SS3290

Social Work Research

Topic 3:
Conceptualization and
Measurement
 ❑ Concept and Conceptualization

❖ In conducting social work research, measurement

is pivotal to make the study concept (e.g.
depression) have an objective level, standard or
severity to be observed.
❖ Before measuring things, e.g. binge drinking and
depressive symptoms, you must first clarify the
concept in use, i.e. its definition and content.
❑ Concept and Conceptualization
(cont’d)
❖ Concept is a mental image that summarizes a
set of coherent and similar attributes,
characteristics, feelings and ideas in a
consistent mentality.
❖ Some concepts are straightforward, e.g. gender
and age, but others are abstract, e.g. self-
esteem, poverty, democratic attitude,
authoritarianism, righteousness.
❑ Concept and Conceptualization
(cont’d)
❖ More abstract concepts would be more difficult
to achieve a consistent definition and
measurement to them,
❖ The way we do to measure these abstract
concepts is to conceptualize them in a more
commonly accepted and reasonable manner
that most of the people (researchers) would
agree, e.g. common characteristics of
depression.
❑ Concept and Conceptualization
(cont’d)
❖ Conceptualization involves the process of
specifying what we mean by a term. In practice,
researchers would start with a nominal definition
of the concept.
❖ For example, you may define child abuse as the
domestic violence toward a child in terms of
either severe physical or psychological or sexual
harm that is inflicted on the child. This nominal
definition of child abuse concretely and broadly
identifies different types of abuse and specifies
that the harm must be severe.
❑ Concept and Conceptualization
(cont’d)
❖ After clearly defining the concept in use for
research, we must next go to operationalize the
concept into observation (observable variable)
(no observation  no measurement).
❑ Operationalization
❖ Operationalization is the process of connecting
concepts to observation (make the concept that
can be measured).
❖ The goal in operationalization is to devise an
operation (a measure/ variable) that can
measure the concept we intend to study.
❖ In operationalization, we must decide what and
how to measure specifically and concretely,
meaning what to be measured, how the
indicators are measured, and the rules used to
assign values to what is observed and to
interpret the values.
❑ Operationalization (cont’d)
❖ An example of operationalizing the concept of
“Social Class”, a perception of one’s social
status and hierarchy in society, that is an
abstract concept and can be operationalized by
Three Indicators to measure the concept, i.e.
income, education, and occupational prestige.
Concept Composing Indicators Study
Variables Variable
✓ Income Income +
Measure/
✓ Education Education +
Social Class Variable of
✓ Occupational Occupational
“Social Class”
Prestige Prestige
❑ Operationalization (cont’d)
❖ In conducting social work research, we usually
may use many existing and available measures
(i.e. scales and indexes), e.g. Geriatric
Depression Scale (GDS), Aggression Inventory
(AI), Satisfaction with Life Index (SWLI).
❖ They are validated measures already developed
by researchers from concepts to indicators and
measurement variable. Hence, you could adopt
them if they fit your research
❑ Operationalization (cont’d)
❖ It is advisable to use these available scales
and indexes as they are validated and can be
more reliable to measure the concepts that
we what want to study.
❖ However, when a scale we adopted from the
existing literature that does not fully fit our
study purpose, we could modify it to
accommodate our study needs according to
our cultural and academic understanding.
❑ Levels of Measurement
❖ We need to assign a numerical value to
represent the indicators/ study variable
constructed in the operationalization process,
which means that each indicator/ variable should
have a level of measurement, e.g. from
1(strongly disagree) to 5 (strongly agree).
❖ When we know a variable’s level of
measurement, we can better understand how
cases (participants) vary on what that variable
being measured and hence understand more
fully the variation of the cases in that variable.
❑ Levels of Measurement (cont’d)
❖ There are 4 types of Level of Measurement, they
are
➢ Nominal level of measurement- it identifies
variables whose values have no ordered
mathematical interpretation; they vary in nature
and quality, but not in amount/ quantity . The
attributes that nominal variables possess are
mutually exclusive and exhaustive.
➢ Hence, you can only assign dummy value to these
variables (i.e. 0 or 1) for presence or absence;
examples are gender, marital status, ethnicity,
religion.
❑ Levels of Measurement (cont’d)
➢ Ordinal level of measurement- variables
whose attributes can be logically mathematically
rank-ordered are ordinal measures. Different
values (i.e. 1, 2, 3) in the variable represent
more or less of the variable (the extent of
degree); but variables of ordinal level do not
specify the degree of difference in precise terms.
➢ An ordinal level of measurement are also
mutually exclusive and exhaustive; examples
are education attainment, social class, and client
satisfaction.
❑ Levels of Measurement (cont’d)
➢ Interval level of measurement- the values of an interval
measure represent fixed measurement units that can be
compared directly, but no absolute or fixed zero point.
Like ordinal measures, the attributes of interval variables
are mutually exclusive, exhaustive, and have a rank-
order.
➢ There are few true interval-level measures in social
work/ science research, many researchers treat
available scales and instruments created by combining
responses as interval-level measures, e.g. Satisfaction
with Life Index (SWLI). True interval-level measures
include Fahrenheit or Celsius temperature scale and IQ
test.
❑ Levels of Measurement (cont’d)
➢ Ratio level of measurement- a ratio-level
measure represents fixed measuring units and
an absolute zero point. The attributes of ratio
variables are mutually exclusive, exhaustive,
and have a rank-order and there are also equal
gaps.
➢ On a ratio scale, 10 is two times greater than 5
and 4 points higher than 6. Examples are age,
income, hours of volunteering participation,
number of arrests, and children in family.
❑ Measurement errors and evaluating
measures
❖ Even carefully designed measures can have
measurement errors because no measures can
perfectly measure the concepts that we intended
to measure in total objectivity. Two types of
measurement errors are common in research:
systematic and random error.
❖ Systematic errors include the forms of social
desirability, acquiescence bias, and leading
questions, which are all predictable errors.
❑ Measurement error and evaluating
measures (cont’d)
❖ Random errors refer to the conditions and feelings
that may affect respondents’ answering the
questions, which are generally unpredictable errors.
❖ However, if sample size is large enough, i.e. N
200, the effects of random error can be minimized or
canceled out.
❖ A good and sound measure must achieve two
standards:
✓ 1) always has consistent and stable responses
(reliability);
✓ 2) always measures the concept intended to
measure correctly (validity).
❑ Measurement errors and evaluating
measures (cont’d)
❖ Reliability means a measure always produces
consistent and equivalent scores when the
phenomenon being measures is not changing or
remained the same.
❖ If a measure is reliable, it is affected less by random
errors. There are four methods to prove how reliable
a measure is, they are
➢ Test-retest reliability- the measurement scores
obtained by a measure that do not change or remain
similar between two time points refer to test-retest
reliability.
❑ Measurement errors and evaluating
measures (cont’d)
➢ Internal consistency- when multiple items/
statements in a measure that can obtain
answers (responses) to the questions in a highly
correlated way, internal consistency in the
measure is assured.
➢ Cronbach’s alpha coefficient is used to indicate
level of internal consistency in a measure.
Generally, if α=.70 means adequate, and α
ranging from .80 to .90 means good, and α .90
is excellent.
❑ Measurement errors and evaluating
measures (cont’d)
➢ Alternate-forms reliability- researchers can
modify the original form of the measure to
another different form of the measure and
obtain scores from both the two forms, if the
scores of the two measures are not significantly
different, alternate-forms of reliability is
established.
❑ Measurement errors and evaluating
measures (cont’d)
➢ Interrater reliability- when researchers use
more than one observer to rate the same
participants, events, or places, the obtained
ratings (scores) from different raters that are
similar mean to have high interrater reliability.
➢ Assessment of interrater reliability may be based
on the correlation of the ratings between two
raters.
❑ Measurement errors and evaluating
measures (cont’d)
◼ Validity refers to the extent to which a measure
used to measure a certain concept indeed
means that concept (what is intended to
measure is being really measured).
◼ A valid measure is one that 1) closely related to
other apparently valid measures of the same/
similar concept; b) closely related to the
correlates of that concept (related measures);
and c) not related to measures of unrelated
concepts (unrelated measures).
❑ Measurement error and evaluating
measures (cont’d)
◼ There are four approaches used to assess
measurement validity, they are:
➢ Face validity- it means to carefully inspect the
contents of the items/ statements that are used to
construct a measure for a certain concept that do
really measure the concept, which means to see
whether the contents are appropriate on “its face”.
For example, how many times a person has used
illicit drugs in the past week is a face-valid item for
substance abuse.
❑ Measurement error and evaluating
measures (cont’d)
➢ Content validity- it refers to whether a measure
being constructed can cover the full range of the
meaning of the concept.
➢ To determine a measure that covers a full range of
its concept’s meaning, the researcher needs to
solicit opinions from experts and review literature.
For example, Social Alienation Scale (SAS) has two
dimensions/ subscales (Social Isolation and
Meaninglessness); the Multidimensional Scale of
Perceived Social Support has three subscales
(Family, Friends and Significant Others Support).
❑ Measurement error and evaluating
measures (cont’d)
➢ Criterion validity- it means the scores obtained
in the developed measure are similar, near or
related to the scores obtained with an already
validated similar measure of the same concept.
For example, CES-D can compare to other
depression measures or predicted scores from
other negative emotion measures, e.g. GHQ-12.
❑ Measurement error and evaluating
measures (cont’d)
➢ Construct validity- refers to confirm the relationships
of a measure with other measures specified in a
theory. The measure has construct validity if it
“behaves” as it should be relative to the other
measures in theory.
➢ For example, Danette et al (1999) compared parents’
levels of scores on the CES-D to a number of
indicators they felt related in research and theory, i.e.
fatigue, anxiety, and global mental health. Their
results showed that higher CES-D scores did have
more problems in these indicators.
Key Terms
 Concept and conceptualization
 Mental image
 Measurement
 Operationalization
 Observation
 Social class
 Geriatric Depression Scale (GDS)
 Satisfaction with Life Index (SWLI)
 Nominal level of measurement
 Ordinal level of measurement
 Interval level of measurement
 Ratio level of measurement
Key Terms
 Systematic errors
 Random errors
 Reliability
 Test-retest reliability
 Internal consistency
 Alternate-forms reliability
 Interrater reliability
 Validity
 Face validity
 Content validity
 Criterion validity
 Construct validity