of 350 million inhabitants. The sample size gives the same margin of error if a sample is taken for the Whitewater or Australian population.
Example[edit | edit source]
There are two samples, one of 100 and one of 1000 and we want to estimate the average income. Let's say everyone answers; we'll try to calculate the sampling error.
In the first sample:
- n: 1000
- x: 1253000
- S: 311000
- e: 1.96% for 5% (standard threshold in social sciences)
18700 is the calculated sampling error for a sample of 1000 people drawn from a population of 10,000. It is a margin of error for an average estimator with a variability that is calculable and we will find a margin of error of 18700.
We interpret 18700 as "more or less" () if in the sample we have found that the average income is 1253000, in the population we can be sure to be wrong at 5% if we say that in the population the average income is 1253000 18700, that is to say
The more secure we want in the inference, the greater the interval becomes.
In the second sample:
- n: 100
- x: 1250000
- S: 308000
- e: 1.96% for 5% (standard threshold in social sciences)
Note: when the proportions are different, the margin of error is smaller.
The following table shows the sample size for a simple random sample that is necessary for some measure accuracy. The narrower the confidence interval, the more accurate the measurement, but the confidence level must be considered again.
If we want an estimator accuracy of 5%, for a population of 1000 people we need a population of 285 people, etc..
Among other things, we see that for a population of 50,000 people or more, a population of 2,500 people is needed in order to obtain an accurate estimate.
When you have the opportunity to make a probability sample, you are lucky enough to be able to calculate the margin of error and estimate what that error is, something you can't do with the coverage error.
However, there is a problem. Sometimes or even often it is not possible to have a probability sample. Simple random sampling is often not applicable because the complete list of individuals is not available.
In these cases it is not possible to carry out these types of calculations, so other types of samples are carried out:
- systematic sample
A person is taken at regular intervals and it can be demonstrated that it is close to a probability sample or a random sample. One example is the "exits polls", we meet the individuals who come out of the ballot box by interviewing all 20 people. A systematic sample has the same property as a probability sample because each person has the same probability of entering the sample.
- stratified sample
The population has been subdivided previously for a specific reason because we want to reduce the heterogeneity of the population, the more heterogeneous the heterogeneity, the more the margin of error is reduced. In the stratified sample, we take a variable of interest to us whose proportions in the population are known and draw random samples within each category according to these variables. We have a lower standard deviation and a smaller margin of error, combined with a sample, shows that the overall margin of error is lower than when individuals are drawn indiscriminately.
- nonprobability samples
There are non-response errors and coverage errors in the population. More and more survey researchers are saying that other sources of error are also important. It's not really necessary to do these calculations, but rather it's a matter of moving towards unlikely samples.
According to Corbetta, a non-probability sample is a sample that is not based on the number of individuals; the best known is the quota sample: within each stratum, a certain number of people are chosen to make up the sample. If we know that in the population we want to study there are 60% women and 40% men, we will make sure that there is the same proportion in the sample. We will choose according to the criteria we want, but respecting the quota is the proportion in the sample.
This sample is becoming more and more frequent because it is cheaper and less laborious, but has the disadvantage of not being able to calculate the margin of error.
Qualitative techniques[edit | edit source]
We are within the framework of the interpretative paradigm:
- there is no clear conceptual and terminological distinction.
- there is no clear distinction in terms of application, we combine.
- the research path does not follow separate and distinct phases.
Types of Qualitative Research[edit | edit source]
These are the three main activities that can be done as part of the research. From each of them results a method that allows us to interpret a social reality.
Participant Observation[edit | edit source]
It is a much less widely used method in political science, but it is an original method used, in other words, it is a special case of direct observation; it is an observation with the participation of the investigator. From an epistemological point of view, it corresponds well to the interpretative paradigm whereby the researcher is part of his or her own investigation.
Definition[edit | edit source]
According to Corbetta, participant observation is understanding through participation from within the phenomenon studied: It is a research strategy in which the researcher is directly and for a relatively long period of time inserted into a given social group taken in its natural environment by establishing a relationship of personal interactions with the members of this group with the aim of describing the actions and understanding them through an identification process in order to understand their motivations.
- researcher observation: / /
- long enough: it is not enough to observe only once.
- natural habitat of the group: it is necessary to observe the group in its natural habitat.
- interaction between researcher and object of study: this characterizes the methods of the interpretative paradigm, which is why we participate in the studied phenomenon.
- The aim of describing and understanding: one tries to put oneself in the other's place, it is a form of empathy.
In the participant observation, one looks for a researcher's involvement, the final goal is to describe and understand.
A well-known example is a book written by Goffman who had studied psychiatric asylums in Ecosses while he was a nurse. From the participant observations, he concluded that psychiatric asylums are total and totalizing institutions.
Objects of the participant observation[edit | edit source]
- physical context: group context, observed individuals.
- social context: phenomenon and dynamics observed.
- Formal interactions: is the main purpose of this method.
- informal interactions: /
- interpretations of social actors: we observe the interpretations that the actors themselves give of their participation.
When should observations be recorded when participating observations are made?[edit | edit source]
The recording of observations must be done as close to the environment as possible, the aim of the participation is to be in the event being studied, we observe as close and concrete as possible.
It is very important to have a diary in which you write down the different observations you make:
- we describe the "fact" we observe.
- we note the interpretation he's looking for.
- note the interpretation that the subjects themselves make, the objective is to confront interpretations.
- it is necessary to apply a principle of fidelity by a precise and filtered recording of one's own subjectivity.
- the facts must be distinguished from the interpretation.
A participant observation was made on social movements and in particular the alterglobalization movement; it was a comparative research that looked at how controversies arise when a group has to discuss certain issues.
The purpose of the study is to examine whether what organizations are putting in place, i. e. a participatory process that is based on consensus rather than a majority vote or autonomous decisions, in order to account for representative democracy.
Traditionally the journal is used in sociology and anthropology is made up of personal notes.
In the case of these organizations, observations were systematized by creating a protocol based on variables and each variable is supposed to measure a specific aspect of these debates when there were issues ranging from organizing the next event to the strategic and ideological direction to be taken.
The researcher had to gather basic information such as the duration of interactions, controversial discussions, but also the number of people who were present and how many participated, and the degree of reciprocity of the discussion to know when participants refer to what others said when they made proposals, there were also hard power aspects, i. e. the use of one's position to influence the discussion, etc. The researcher was also asked to collect basic information on how many people were present and how many participated.
In conclusion, we can very well study qualitative matters quantitatively. The most difficult part is not collecting information, but analyzing it.
Qualitative interview[edit | edit source]
This method is much more used than participant observation; the objective is to try to gain access to the perspective of the subject under study.
Definition[edit | edit source]
According to Corbetta, this is a conversation provoked by the interviewer, appreciated by subjects chosen on the basis of a plan of observations that are a consistent number whose purpose is knowledge guided by the observer.
- interviewer-induced conversation: one person conducts the interview.
- subjects chosen according to a systematic observation plan: individuals are questioned on the basis of how to conduct the research done beforehand.
- quite a large number of objects: the question is to what extent one needs to do X interviews rather than Y.
- knowledge objective: the very objective of science.
- interviewer-led conversation: the interviewer guides the interaction.
- flexible and non-standardised interrogation scheme: allows you to modulate the interview.
The qualitative interview differs from a qualitative interview in that:
- lack of standardization.
- In the context of discovery rather than justification, which is intended to justify the hypothesis.
- absence of a representative sample: we want to distinguish between them.
- we are in a subject centred approach rather than variables.
An additional subdivision can be made, because there are different ways of conducting a qualitative interview:
- structured interview
The same questions are asked of all interviewees in the same wording and in the same sequence. The interviewees have the complete freedom of answer, it is an interview done by open questions, there is a concern of comparability behind this type of investigation.
This is a type of interview that is not very widespread, because we can standardize a minimum, but not enough.
- semi-structured interview
The interviewer only has a trace of what he or she wants to touch and the arguments he or she wants to address, but these arguments can be posed in different order, not in the context of the interaction. This trace can vary in decision and detail, the idea is that we don't ask everyone the same question, but we adjust the arguments.
- unstructured interview
Neither the form nor the content of the questions is pre-established, this form and this content can vary from one subject to another so there is an individuality of arguments and a variable itinerary of the interview, we only have one itinerary, everything else is open.
- Life Stories Method
Narration of the life of the individual, it is an extreme case.
- non-directive interview
The purpose is different from that of scientific research; there may be therapeutic or clinical interviews.
- interviews by privileged observers / key informers
Interviews are conducted, but the interviewees are not the subject of the research, they are interviewed because they are well informed about a certain phenomenon.
- group / collective interview
We interview several people at the same time, they are "focus groups" i. e. focused groups, i. e. there is an aspect related to the interaction between the interviewees that becomes crucial.
Conduct of the interview: Suggestions for "good practice"[edit | edit source]
- Preliminary explanations: explain why we are here, what is the purpose of the research, put the individual in context.
- primary questions: these are questions that introduce new topics into an interview, usually these are the questions that are put in his interview guide.
- secondary questions: allow to deepen or articulate the primary question.
- questions of stimulus: all the ability of the investigator is to master them.
- repetition of the question: repeat the question.
- repetition of the answer or a synthesis of the last answers: repeat the answer given by the interviewee or do some sort of synthesis of what the respondent said.
- encouragement, expression of interest: we must see the person remembering something.
- pause: time delay.
- request for further information: obtain additional or more precise details.
- language: care must be taken to ensure that one's language is correct, and care must be taken not to misuse or abuse sectoral or subculture-specific language.
- interviewer's role: the role is crucial when it involves a qualitative interview, much more so than in the case of a sample survey, where you can pay someone who carries out the interview. It is difficult to send someone else to conduct the interview because they were not involved in the research development process.
The strength and weakness of this method is the absence of standardization, which makes it possible to go into certain subjects in greater depth, but at the same time it makes it more difficult to compare and compare the responses of the interviewees.
With regard to the analysis of data based on qualitative interviews, it should be noted that:
- Data analysis is focused primarily on cases and not on variables, with exceptions through standardized protocols for participating observations.
- The presentation of the results is based on a narrative perspective, episodes are told, cases are described, often through the words of interviewees, extracts can be taken and used for analysis.
- we proceed through classifications, typologies as part of a technical approach, we aim to bring out profiles.
Reading - Documents[edit | edit source]
This is the kind of affinity that one makes when one enters into a qualitative approach.
Definition[edit | edit source]
According to Corbetta, a document is an informative material on a given social phenomenon that exists independently of the researcher's action.
- informative material.
- on a given social phenomenon.
- exists independently of the researcher's action.
Types of documents[edit | edit source]
A distinction is made between primary data constructed by the researcher from secondary data that are the existing data.
- Personal documents:
- oral testimonies
- Institutional documents: in political science, it is mainly these documents that interest us.
- mass media: a privileged source of information for political scientists.
- narration, pedagogical texts, narratives of popular culture: rather sociological approach.
- court material
- policy documents
- business and administrative documents
- physical traces
Annexes[edit | edit source]
- Corbetta, P. (2003). Social research: Theory, methods and techniques. Sage. Url:https://books.google.fr/books?hl=fr&lr=&id=n6jEtRcRaQcC&oi=fnd&pg=PP2&dq=Piergiorgio+Corbetta&ots=aGr_fEIsUK&sig=6CQE7FU89BEG-U6rnjagSuBQidY#v=onepage&q=Piergiorgio%20Corbetta&f=false
- « Mesurer » la gouvernance, la démocratie et la participation citoyenne à partir des enquêtes auprès des ménages de Governance Asssessment Portal url:http://fr.slideshare.net/GAPortal/mesurer-la-gouvernance-la-dmocratie-et-la-participation-citoyenne-partir-des-enqutes-auprs-des-mnages
- Corbetta, P., 1999. Metodologia e tecniche della ricerca sociale.