Download Social Research Methods: An Overview and more Exams Social Work in PDF only on Docsity!
Social Research Methods
Social Research - ✅Process of using empirical observations (DATA) to answer questions about the social world. Research subjects are people or groups of people Focused on patterns or regularities Conclusions are about aggregates, not specific individuals Empiricism - ✅A broad epistemology that says we gain knowledge by observing the world around us (i.e., gathering data). Epistemology - ✅Refers to how we come to know something how we justify that knowledge to others. Central to the philosophy of science. Pure Research - ✅Social scientific research that tries only to increase our knowledge of the world. Primary audience = academic types. Applied Research - ✅Research that tries to identify a social problem (with an eye to fixing it) or offer to solutions to an already-identified problem. Primary audience = practitioners. Mistaken Observation - ✅See something that isn't there Careful collection of data and measurement (mistaken/selective observation) Selective Observation - ✅Only seeing what we want to see/"cherry-picking" Overgeneralization - ✅What's true for a few is true for all Systematic sampling (overgeneralization) Illogical reasoning/faulty assumptions - ✅Ex. gambler's fallacy
Explicit criteria for determining cause and effect (illogical reasoning) Resistance to change - ✅Commitment to our "priors" Maximizing evidence for an minimizing evidence against over favored positions Methodology - ✅The study of "finding things out." Can be viewed as either a branch of epistemology, or a related field. Two main ways to distinguish between the various methods of social research. - ✅Focus on the strategy that the researcher uses. Focus on the kind of data involved. Obtrusive Research - ✅The researcher directly elicits data from the people she wants to learn about. These people (research "subjects") are typically aware that they are being studied. This is the bread and butter of social science. Ex: Survey Research, Experiments, Fieldwork (usually) Unobtrusive Research - ✅The researcher avoids directly interacting with the research subejcts. Instead, she looks for data sources that subjects create in the normal course of their lives, or for data that were obtrusively collected by someone else for a different purpose. Ex: Trace evidence, Content analysis, Secondary analysis Quantitative Research - ✅Data are numeric Examples:
- Most survey data
- Most experiments
- Census records
- Election returns
- Financial records Quantitative research generally accepts that there is an objective truth to identify. Qualitative Research - ✅Data are non-numeric (i.e., words) Examples:
- Interviews
- Most fieldwork (e.g., participant
Ex: Marxist theory, conflict theory Middle-Range Theory - ✅Theory that deals with individual aspects of social behavior, but not individual people/groups. Bread and butter of social research Not the same as microtheory Ex: risk-taking behaviors, social networks, learning, empathy Paradigm - ✅"a model or frame of reference through which to observe the world" (Babbie) An agreed-upon set of assumptions (a "tradition") that guides scientific inquiry Paradigms represent our prior beliefs, or first principles, about the world Theories are built on a paradigmatic foundation If we agree about these principles, scientific communities develop theories faster Paradigms related to the inquiry - ✅Deals with how scientific research should be done Suggest what kinds of questions are answerable and how to answer them Sometimes called "inquiry paradigms"
- Ontology 2) Epistemology 3) Methodology Paradigms related to the explanation - ✅Deals with fundamental explanations about how the world works (i.e., which social forces matter?) Grand theories typically have associated paradigms Ex: "We should understand the social world as a battle for dominance over others" (conflict theory) Occasionally called "substantive paradigms" Ontology - ✅What is the nature of reality? Assumption Epistemology - ✅How do we generate knowledge about that reality? Assumption
Methodology - ✅What tools should we use to generate knowledge? Implication Positivism - ✅Ontology- The world has objective, generalizable regularities (rules) Epistemology- Objective empirical observation allows us to identify those rules Methodology- Usually quantitative Interpretivism aka Antipositivism - ✅Ontology- Reality consists of subjective meaning not objective rules Epistemology- We should try to understand how people interpret their surroundings Methodology- Exclusively qualitative Researchers can never be completely objective because their subjective biases impact:
- How they ask questions
- How they define and interpret data
- How they draw conclusions - ✅Because of the above, we can't be certain in our conclusions. Conclusions about the truth are probabilistic at best Post-positivism - ✅Ontology- The world has objective, generalizable regularities (rules) Epistemology- Empirical observation allows us to learn about those rules but we will never have complete, unbiased information Methodology- Quantitative and qualitative Post-positivism asks researchers to be honest about their own biases Most quantitative and a large amount of qualitative social research fits this inquiry paradigm Substantive paradigms can imply: - ✅A set of questions to be asked about a given topic A general theoretical approach explaining the nature of the topic (e.g., "class struggle") A set of smaller theories with a common theme Conflict Theories/ Conflict Paradigm - ✅Power is the driving force in human societies and human behavior can be understood as an attempt to dominate or to avoid being dominated by others. Symbolic Interactionism - ✅Human behavior involves assignation of symbolic meaning to objects, behaviors, and other people. These symbols color future behavior. Rational choice - ✅People decide on various behaviors by weighing the relative costs and benfits of each behavioral choice and choosing the one that is best for them.
Logic - ✅Scientific expectations (potential explanations) are deduced from general theoretical premises and then tested with data Motivation - ✅Check a specific explanation for the phenomenon Generate potential explanations for the phenomenon Verification: The Deductive Research Pattern - ✅1 Research question about a topic 2 Theory related to the topic 3 Hypotheses about what we will observe 4 Observation: Collect and analyze the data 5 Confirmation: Is the data consistent with the theory-generated hypotheses? Mode - ✅Begin with the data, describe it, and develop a theory to account for it. Generation: The Inductive Research Pattern - ✅(General) research question about a topic
- Ex. How do homeless adults understand homelessness? Observe patterns in the data Tentative hypotheses about what's going Substantive theory about the topic Concepts - ✅Abstract ideas about a phenomenon. "Mental images we use to bring order" to social phenomena Latent Jargon Examples: Sociology: power, wealth, class, gender Psychology: intelligence, perception, Political Science: participation, representation Economics: supply, demand Propositions - ✅Conclusions about the relationships between concepts that are derived from the theory. Poor youths are more likely to break the law to gain material comforts than rich youths.
Judges who have daughters are more likely to be sympathetic to so-called women's issues. Concepts in the examples: poverty/wealth, obedience, gender, empathy Variables - ✅Literally, a characteristic that can vary. Measureable concepts. Independent variable: causes/relates to variation in something else; a predictor Dependent variable: something that depends on the influence of another variable; an outcome Hypotheses - ✅Expectations/beliefs about the relationship between variables. Testable propositions. Null hypothesis: Variables are unrelated Alternative hypothesis: A particular relationship exists. Responsibility to science - ✅Academic integrity Admit errors when made Pursuit of knowledge should be more important than personal gain or promotion of a personal philosophy Responsibility to society - ✅Social research should benefit society (in the form of new knowledge) while minimizing harm Responsibility to subjects - ✅The single biggest area of ethical concern deals with how we treat human subjects. Belmont Report: In the 1970s, there was a US federal commission to examine abuses in research.
- Respect for Persons, a.k.a. Autonomy
- Beneficence, a.k.a. "do no harm"
- Justice: Most relevant to receipt of placebos in medical studies Autonomy - ✅Voluntary Informed Consent Voluntary: Subjects are not directly or indirectly coerced into participating. I.e., they feel as though they can say "no." Informed: Subjects understand what they're getting into, including any risks. Consent: Subjects actually agree to participate Beneficence -
Reviews reserch proposals and ensures:
- Minimization of risk of harm to research subjects
- Risk of harm is proportionate to the importance of the project
- Researcher's procedures will produce informed consent
- Researchers comply with appropriate ethical standards (such as the special rules for children) Limitations of the IRB - ✅Ethical issues are complicated; different IRBs might come to different decisions Example: Montana Campaign Mailer Experiment IRBs spends an inordinate amount of time reviewing harmless research Can focus on minutiae while missing big things IRB review imposes special burdens on some kinds of qualitative research Ethnography is hard to predict beforehand Reports of IRBs being biased against qualitative work Assessing potential harm - ✅Method: Am I interacting with participants? Topic: Am I touching on private matters? Participants: Am I in a position of power? Dissemination: Will I be sharing this information? Pure deception - ✅Providing false information (actively lying) to research subjects about the nature of their participation in the study. Incomplete disclosure - ✅Withholding information about the real purpose of the research. Debriefing - ✅Efforts by researchers to provide information and correct subjects' misconceptions about the project and their participation. Debriefing should be immediate if there is any sort of risk of distress (e.g., Milgram) Economics and deception - ✅Most econ journals will not publish deceptive research. Agree that deception should be avoided; disagree about whether its ethical Primary argument is methodological Typically use pools of subjects in experimental labs Concerned about getting bad data
Research question - ✅A question about the nature of the relationship between concepts that can be answered directly or indirectly using data. Ideographic explanation - ✅Tries to identify all of the explanations for a phenomenon in a small number of cases. Deterministic - once you understand the situation so well, the answer is obvious. More typical in qualitative fieldwork Nomothetic explanation - ✅Tries to find a the most important explanations for a broad range of cases. Probabilistic - our explanations are only "likely" More typical in quantitative research Characteristics of "Good" Research Questions - ✅1) Specific and unambiguous
- Answerable with data (i.e., empirical not normative)
- Nomothetic: Suggest a relationship between concepts
- Nomothetic: Possible relationship between the concepts is meaningful (i.e., causal or correlated for a reason)
- The answer matters (i.e., interesting) Stages of a research project - ✅1 Formulate the research question. 2 Set the framework of the research design 3 Develop a measurement strategy 4 Select the sample 5 Collect the data 6 Process and analyze the data 7 Present the results Case - ✅A single instance of a phenomenon that we might want to explain or describe. Examples: A person deciding whether to vote; a couple deciding whether to have children; a person acquiring a second language; a jury deliberating over a criminal conviction. Unit - ✅Refers to a level of social life that identifies a single "point" for study. A case may work at this level, such that each case is one point, or a single case may have many points. Unit of analysis -
Actions - ✅Behaviors or interaction Unit = Person: voting, drinking Unit = Group: corporate philanthropy, passage of laws Why do aspects matter? - ✅Different designs are better for different focuses. Surveys are good at gauging characteristics and orientations, but worse at gauging actions. Experiments are good at gauging actions, worse for orientations and awful for characteristics. Cross- sectional designs - ✅Look at many units at a single point in time. Appropriate for description and for explanation during a single time period. Useful for establishing correlations, but causality is more difficult. Longitudinal designs - ✅Look at multiple points in time. Sometimes involve a smaller number of units. Trend/time-series studies Cohort studies Panel studies Retrospective study Pooled designs - ✅Look at multiple points in time, but treat them as if they were simultaneous. A "fudge" used in some studies here a process is believed to be stable over time. Can be used to increase the number of data points to support certain statistical methods. Trend/time-series studies - ✅Look at pattern of change in a variable over time. If explanatory, the IV is time. Cohort studies -
✅Cohort = set of people who experience the same event at a given time, such as birth year or graduating class. Compare cross-sectional studies at different times. Individual subjects are not the same. Panel studies - ✅Reinterview same group of people (the panel) at multiple points in time. Great for causality since the future can't cause the past Extremely expensive and time-consuming Attrition — people leaving the study — poses analysis problems Retrospective study - ✅"Fakes" a panel design by asking people questions about the past. Inferior because people (a) forget and (b) lie. Correlation - ✅An empirical association or correspondence between two variables. Does not imply that one causes the other... But you can't have cause and effect without correlation! Three characteristics to watch: Direction, Form, Strength X and Y must be related Positive correlation - ✅E.g., as X does something (e.g., gets bigger), Y tends to do the same thing. Negative correlation - ✅E.g., as X does something (e.g., gets bigger), Y tends to do the opposite thing (e.g., gets smaller). Zero (Null) correlation - ✅E.g., as X does something, Y does nothing. Linear correlation - ✅Same relationship at all values of X Logarithmic correlation - ✅Relationship levels off eventually. Quadratic correlation - ✅Relationship changes direction at a certain point.
Necessary Cause - ✅X must happen in order for Y to happen. Sometimes findable in nomothetic work. Repressive or ineffective government seems to be a necessary cause of revolution. Sufficient cause - ✅If X happens then Y will happen, but Y might happen without X. Ideographic study of one case can find it. Necessary and Sufficient - ✅X is the only cause of Y, and it causes Y every time. Never see this with nomothetic relationships (because we don't go deep enough into any individual case) Hypotheses - ✅Expectations/beliefs about the relationship between variables. Testable propositions Example: Adults with higher levels of education will earn greater annual incomes than adults with lower levels of education, ceteris paribus. Null Hypothesis (H0) - ✅Research design should attempt to falsify the "null hypothesis" X is unrelated to Y Alternative Hypothesis (Ha) - ✅X has some specified relationship with Y. Falsifiability - ✅We can imagine a set of circumstances that we might observe under which H0 is false. Why focus on whether H0 is false? - ✅Motivational reason: Forces us to ask answerable questions. Logical reason: If H0 is false, it is not necessarily true that HA is true. Analytic reason: We don't have tests that can tell us how likely we are to observe HA due to random chance! Quantitative Inferences About Hypotheses - ✅1 If H0 is true, we should observe a null correlation. 2 If the correlation is not null, how likely is this just random chance?
3 If that likelihood is sufficiently low (e.g., < 5%), then we say H0 is probably false. Stages of a Research Project - ✅1 Formulate the research question 2 Set the framework of the research design 3 Develop a measurement strategy 4 Select the sample 5 Collect the data 6 Process and analyze the data 7 Present the results Measuring social concepts - ✅No common unit of measurement Lack agreed-upon standards against which to measure social phenomena. Have to justify that our measures are valid Exception: Economics has money Latent concepts - ✅Not something we can observe directly (e.g., personality, legitimacy of institutions) Research questions involve statements about abstract concepts. - ✅These are NOT empirical How do we make them empirical? Conceptualization: clarify the concepts Operationalization: specify how they are measured Conceptualization - ✅Clarify the concepts Generally, we rely on prior work to compose nominal definitions. Selecting a nominal definition for our concepts Operationalization - ✅Specify how they are measured Specifying the exact operations involved in measuring a variable The manner we have chosen to measure the concept Dimensions of concepts - ✅Latent but suggestive
Operational hypothesis: Testable proposition to the research question stated in terms of indicators! Levels of Measurement - ✅Refers to how we assign numerical values to the attributes of a variable.
- Nominal 2) Ordinal 3) Interval 4) Ratio The level of measurement of the dependent variable determines which statistical analyses we can perform.. Nominal Measures - ✅Attributes differ in quality but not in amount. Values have neither cardinal nor relative meaning.
- It doesn't matter what the values are, just that they're different from one another. Used for things like race, gender, religion, political party Dichotomous measures—measures that look for the presence or absence of one trait—are often called "dummy" variables due to their simplicity. Two Requirements for Measures - ✅Measures must be exhaustive and mutually exclusive
- Required for all measures, but comes up most often for nominal. Every observation must fit into one, and only one category. Example: Mixed-race individuals and the race variable Option 1: Mixed-race category Option 2: Separate yes/no variable for each racial group Ordinal Measures - ✅Attributes can be said to denote an amount/order. Values have relative but not cardinal meaning.
- It matters that one is bigger than the other, but you can't do mathematical operations (e.g. addition) on them.
- Distance between two values is not necessarily the same. Interval Measures - ✅The logical distance between attributes is meaningful. Values have both relative and cardinal meaning, but there is no meaningful 0 point.
- Thus, we can do addition/subtraction
- But not multiplication/division Very rare in social science. Example: Fahrenheit Temperature 90 F is 30 degrees warmer than 60 F.
It's not 3 times as hot. 0 F is not meaningful because it doesn't indicate a lack of temperature. Ratio Measures - ✅Most precise level of measurment Values have cardinal and relative meaning There is a fixed, absolute zero point
- Can be multiplied and divided Examples: Age in years Income in dollars Hours worked per week Kelvin temperature Reliability - ✅Will the measure yield provide consistent answers? Validity - ✅Is the measure trustworthy? Does it actually measure what it professes to measure? Stable Reliability - ✅If we take the same measurement from the same subject at different points in time, will the score be the same? This is often called test-retest reliability Consistency Reliability - ✅If we measure a concept in multiple ways, do the answers point to the same conclusion? Inter-item reliability: Agreement of multiple indicators for the same concept/dimension. Inter-coder reliability: Agreement of multiple observers looking at the same thing and using the same definitions. Face Validity - ✅Does the measure seem like a reasonable way to operationalize the concept on its face? Content Validity - ✅Does the measure cover all of the relevant aspects of the concept/dimension it claims to measure? Construct Validity - ✅Does the measure behave in the theoretically expected way? A measure of self-esteem should be positively correlated with a measure of confidence.
