SUSS SSC 109 Intro to Social Science Study Unit 3
SUSS SSC 109 Intro to Social Science Study Unit 3
SOCIAL SCIENCES RESEARCH TERMS: Definitions and applications
A hypothesis is a tentative (i.e., untested) statement describing the relationship between facts and events.
typically between two or more
e.g.: Consider the hypothesis: ‘Absenteeism rate of primary school students depends on the distance between the child’s home and the school.’
The independent variable is the ‘distance between the child’s home and the school.’
The dependent variable is the ‘absenteeism rate of primary school students.
A variable is something which is thought to influence (or be influenced by) something else.
It is a conception that contains within it the idea of change (i.e. it is not static).
(IV) are variables that influence another variable.
i.e. Colour variant in a stroop task
(DV) are variables that are affected or influenced by the
i.e.Duration to identify colour in a Stroop task.
Reliability is an
assessment of the accuracy
of the measurements of the variables.
Reliability depends on how
is the data collected and how
are the experiment’s outcomes.
Consider the research question “What is the primary reason why Singaporeans below age 50 choose to marry?”
The data collection methodology must consider the following points:
To be reliable, the sample group must be Singaporean, below 50 years old, adequately large in size, and reflects Singapore’s demographic makeup.
If a survey/interviews were used, the same set of questions must be put to all responders. No leading or loaded questions are to be used.
Data needs to be suitably granular to identify differences in the primary reason between men and women, people of different age groups, different educational, income, occupational backgrounds, etc
Validity refers to the degree to which a measure reflects the reality of the subject(s) tested.
Questions on validity often occur when doubt emerges on the research’s underlying assumptions
consider the following statement: “IQ tests are often used to measure a person’s intellect or intellectual potential.” Questions on the validity of this measure include:
Does the IQ test measure ‘potential’ or just reflect the intellectual skills acquired by the person at that point in time?
Does it over-emphasise a particular set of intellectual skills – i.e., logic or spatial skills – at the expense of other skills such as empathy and observation skills.
“Deduction has to do with using the logic of a
theory to generate propositions
that can then be tested”
A deductive approach begins with an expected pattern that is tested against observations.
“Induction refers to building
the accumulation and
summation of a variety of inquiries
An inductive approach begins with observations and seeks to find a pattern within them.
A theory is a
that has demonstrated real explanatory power but cannot be proven to be a fact or natural law. The successful outcomes of all social science findings, which are not related to biology, are theories.
Social theories are thus accurate generalisations of human behaviour.
Some theories have very broad applications, and are often contingent on the place, culture, and time they were developed.
A model is a simplified framework for describing or explaining observable and complex behaviour or phenomena.
Models are composed of and
reflect interrelated groups of theories.
FRAMING A BASIC SOCIAL SCIENCE RESEARCH STUDY: An indepth look into the scitific method
1. Identify topic and variables
1a. Identify Topic
/ social phenomena for study
i.e. falling birth rates, cost of living, ageing population, crime, etc
Identify relevant variables
for the study and prioritize them.
Identify the independent and dependent variables.
For example, in the research into the cause of low fertility rate in Singapore, you may want to consider the cost of living, lack of time, work stress, career expectations, late marriages, better-educated women, changing cultural and lifestyle expectations, etc, as relevant variables to study.
However, a researcher must necessarily prioritize and identify the most important independent and dependent variables.
Variables: The different dimensions
Examining the different dimensions gives us a more in-depth understanding and
new or different measurement
possibilities for extracting more information from the variable.
Something which is latent is
present but hidden
. It may develop and become
more significant over time
or in the future.
Latency is often identified through analysing macro-scale quantitative research data and confirmed through qualitative field research (vice versa is also possible, but relatively less frequent).
most demographic changes are latent because it can take a long time to develop
. Consider an
. If most young, married Singaporean adults today have less than two children, then in 30 to 40 years’ time, Singapore will have an ageing population. This future probability is only partly visible today; hence, it is latent.
how important an opinion is in relation to other opinions a person holds
. Salience is often
used to test the intensity and stability of an opinion
. It is very useful in opinion surveys to test for and
politically correct (or
Salience questions are
typically tough questions
that make respondents think through their opinions more carefully.
appear frequently where there is a controversial opinion.
For example, given a limited government budget and pressing social need, should more be spent on education, housing, or public healthcare? [i.e., to develop future generations, provide homes for young adults, encourage people to marry and procreate, or to care for the older generation of people who helped build the country]. If hungry people steal food, should they be punished severely? (Underlying question here is: compassion or social justice versus social order or sacredness of the law – which is more salient or important?)
how strongly held an opinion is
. Intensity is often reflected on scales. Note that
location and intensity
of opinion are
Intensity concerns itself more closely with how weighted opinion is towards the two ends (and centre) of a scale.
An example of a
highly intense opinion
: most Americans are convinced that democracy is the best political system in the world and will defend it when attacked.
An example of a
low or no intensity opinion
: although the workers in this factory are made to recite their company’s “Mission Statement” every morning, they do not care much about it.
Refers to how
changeable or volatile is an opinion
Note that strongly held (i.e., intense) opinions can be highly changeable/unstable. Opinions held with moderate intensity can be very stable.
An example of an
: after committing his life to studying law, Jack became a bartender, and then a ballet dancer; today, he sells home-made perfumes.
An example of a
: despite many disappointments and defeats, the soldiers remained devoted to defending their country
an opinion is located
on a scale of ‘for or against’
For example, a scale showing: Strongly Agree – Agree – Neutral – Against – Strongly Against. If most of the opinions fall within ‘Agree’, then ‘Agree’ is the location of the opinion.
Refers to the “for-ness or against-ness” of opinion for any given question. How many or
what proportion of people agree or disagree
with the given issue/question/statement? The side on which more people put their opinion is its direction.
For example, in a survey of 1,000, 550 voted ‘for’; 450 voted ‘against’: the direction of opinion was thus 55% towards approval.
Test the hypothesis
operationalising its variables
and measuring changes
in the variables.
3a. Operationalise Variables
To operationalise variables is to convert them into a form that is
measurable and testable
This is done by
formulating questions with measurable answers
that accurately reflect the variable being tested.
Operationalisation also allows measurements of the variable to be replicated by others.
Vague and/or complex variables should be
The operationalised variable or the questions posed, must relate to the original
variable as closely as possible to ensure its validity
(Dont ask irrelevant questions).
3b. Test hypothesis
the best research
to test the hypothesis.
This selection is usually a combination of quantitative and qualitative research methods.
combination of methods
helps in testing the consistency of the results
If the results are inconsistent, it is likely that an alternative variable is present .
must be collected in a way that is
reliable and consistent
the measured relationship between the variables by comparing with
the original hypothesis
4a. Evaluate the test results by comparing findings against original hypothesis
If the test results are valid and survive a third-party independent testing, the hypothesis can be upgraded to a generalisation.
If the test results do not support the hypothesis, a null result is given.
A generalisation is a hypothesis that has been scientifically tested and shown to be valid.
4b. Alternative Variables
Alternative variables are variables that provide the real explanation for the relationship between two or more separate variables which have been tested to be correlated.
For example, in the US, there is a strong correlation between high educational attainment and voting for the Republican party. However, the real reason was that many highly-educated people come from wealthy families who vote Republican (this is a more pro-wealthy or pro-business party).
The alternative variable here is thus ‘wealth’, which is the more accurate variable influencing why most highly educated US citizens tend to vote for the Republican party.
Develop a hypothesis
explaining the relationship between the variables
Avoid forming hypotheses that you cannot test given your resources
For example, if you have two weeks and a small research budget, it is better to choose a narrower topic and form a hypothesis with a narrower scope
Exercise empirical objectivity
Is the hypothesis testable
, so that evidence of the relations contained in it can be observed, demonstrated, and falsified?
For example, “The human body is sustained by a mysterious substance called ether.” Hypotheses like this cannot be empirically tested and/or falsified at present. This is the case even if the statement is true.
in the hypothesis clearly
specified and measurable
by a technique you know?
For example, How do you measure ‘political participation’, ‘social cohesion’, etc.? Can these variables be measured using a survey questionnaire or field research methods? Is secondary source or archival data available?.
Which hypotheses are crucial
to the solution of the whole problem?
There are many possible hypotheses that could be used to answer a research question posed in the social sciences.
Social scientists need to
isolate the most likely and fundamental causes and start experimenting from these
Propose theoretical significance
Relate new findings to other existing generalisations or theories.
Can new theories be formulated from the new information? Consider if the inquiry suggested any other possibilities or hypotheses.
QUALITATIVE & QUANTITATIVE
To quantify is to make it measurable
, and establish a measurable unit or category as a basis of measurement (this is what meant by ’establishing a standard amount‘).
For example, money (the dollar symbol $), Temperature (degree Celsius °C), distance (centimetre, metre, kilometre, etc.), time (hour, day, month, year) the voter, the patient, the foreign worker, the civil servant, etc.
A quantitative approach relies heavily on gathering and analysing statistical (numerical) data.
Two types of quantification.
is used to count the “presence or absence” (i.e. the quantity) of a measurable unit.“How many?” is the key question in discrete quantification.
An example of discrete quantification is “This class has 48 students in it.” ‘Student’ is a measurable unit and there are 48 of them in this class.
can take any value. It is used in practice mainly to
capture the range of values in each category
For example, in a survey questionnaire. Imagine we are doing a survey for a clinic and everyone should fill in their height, weight, and age. Someone’s age might be 9, another person’s 62. The collected survey would provide information on the range of ages of the clinic’s patients, and the proportion of patients in an age group.
Critical look at this approach
Quantitative approaches rely on empirical evidence and are scientific. However, it has
limited interpretive power
and is also vulnerable to
like non-scientific qualitative approaches.
Data on its own provides only factual statements.To analyse and interpret data is to introduce a qualitative element to it.
Data is not objective. It must be selected and interpreted
by people who are subjective.
How data is collected, and which data is emphasised is also subject to bias
Qualitative approaches rely on the collection of
to account for social phenomena.
, information gathered by conducting
, through observation during field research, through
logical inference or deduction
of both observations and raw information.
Critical look at this approach
Overall, qualitative approaches rely on empirical evidence. But they are
not all strictly scientific
For example, field research that depends on the direct or personal experience of the research subject is an effective research method, as it builds up a body of empirical data that is subjectively interpreted with limited control over the research environment.
However, the deep insight and value that field research has brought to the social science are indisputable, even though its finding is subjective and not fully scientific.
Comparisons and Complementarities
approaches address the ‘
‘why’ and ‘how’ questions
For example, statistical data (quantitative) can tell you, for instance, that the number of young adults seeking psychiatric treatment has been increasing annually for the last five years.
However, it cannot tell you why this is happening, and how this has happened. It can only tell you what has happened. The ‘why’ and ‘how’ questions require qualitative treatment.
From this example, you can see that
quantitative and qualitative
approaches are not necessarily separate approaches,
but can be complementary.
approaches are (often) used to
formulate new hypotheses
and postulate theories whereas
approaches are (often) used to
For example, consider the hypothesis: “The rise of the hand phone has made pagers obsolete.” This hypothesis comes from a qualitative analysis and observation that linked the pager’s decline with the rise of the hand phone. Quantitative analysis techniques are then put to work to see if there is any empirical support for this hypothesis.
COMMON RESEARCH METHODS
involves direct observation
of people, events, and societies under research.
It is used
to gain an in-depth understanding
of the fundamental causes or dynamics of social phenomena.
Field research is used heavily by anthropologists and, to some extent, political scientists and sociologists.
The researcher both participates and observes first-hand, from the inside, the behaviour and activities of the subjects studied.
Ethnography, which is the systematic description of a society’s customary behaviours, beliefs and attitudes by an anthropologist, is a product of participant observation.
One strength of participant observation is the
increase in depth of understanding
as the social phenomena are studied from the inside-out.
Another strength is that the researcher is
more likely to identify
harder to catch
and interconnections between variables.
the behaviour being studied
but tries not to intrude or take part
in the behaviour.
For example, if a sociologist were to study youth gangs, he or she observes them from a distance without them knowing.
Non-participant observation helps to answer the ‘what’ questions and test statistical and survey data.
For example, if a survey reveals that wearing hats has become fashionable among Singaporean youths, non-participant observation can help to test if this is true.
Non-participant observation can
better capture the ‘natural behaviour
’ of its subjects. There will also be a reduced risk of the Hawthorne and Halo effects.
Non- participant observation is
less effective in explaining the ‘why’
There is increased
in trying to explain phenomena and a higher probability that incorrect correlations would be made.
if a non-participant observer notices that more secondary school students prefer eating homemade sandwiches in 2010 than in 1990, he
may have to infer or deduce
rom other observations
the following questions: Is this a new trend? Has the cost of buying food gone up? Are they saving money to buy other items, etc?
These remain guesses. Participant observation, on the other hand, would be better placed to answer these kinds of questions.
Controlled experiments involve the
comparison of specific changes
in two or more carefully selected groups that are identical in every way except that one group has been given the programme or treatment under study while the rest have not.
Experiments are conducted in a
controlled environment with controlled subjects
As such, the
likely to be accurate
and relatively more objective.
Control/Experimental Group and Result comparison
The control group consists of people of exactly the same profile as experimental group but which
will not undergo
The control group is ideally
given a placebo
(some activity or substance which is known to have no effect) to prevent subjects from knowing who is in the control group, and to prevent this knowledge from influencing the subject’s behaviour
of the experiment are measured by comparing the behaviour of the experimental and control group.
If there are
, the scientist infers a
link between the experiment and changed behaviour
If there are
no significant differences
, the scientist posts a
– a hypothesis that the activity or treatment has no effect.
To avoid biases/confounding covariates, scientists tending to the subjects may not be told which is the experimental group and which is the control group.
This is to prevent scientists from unconsciously influencing subject behaviour. This is called a double-blind experiment.
The experimental group consists of people that will undergo the programme or treatment
In survey research, a sample population does the survey, and they are taken to represent the wider population under study.
It involves a mix of quantitative and qualitative research methods.
Challenges and drawback
Questions must be
neutral to avoid suggesting
For example: “Tan Tock Seng Hospital is the best hospital in Singapore.”
This is a
loaded and leading question
which might encourage people who do not really have an opinion or who do not have a strong position to say “Tan Tock Seng” even if they do not mean it.
Better ways to phrase the question are “In your opinion, which of the following hospitals would you prefer?”
Doubtful quality of
, where people tend to behave
differently or unnaturally when under observation;
, where people are more likely to give ‘model citizen’ responses when they have no strong opinion on an issue
If sampling is done wrongly, the entire survey could be rendered worthless.
For example, in a survey on tastes and preferences of teenagers in Singapore, if by pure chance 40% of the sample population comprised of very wealthy Indian teenagers, the
results would no longer be representative
of the tastes of Singapore teenagers in general.
LIMITS OF SCIENCE/ SOCIAL SCIENCE
limits arising from the subject of study as well as the scientist being human (i.e. Biases)
Science and the Scientific Methods
Social scientific facts, generalisations, and theories are tentative and
do not represent ‘truth’
They are only as good as their verification method and are
subject to change
as new ways of measuring reality emerge.
Social scientific theories are
not absolutely predictive and accurate
They are always probabilistic as they deal with human beings who are inherently subjective and have the free will and creative power to adapt their thinking and action to new situations.
human behaviour is not fully predictable
The scientific method is very effective at testing empirical data but
at testing aesthetic, moral, spiritual and other
For example, Science has proven unsatisfactory in explaining and defining concepts like ‘truth’, ‘good’ and ‘beauty’ as well as religious experiences and subjective personal experience.
Science is thus
not a moral-ethical system
. It is only a technique of verifying
hypotheses and a tool to help people better understand their surroundings.
Do not deliberately hurt or harm the people in your experiment. (i.e. Chemical weapons testing)
Do not use science to exploit or manipulate people. (i.e. MK Ultra)
Do not use science to make unproven claims.
Do not make science into a moral-ethical system.
DIFFERENTIATING GENERALISATIONS FROM THEORIES AND FACTS
Theories represent the fruits of social science research, which provide
(though not absolutely valid)
explanations about the realities of human life
Theories are subjected to change or improvement, but they are
not mere hypotheses.
Examples of theories include Maslow’s Hierarchy of Needs Theory, William Glasser’s Choice Theory, etc.
has two meanings in social science, which we must be careful not to confuse.
Firstly, it refers to the
extent to which a theory can be applied
to understand human behaviour (i.e., how generalisable is the theory).
Secondly, it refers to
a hypothesis that has survived testing
and has been proven to be valid.
A theory is usually comprised of several interrelated generalisations.
For example, Aristotle’s ‘Man is by nature a political animal.’ This theory is a valid generalisation as it has been found to be accurate after testing. It is generalisable as it offers an accurate prediction of human behaviour for all places and time.
In social science, we understand ‘fact’ to mean something that has been found
tentatively true from scientific testing
Something factual may not necessarily be true, as a fact is only as good as the method used to verify it.
Understanding and Applying PROBABILITY
refers to the
or odds of an event
. It is at the centre of all social science theories.
and data collected in the social science are
, to a higher or lower degree, as it deals with human behaviour which always has exceptions and variations.
1. Statiscal significance
whether experiment/survey results are meaningful
, by measuring the odds that the results occurred by chance.
For example, ‘tests have shown a 70% chance that if you smoke more than five unfiltered cigarettes daily over a period of 30 years, you will develop lung cancer.’ What is the probability that replicated tests of this theory will not show 70%?
If there is a very low (1-5%) chance that incorrect results occur due to a random occurrence, then the data can be said to be statistically significant
2. Statistical/sample range
establish the range
within which experiment or survey
result is likely to be accurate
. Sampling error is a probabilistic measure of the range within which a survey result is likely to be accurate.
For example: “Surveys have shown that 80% of Singaporean men prefer blue to pink.” If the sampling error is +/- 5%, then the survey result could fall within the range of 75-85% if the experiment is replicated with another group of Singaporean men
Sample and Sampling Techniques
represent a larger population
For example, if you want to test how popular McDonald’s is in Singapore, you would select a sample group of about 1,000 Singaporeans to conduct the test.
Two common sampling method
Both stratification and random sampling methods can be used in parallel to compare results and spot errors. If results derived from different methods are very different, there is a possibility that one is an anomalous result, and the test is repeated.
This is a sampling technique which involves trying to reproduce a large population by
its composition in the sample.
For example, in a survey on housing preferences of adult Singaporeans, a stratified sample group of 1,000 Singaporeans might appear something like this: 7% Indian, 15% Malay, 70% Chinese, 8% ‘others’, of which 10% come from wealthier families, 70% are middle class, 20% are in the low-income group, which approximates the social structure of Singapore
is the act of
selecting at random a sufficient sample of the population such that there is a high probability of reproducing the essential characteristics of the total population
. Random samples cannot be too small or sampling error will be huge, rendering the sample useless.
of a sample can be determined mathematically to a point
where further sample size increases will have little effect on sampling error
. The optimum size need not be large. Random sampling
works on the principle of probability
For example, think of a coin flip. There is a 50-50 chance of getting heads or tails. However, we can only get this 50-50 distribution if we flip the coin enough times. If we flip just three times, we might get heads thrice in a row. If we flip 1,000 times, we will probably get a result much closer to 50-50. If we flip 3,000 times the result will still be around 50-50.
Hence, the optimum sample size is the smallest sample that will give a close approximation to the correct distribution.