1

CRIMINOLOGICAL RESEARCH AND STATISTICS

What is research?
Probably, your first question in this course might be: What is research? Before answering
this question, let me first illustrate what is not research.

Research is NOT . . .

 when you are told by your instructor to do “research” in the library about types of
crime, for example;
 when you tell your parents that you need money because you are going to an
internet café to do “research” about types of crime; or
 when you ask permission from your parents to go to your classmate’s house and
go home late because you are “researching” types of crime.

The term “research” that you usually use in these situations connotes another term.
Instead of using “research,” the best word you can use is “literature review.” A literature
review is an act of browsing the literature (e.g. books, research journals, and articles in
the internet) for a specific topic, say types of crimes, and collating what these sources say
about types of crimes through writing an essay. It should be noted, however, that
reviewing the literature about a topic is a requisite of research. In order to not reinvent the
wheel, the researcher must review the literature, especially previous research. By
reviewing previous research, one will know the gaps of the literature, and, therefore,
he/she can situate his research within the big puzzle of the literature about his chosen
topic.

Research- the scientific investigation into or of a specifically identified phenomenon and

is applicable to recognizable and undiscovered phenomena ( Dantzker, et. al, 2018). It is
also defined as a “careful, systematic study of knowledge in the field of criminology or
criminal justice, undertaken to discover or establish facts or causations of crime” (p. 2,
Soriano, 2011).

Researchers use scientific methods in solving problems. But what do we mean by the
word ‘problem’? A problem is “anything that a person finds unsatisfactory or unsettling,
a difficulty of some sort, a state of affairs that needs to be changed, anything that is not
working as well as it might” (Fraenkel & Wallen, 2006: 26). There are two types of
problem that researchers attempt to study and solve – practical and research problem
(Booth, Colomb & Williams, 2003).

Practical problems. This is probably your most known problem. Most of the time we
encounter practical problems in the family (e.g. conflict and finances), in work (e.g.
backlog of job assignments), or in any area of your life. In criminology, we see crime as
our primary problem. Booth et al. (2003:59) thus say: “A practical problem is caused by
some condition in the world, from e-mail spam to terrorism, that makes us unhappy
because its costs us time, money, respect, security, pain, even our lives.” Clearly crime is
a practical problem. It makes us unhappy because it endangers our lives and it costs us
 2

lots of resources. However, no one study could solve a practical problem. Instead,
researchers solve practical problems by first solving research problems.

Research problems. These problems are quite conceptual and sometimes hard to grasp
because all of us are used to encountering practical problems, and few ever heard of
research problems. According to Booth et al. (2003:59), “A research problem is
motivated not by palpable unhappiness, but by incomplete knowledge or flawed
understanding.” In other words, research problem pertains to what we do not know about
a practical problem. These things that we do not know answers questions of what, when,
where, how and why.
Let us take for example crime as a practical problem. One of the commonly agreed
solution to this problem is random mobile patrol. However, some police officers in
Kansas City doubted the effectiveness of this strategy in reducing crime and improving
perceptions of community members towards the police and fear of crime (Kelling et al.,
1974). They did not know (a research problem) whether this strategy could in fact lead to
lower crimes and improved community perceptions. They had no empirical (based on
facts generated by observations) evidence for its effectiveness. Thus, Kelling et al. (1974)
evaluated the effectiveness of random routine patrol of police officers across the city of
Kansas, USA. They wanted to determine whether intensive random patrol decreases more
crime and achieves more of some other measures of police
effectiveness than reactive patrol and normal levels of patrol. They compared three
groups of areas: control, reactive, and proactive. In control areas, they maintained normal
levels of patrol; in reactive areas, they did not employ any form of patrol; and in
proactive areas, they increased police visibility two to three times than normal levels. To
their surprise, after one year, they found no significant differences in crime levels and in
other criteria among the three areas, thereby casting doubt on the effectiveness of random
patrol in reducing crime. Perhaps, the problem of random patrol is that it is not systematic
and not directed to high crime areas. Recent experiments in criminology show that hot
spots policing, which are directed to high crime areas, does in fact reduce crime (see
Braga, Papachristos & Hureau, 2012).

These are the major steps in scientific method that researchers use to solve research
problems discussed above (Dantzker and Hunter, 2006; Maxfield & Babbie, 2005):

1. Identifying the problem

Here the researcher identifies the problem and formulates the arguments and speculations
to answer the problem. For example, if the topic is about bomb threats, after reading the
literature and finding that previous research focused primarily on characteristics of
offenders, he may reason out that in order to advance knowledge about bomb threats
there is a need to explore the spatial and temporal dimensions of these events. The spatial
and temporal patterns of bomb threats are things which researchers do not know – this is
the research problem. He will argue then that bomb threats are concentrated in space and
time, thus, non-random like other crimes.

2. Research design
The researcher then plans how to support this argument and to collect the data. She/he
 3

can choose among several research methods (i.e. survey, experiment, or secondary data
analysis). Next, he/she will decide who to study, when, what, and where. In this case,
since doing a national survey is not feasible considering logistical and financial
requirements, and experiment appears inappropriate, the researcher can use a secondary
data analysis of reports about bomb threats. He then decides the time frame, say 2005-
2014, and the place of study, say the whole Philippines not his/her province, to collect
more data since these events are rare.

3. Data collection
The researcher now collects data he/she can analyze to support his arguments. If reports
of bomb threats are routinely collected by a law enforcement agency, he can ask for such
data from such agency. If not, he/she can look for an alternative data source, like the
news reports because bomb threats are high intensity events like terrorist acts which gain
much attention from journalists and, thus, reported in the media. He can locate online
reports, instead of newspaper reports, about bomb threats to expedite the data collection.
After collecting the news reports, it is time for the researcher to code the news reports
according to the desired data, such as time, date, location, type of target, presence of
explosives, and words used by the perpetrator.

4. Data analysis
When the data are ready for analysis, the researcher then uses statistical software to
summarize the data, especially to determine whether bomb threats are non-random and
tend to concentrate in space and time. If such concentration of bomb threats occurred, he
can then create graphs, tables, and other statistical outputs of his/her analysis.

5. Reporting of results
The researcher then writes a research article (by the way this is just an essay) arguing that
bomb threats are non-random in space and time. To support his argument, he/she will
embed the graphs, tables, and other statistics in the research article, and try to interpret
and build conclusions about the results. He may relate the findings to the literature on
bomb threats and suggest some recommendations to authorities directly involved in
controlling such events

Why do research?
Dantzker and Hunter (2006) enumerate common reasons for conducting research in
criminology and criminal justice – curiosity, social problems, and theory testing

Who are researchers?

Researchers are men of science; thus, researchers are scientists. If you are to think,
however, of a scientist, you might visualize a person with eyeglasses in a laboratory
doing experiments involving chemicals and lots of laboratory instruments. The name
scientist is not strictly only for these people. Actually, a person is a man of science if
he/she uses scientific method in solving a problem; and a researcher uses the steps of
scientific method in conducting a research. It follows then that anyone can become a
researcher as long as he is well-versed in scientific method and uses it to solve problems.
 4

An undergraduate criminology student, criminology professor, police officer, correctional

officer, lawyer, or even a civilian can become men of science if they use the steps in
research to solve a crime problem.

Goals/Purposes of Research

According to Amante (2013) research is designed to:

1. Discover new facts about known phenomena.
2. Find answers to problems which are only partially solved by existing methods and
information.
3. Discover previously unrecognized substances and elements.
4. Improve existing techniques and develop new instruments and product.
5. Discover pathways of action of known substances or elements.
6. Satisfy the researcher`s curiosity.
7. Find answers to queries by means of scientific methods.
8. Etc.
Classifications/Categories of Research
1. According to purpose- Travers (1978) identifies three broadly different kinds of
research according to purpose, namely:

a. Predictive or Prognostic Research- has the purpose of determining the future

operation of the variables under investigation with the aim of controlling or
redirecting such for the better. It proposes to give the result from one specific
educational practice and seeks to establish a close statistical connection between
characteristics of the student and prediction of eduacational outcome.
b. Directive Research- determines what should be done based on the findings to
remedy an unsatisfactory condition, if there is any.
c. Illuminative Research- is concerned with the interaction of the components of
the variable being investigated by showing the connections.

2. According to Goal
a. Basic or Pure Research- is done for the development of theories and principles.
It is conducted for the intellectual pleasure or learning much of this kind of
research has been done in psychology and sociology. The findings from basic
research often have little or no applicable use in the field of criminal justice.
However, such research may be used as the foundation for subsequent applied
research and criminal justice policy. For example, exploring what process was
being used by the largest 17 municipal agencies for psychological pre-
employment screening of police officer candidates.
b. Applied Research- is the application of the results of pure research. This is
testing the efficacy of theories and principles. It is concerned with the acquisition
of new information for the purpose of helping develop the scholarly discipline or
field of study in which the research is being conducted.
 5

c. Multipurpose Research- is the scientific inquiry into the issue or problem that
could be both basic and applied research. This type of research generally begins
as basic, but the results are found to have an applied purpose. For example, level
of job satisfaction. From basic perspective, the data may simply describe how
officers perceive satisfaction with different aspects of their jobs, becoming
descriptive in nature. However, the same findings could be used to evaluate the
police agency by examining those areas where satisfaction is the lowest and
leading to efforts to determine how to improve these areas (Dantzker et. al.,
2018).

3. According to the level of iinvestigation- Vockell (1993) categorizes research

according to the levels of investigation into:
a. Exploratory Research – the researcher studies the variables pertinent to a
specific situation. It is designed to describe an existing problem situation and
examine the underlying factors that contribute to the emergence of the
problem, the nature of which is not yet well known. This type of research is
loosely structured but valuable methodological strategy.
b. Descriptive Research- the researcher studies the relationships of the
variables. It finds answer to the questions who, what, when, where, and how.
What may be described are characteristics of individuals or groups.
Descriptive research aims to describe precisely the dimensions of a
phenomenon. This type of research answers questions like: What is the
average number of police officers in the police stations in Negros Oriental?
What is the average caseload of police investigators in Negros Oriental? What
are the hot spots of carnapping in Dumaguete City? How does a typical crime
scene investigation take place? Results of descriptive research are also aimed
to inform policies. For example, to formulate information-based programs to
curb drug abuse in the country, the Dangerous Drugs Board (2008) conducted
surveys that are aimed to describe the current situation of drug abuse among
Filipinos. Also, a chief of police may ask a report on the hot spots of crime in
his jurisdiction on which police patrol are directed.
c. Experimental research- it studies the effects of the variables in each other. It
involves the administration of research stimuli to participants in a controlled
environment.
d. Explanatory Research- Instead of answering questions such as ‘what is/are’
or ‘how’, explanatory research answers ‘why’ questions. This type of
research, however, may first describe a phenomenon, say crime, and proceeds
to explain why such phenomenon occurs. Research in criminology and
criminal justice abounds with explanatory research. For example, after
findings that some barangays are hot spots of violent crimes, the researcher
may attempt to explain why such hot-spotting occurs. He/she may use
variables such as poverty incidence or population mobility. While finding that
poor neighborhoods have higher violent crime levels than rich neighborhoods,
the researcher may proceed to another ‘why’ question. This is what Sampson,
Raudenbush and Earls (1997) did: They use the concept of collective efficacy
to explain why such pattern occurs. They defined collective efficacy “as social
 6

cohesion among neighbors combined with their willingness to intervene on

behalf of the common good” (Sampson et al., 1997: 918).

4. According to type of Analysis- Wiersma (1995) classifies this into:

a. Analytical research- the researcher attempts to identify and isolate the
components of the research situation.
b. Holistic research- begins with the total situation. Focusing attention on the
system first and then on its internal relationships.

5. According to Scope
a. Action Research- this type of research is done in a very limited scope to
solve a particular problem. It is an on the job type of problem solving.

6. According to choice of answers to problems

a. Evaluation research- all possible courses of action are specified and identified
and the researcher tries to find the most advantageous. Almost every year there is
a new program in the agencies that constitute the criminal justice system.
However, in order to avoid wasting resources by an ineffective program’s
continued implementation, policymakers and researchers must evaluate the
effectiveness of criminal justice programs, whether they achieved what they
intended to achieve. Earlier, we discussed the experimental study of Kelling et al.
(1975) to assess the effectiveness of random patrol.
b. Developmental Research- the focus is on finding or developing a more suitable
instrument of process than has been available.

7. According to Statistical Content

a. Quantitative research- is one in which inferential statistics are utilized to
determine the results of the study. In this study, concepts are assigned numerical
value.
b. Qualitative research- the use of quantity or statistics is practically not utilized.
Descriptive data are gathered rather than quantitative data. In this study concepts
are viewed as sensitizing ideas or terms that enhance our understanding.

8. According to time element

a. Historical research- describes what was.
b. Descriptive research- describes what is.
c. Experimental research- describes what will be.

Idiographic and Nomothetic Explanations

One of the most important uses of research is to explain something (in criminology,
crime and justice). Why do some persons commit crime? Why does a husband maltreat
his wife? Why do police officers commit misconduct at work? Why do some places have
higher crime levels compared to other places? Knowing the answers to these questions
can, of course, inform policy makers; and answering these questions using research can
be addressed using two types of approach - idiographic and nomothetic.
 7

Idiographic and nomothetic researches differ in terms of the number of study subjects, the
number of explanatory factors used to explain something, and the generalizability of the
findings. Idiographic research studies one case or instance to have an in-depth
investigation of numerous factors to explain something, say crime. However, since only
one case is to be studied, the researcher is only capable of explaining such case and
cannot use his findings to explain other cases in other situations. On the other hand,
nomothetic research studies several cases but uses few factors to explain something; the
findings of which, however, are generalizable to a range of cases.

We will illustrate the difference between idiographic and nomothetic research by

answering the research question: Why do husbands maltreat their wives? To answer this
question we will use as an illustration the studies of Ysmael (2009) and Babu and Kar
(2010). Ysmael (2009) studied the case of Lumen, a battered married woman in Cebu
City. He used the case study method to discover the factors that contribute to the
victimization experiences of Lumen. The focus on one woman provided him with very
rich information about the experiences of Lumen before, during, and after the abuse.
However, one pitfall of his research is that his findings could not be generalized to other
cases of battered women. In other words, what he discovered could not be used to explain
the cases of other abused married women in Cebu City. On the other hand, Babu and Kar
(2010) used the survey method to study 1,718 married women and 1,715 married men in
Eastern India. Although their findings are not that rich compared to Ysmael’s (2009),
their findings could be generalized to several cases, possibly, even up to the whole
population of married men and women in Eastern India.

Inductive and Deductive Reasoning

We will now turn our attention to another dichotomization of research – inductive and
deductive reasoning. In this dichotomization, we shall try to familiarize ourselves with
the role of theory in criminology and criminal justice research. Although there is no
agreed upon definition of theory, we shall agree with Champion (1993) who synthesized
the definition of theory provided by social scientists Robert Merton and Arnold Rose.
Theory refers to “an integrated body of assumptions, propositions, and definitions that
are related in such a way so as to explain and predict relationships between two or more
variables” (Champion, 1993:19). (We shall elaborate more on theory in the lesson about
research problem and measurement.)

Inductive reasoning goes from specific details to general propositions, while deductive
reasoning goes the opposite. Let us use the reasoning used in criminal investigation and
criminal prosecution in court to enlighten us about these research approaches. Criminal
investigators often use inductive reasoning and proceed from specific details to general
propositions. For instance, after the investigator has collected evidence (the weapon used,
samples of blood, fingerprints, and other trace evidence), he learns that the weapon
belongs to Juan dela Cruz and the samples of blood, fingerprints, and other trace evidence
match with that of Juan, the investigator then can speculate that Juan might be the one
who has committed the crime. Note that the investigator begins from specific details
 8

(pieces of evidence) and proceeds to a general proposition (that Juan might have
committed the crime).

On the other hand, a criminal prosecutor proceeds in the opposite way when they
prosecute an accused in court. As you have learned from your criminal procedure class, a
prosecutor files an information (a written allegation) in court. In this information, he/she
alleges that the accused might have committed the crime. In the trial, the prosecutor rests
on this proposition and tries to argue using pieces of evidence that the accused might be
the one who committed the crime. Thus, in trial, he/she presents evidence just to support
his claim. Note that he/she goes from a general position (that the accused might have
committed the crime) and proceeds to specific details (pieces of evidence) to support the
proposition.

Research can also be either inductive or deductive in its approach. In inductive research,
the researcher analyzes data, and out of this analysis, he/she attempts to build a
proposition or a theory. Inductive research uses some qualitative research methods or
quantitative data mining strategies. In deductive research, the researcher analyzes data to
test a given theory. Researchers often experiment and conduct surveys to test theories.
Looking back to our previous example of the study of Donner and Jennings (2014), we
will find an example of a deductive research. At the start of their research they already
had their proposition from self-control theory – that police officers who had low self-
control would commit more police misconduct that those who had high self- control. To
test this theoretical proposition they analyzed survey data on more than one thousand
police officers in Philadelphia. The results support their proposition.

What is research design?

Research design is the “blueprint” of your study (Dantzker & Hunter, 2006). When you
design a study you specify “exactly who or what is to be studied, when, how, and for
what purpose” (Babbie, 2001:90). Designing research requires the researcher to be very
specific in identifying the general purpose of the study – whether the researcher wants to
explore a little known phenomenon, describe a phenomenon, or explain why such
phenomenon occurs (purpose of the study). The researcher must also decide who or what
he/she is going to explore and describe, or whose actions is he/she going to explain (unit
of analysis). He/she also decides the time he observes what or who he observes – once
only or overtime (time dimension). Finally, he plans how he is going to collect the data
needed in his study (research method). We will start with deciding the overall purpose of
your study.

Ecological fallacy occurs when a researcher uses his findings on the characteristics of
groups to explain characteristics of individuals. For example, when you compare cities in
terms of homicide rates and poverty incidence and you find poorer cities having higher
homicide rates, it is an ecological fallacy to conclude that poor individuals are violent
compared with rich individuals. In this case, you do not really know who committed the
killings in poor cities; it might be possible that rich individuals are responsible for the
large proportion of those killings. Thus, you can only say that poor cities tend to have
 9

higher homicide rates compared with rich cities. Likewise, we all know that among the
three island groups (Luzon, Visayas, and Mindanao) in the Philippines, Mindanao has the
highest level of bombing incidence. However, you would commit a fallacy when you
conclude that Dapitan City (a city in Mindanao) has higher bombing incidence than any
city in Metro Manila, Luzon. It is because bombings in Mindanao are not randomly
scattered across all cities in its islands; rather, bombings are clustered in few
cities/municipalities other than
Dapitan (Relatorres et al., 2013).

Individualistic fallacy. This fallacy occurs when one generalizes

his/her observation about an individual to groups or organizations. For
example, when one hears from the news that a certain police officer has
been found positive of using dangerous drug, it is a fallacy to generalize
and say: I am afraid to report to the police station drug crimes in my
neighborhood because the police are even involved in drug use.

Reductionism occurs when researchers try to disregard other units of analysis and
concepts from other fields. This is sometimes the case of crime causation because the
field criminology is multidisciplinary; it involves researchers from criminology, criminal
justice, sociology mathematics, biology, economics, chemistry, and political science, to
name a few. Some researchers may limit their focus on what tend to cause crime.
Economists might focus only on cost and benefits (economic reductionism); biologists
might focus on genetic influences (biological reductionism); and sociologists might focus
on cultural factors (sociological reductionism). Also, some researchers focus solely on
one unit of analysis and discredit other units. For example some psychologists focusing
on the individual might argue that those researchers studying variation of crimes across
geographic areas run the risk of committing ecological fallacy. Or, those focusing on
places might discredit those researchers studying individuals. These acts result to
reductionism of units of analysis and concepts on crime causation. We should note,
however, that crime occurs because of multiple processes both in the individual and
group level. Studying crime and justice, therefore, requires focus on different units of
analysis and concepts.

How many times should I collect data?

Because of theoretical and logistical restrictions, researchers carefully choose between

two time dimensions of observations: cross- sectional studies and longitudinal studies
(Babbie, 2001; Maxfield & Babbie, 2005).

Cross-sectional studies

Almost all studies cited so far in this module are cross-sectional studies. Cross-sectional
studies collect data at one time, or the data collected cover only one short period, say one
year to two year city crime data. By choosing this type of study, researchers can save a
lot of their time, effort, energy, and, of course, money. One classic example in place and
crime research is the study of Sherman, Gartin and Buerger (1989) on the spatial
 10

distribution of predatory crimes in Minneapolis, USA. They analyzed crime data from
December 15, 1985 to December 15, 1986 and found that only 3% of the city streets
accounted for 50% of predatory crimes. Note that they studied a one-year crime data;
thus this is a cross-sectional study. Most surveys are also cross-sectional in nature. For
example, the Dangerous Drugs Board conducted a national survey on the drug use
situation in the Philippines in 2008; a data gatherer surveyed a particular respondent at
one time only. The ease of administering cross-sectional studies makes it attractive to
some researchers who wish to explore, describe, or explain a phenomenon.

However, this type of study provides a weak evidence when the researcher wishes to
investigate causation. To argue that one variable (e.g. low self-control) causes another
(e.g. violent crime), three requirements are needed – correlation between the two
variables, the causal variable precedes the effect, and the relationship must not be
spurious (Maxfield & Babbie, 2005). Cross-sectional studies seem limited in complying
with the second requirement. Cross-sectional studies cannot establish that the
independent variable, say low self-control, precedes crime because the data are collected
at one time only. This limitation is what some longitudinal studies are trying to fulfill.

Longitudinal studies

Some researchers use data collected through longitudinal studies.

Longitudinal studies collect data from a respondent over a long period
of time or the data cover a long period of time, say 10 or more years
(Babbie, 2001; Maxfield & Babbie, 2005). There are three major types of longitudinal
studies – trend studies, panel studies, and cohort studies.

Trend studies display the changes of the frequency of an event over a long period of
time. A city mayor, for example, may request a report from the chief of police to show
the crime rate trend in his city to see the overall picture of crime in terms of 10 or 20
years. Above, we discussed the study of Sherman et al. (1989) on the spatial distribution
of crimes in Minneapolis for one year; however, this study may suffer from a “snapshot
effect.” They cannot really tell whether the distribution remains true in a range of years.
To overcome that limitation, Weisburd et al. (2004) conducted a longitudinal study on
crime concentration in Seattle, USA from 1989 – 2002. Their study confirmed the earlier
findings of Sherman et al. (1989) that crimes are non-random and concentrated in few
areas of the city even up to 14 years.

Criminologists who study causes of crime tend to favor longitudinal panel studies which
collect data from the same sample respondents for 2 or more periods of time. For
example, Thornberry et al. (1994) used the data collected by the Rochester Youth
Development Study in the US to test the propositions of interactional theory of crime.
They used the data collected in the Wave 2, 3, and 4 of the study. In Wave 2, adolescents
were interviewed in their 8th or 9th grade; in Wave 4, these same adolescents were
interviewed in their 10th grade.

Finally, cohort studies research on “specific subpopulations, or cohorts, as they change

 11

over time” (Babbie, 2001). This type of study is similar to panel study in which collection
of data from respondents are done in two or more periods. However, one major difference
between panel and cohort studies is that in panel studies researchers collect data from the
same respondent over time; while in cohort studies, researchers collect from another
sample of cohorts and, thus, not necessarily the same respondents with the first data
collection. For instance, if you want to study a cohort of babies born in Dumaguete City
in 2010, you generate a sample of 500 out of 2000 babies. In 2015, you conduct another
data collection; this time, you generate another 500 five year old children born in 2010 in
the City. Then, you do the same procedure in 2020, 2025, and 2030. Note that by
generating another sample of 500 children, it is not necessary that the same respondents
in the first sampling are chosen in the second sampling of respondents. The only
important criterion is that the respondents were of the same subpopulation (in this case,
those babies born in 2010).

However, there are times when criminologists attempt to approximate longitudinal

studies using cross-sectional studies. This is because longitudinal research is more
expensive, time consuming, and energy exhaustive than cross-sectional studies. For
example, Gottfredson and Hirschi (1990) state that the cause of having low self-control is
poor parenting styles during the first eight years of a child. To study this, researchers
would conduct a longitudinal study. They may collect data about the parenting styles of
parents when the child ages below eight years old; and collect data on self-control level
of the child when he/she ages above eight. However, some researchers do not have
enough funding to conduct a longitudinal study (which could possibly amount to
millions!). Thus, they resort to retrospective studies to approximate longitudinal ones.

Retrospective research “asks people to recall their pasts” (Maxfield & Babbie, 2005:
99). For instance, Morris, Wood & Dunaway (2007) tested whether parenting styles
affected levels of self-control among American Indians. They asked high school students
on the parenting styles of their parents when they were younger. One limitation of this
method, however, is that the respondents might not remember totally how they were
supervised by their parents when they were younger; instead, the students’ response
might reflect the current parenting styles. Nevertheless, when longitudinal data is not
available or accessible, this method is acceptable, but the researchers must point this out
in their research article as one of the limitations.

How should I collect the data?

Finally, we have arrived at the last major consideration in designing a research study.
After deciding what your research is for, who or what to be studied, and when to collect
data, you might ask: How should I collect the data I need? To answer this question, you
should first familiarize the different research methods. A research method is a process
of collecting the data needed to answer a particular research problem or question. You
can choose from a range of quantitative research methods – such as experiment, survey,
and secondary data analysis – and qualitative research methods – such as narrative
research, phenomenology, grounded theory, ethnography, and case study (Babbie, 2001;
Creswell, 2007; Fraenkel & Wallen, 2006; Maxfield & Babbie,
 12

2005).

Note, however, that none of these methods is superior to another method. Your choice
depends on your research problem. For example, if you want to determine the
effectiveness of different policing strategies in reducing crime at crime hot spots, it is
obvious that you need an experiment (you cannot do this using a survey). If you want to
determine the perceptions of police officers on their job assignments, you should use a
survey. And, if you want to study the techniques of budol-budol gang on how deceive
passive victims, then you can use qualitative case study or phenomenology. In this case,
you cannot conduct an experiment or a random survey because the population frame of
budol-budol gang members is not known.

Quantitative methods

As you can remember, this research approach is oriented towards the use of numerical
data. There is an array of quantitative research methods used in criminology and criminal
justice, but we will only discuss the most common ones – survey, secondary data
analysis, and experiment.

Surveys

Perhaps, the most common quantitative research methods in criminology and criminal
justice is survey research. Survey research is a method that collects data from a sample
of a population by asking questions in order to describe some characteristics of that
population (Bachman & Schutt, 2006; Fraenkel & Wallen, 2006). Survey is one of the
oldest methods of research (Babbie, 2001) and is often used for descriptive as well as
explanatory research in criminology and criminal justice (Maxfield & Babbie, 2005). One
drawback of surveys, however, is that respondents might not tell the truth; this is
common especially in criminological research which investigates on the criminal acts of
the respondents.

Since conducting an experiment on crime and justice issues is oftentimes unethical,

criminologists use surveys to test theories. For example, Agnew (1992) asserts that
persons who experience strain (negative events) in their lives are likely pushed towards
delinquency. You cannot do an experiment by sampling individuals and letting half of
them experience negative events while the other half do not, and then observe whether
those who experienced negative events indeed commit crimes. This is desirable but
purely unethical!

Secondary data analysis

There are times, however, that researchers can take advantage of readily available data
though secondary data analysis. In secondary data analysis, researchers analyze data
that have been routinely collected by different criminal justice agencies or other
organizations that conduct surveys. The researcher can save a lot of resources through
this method. Some of the studies I have cited in this module used secondary data. For
 13

example, Sherman et al. (1989) collated and analyzed data from Minneapolis computer-
aided police dispatch system to study the hot spots of predatory crime. Some agencies in
the Philippine criminal justice system can provide rich data for your research; these
include – the Philippine National Police, Bureau of Jail Management and Penology,
Bureau of Fire Protection, Philippine Drug Enforcement Agency, National Bureau of
Investigation. Other agencies outside the criminal justice system (e.g. National Statistics
Office, City Civil Registrar, City Engineer’s Office, Traffic Management Office, and
private agencies that provide assistance to victims of abuse) can also give you data. You
can also
locate internet databases (e.g. United Nations Office on Drugs and Crime
website www.unodc.org) that provide data on criminology and criminal justice issues.

Experiments

Perhaps, the ‘gold standard’ in research is the experimental method. Experimental

research investigates whether an independent variable (e.g. police patrol) produces an
effect on another variable (e.g. homicide rate) by manipulating the independent variable
(Babbie, 2001; Maxfield & Babbie, 2005; Fraenkel & Wallen, 2006). One classic
example of experiment in criminology and criminal justice is the study of Kelling et al.
(1974) which we have already discussed in the previous lessons.
The following illustrate how Kelling et al. (1974) conducted the study in
Kansas City:

Initially, they randomly divided the 15 beats of South Patrol Division into three groups.

Then, five beats were designated as reactive groups; another five beats as control groups;
and, the remaining beats as proactive groups. However, they made sure that none of the
reactive beats were contiguous.

Next, they applied different dosage of mobile police patrol in each group. In the reactive
beats, police patrol was withdrawn; police officers only went to the beats when someone
called them for help. In the control beats, the normal patrol effort - one patrol car – was
conducting preventive patrol. In the proactive beats, the dosage of patrol was increased to
two or three times.

To measure the effects of the experiment, they gathered data (before, during, and after the
experiment) on monthly reported crime, victimization experiences, attitudes and fear of
crime from household members, victimization experiences and satisfaction with the
police from businesses, police-citizen encounters, traffic accidents, arrests, and police
response time.

You might notice that Kelling and his associates also used surveys to collect information
on community members’ perceptions and experiences before, during, and after the
experiment, even though they were using primarily experimental method. It should be
noted that using multiple methods can strengthen a particular research project. In the case
of Kelling et al. (1974), effectiveness of preventive patrol was not only measured in terms
 14

of crimes reported to the police but also on other measures of effectiveness like
unreported crimes, fear of crime, and satisfaction with the police services. With this, they
speculated that they could provide a more comprehensive evaluation of random
preventive
patrol.

Qualitative research methods

Some criminology and criminal justice researchers also use qualitative methods in
studying crime and justice related problems. Qualitative researchers often argue that
quantitative techniques are superficial; thus, they turn to analyzing participants’ words
and actions. Among the most common qualitative research methods are narrative
research, phenomenology, grounded theory, ethnography, and case study (Creswell,
2006).

Narrative research

Narrative research is often used in doing research in crime and justice issues, especially
when trying to collect stories from criminal justice actors. Narrative research is a
qualitative method that explores the life of an individual/s through interviews and
collection of documents (Creswell, 2006). This type of research tells the story of one or
more individuals’ life. In criminological research, criminal justice actors (e.g. criminals)
are often given a chance to tell their stories using their own perspectives. The researcher
then looks for themes and attempt to “restory” the participants’ life. For example, Dilmon
and Timor (2014)
conducted a narrative research on 12 men serving sentences for killing
their female partners in Ramla, Israel. The men were made “to express
how they perceive, describe, and explain the murder of their partners”
(Dilmon and Timor, 2014).

Phenomenology

Another qualitative research that is gaining attention in criminology is phenomenological

research methods. Phenomenology studies several individuals who have experienced the
same phenomenon and attempts to extract the essence of their lived experience
commonly through interviews (Creswell, 2006). The essence “focuses on the common
experiences of the participants” (Creswell, 2006:62). Unlike narrative research which
often studies only one participant, phenomenological researchers conduct interviews on 5
to 25 respondents. One example of phenomenology is the study of Narag,
Maxwell, & Lee (2013) on 20 participants of a Driving Under the Influence or Driving
While Impaired (DUI/DWI) Court program in the US. The participants, who had at least
two drunk-driving offense volunteered to undergo the program. By using
phenomenological inquiry, Narag and his colleagues found that the program was
effective in reforming the offenders. The participants generally viewed the program
facilitators as friendly, supporting and caring. These experiences help most participants to
successfully complete the 36-week program.
 15

Grounded theory

There are times, however, when criminologist and criminal justice researchers want to
build a theory based on the views of research participants. These researchers would need
to use grounded theory to explain processes and actions of the participants. Grounded
theory is a qualitative inquiry that employs interviews and aims to develop a theory
based on the views of several participants. This type of study uses interviews of 20-60
individuals. Lopez and Emmer (2000) uses grounded theory in building a theory on
delinquent crime contexts upon learning that theories of delinquency concentrated more
on individual differences as causes of crime and neglected situational precipitators of
serious juvenile offending. To generate the crime contexts that most likely pushed
juveniles to commit serious offenses, they conducted semi- structured interviews on 24
male juvenile offenders at the Texas Youth Commission. The participants were asked to
recall and share their experiences when they committed a crime; then additional, probes
of 5Ws and 1H were used to elicit information on the immediate, situational contexts
wherein the offenders were motivated to commit a serious offense. Grounded theory is
helpful when the researcher wishes to provide a theoretical framework grounded on the
participants’ own views.

Ethnography

Sometimes, in order to obtain more in-depth analysis of a phenomenon, some

researchers go beyond “snapshot” qualitative research (e.g. narrative, phenomenology,
grounded theory) and use ethnographic research. Ethnography is a qualitative inquiry
that describes and interprets the patterns of a group’s culture over time (Creswell,
2006). To describe the culture of a group, the researcher conducts observation
techniques, interviews, and, sometimes, collects other sources of data, for an extended
period of time. One good example is the study of Valdez and Sifaneck (2004) on
Mexican-American gang members in drug markets. Of course, there is no official list of
gang members and gang members are hard to access for a survey; thus, Valdez and
Sifaneck (2004) used ethnographic research using field observations and interviews.
Field work in their study lasted for four years. In two areas in a US Southwestern city,
two indigenous field workers conducted social mapping for six months to identify the
gangs and the places they frequented. After this, the field workers entered into the social
networks of these gangs. The field workers then recorded rich information through field
notes while conducting field observations, and later, conducted interviews on known
gang members about their role in drug selling and drug use patterns.
Ethnographic research is appropriate when enough resources are available and
researchers wish to have a more in-depth analysis of a group’s culture.

Case study

Finally, some researchers use case study in criminological research. Case study is a
qualitative inquiry on an issue or problem by conducting in-depth analysis on an
individual/s, event, program, or activity (case/s). Similar to other qualitative research
 16

methods, case study researchers use extensive observations, interviews, and documentary
analysis on a case or cases. To study the problems and challenges facing crime task
forces, Philipps and Orvis (1999) used the intergovernmental relations conceptual
framework in examining East Texas Violent Crime Task Force. They used semi-
structured interviews on members and administrators of the said task force, and they also
gathered and analyzed legal documents and memoranda concerning the task force. After
these, they discussed the problems and challenges facing crime task forces using the East
Texas Violent Crime Task Force as an example.

FORMULATING THE PROBLEM

What is my area of interest and topic for investigation?

Area of Frame of reference Unit of analysis Topic for investigation

interest
Criminal Environmental Person Decision-making of offenders
etiology criminology, Rational Streets Hot spots/hot times of crime
choice perspective
Criminal sociology Neighborhoods Social disorganization, collective
efficacy, fear of crime, social
disorder and crime, and crime
rates
Cities Social disorganization, collective
efficacy, and crime rates
Groups (Gangs) Subculture, social learning
Persons Strain, social control, parenting
and crime
Biological/chemical Persons Genes and crime, exposure to
criminology chemicals and crime, diet and
crime
Psychological Persons Aggression, self-control,
criminology personality and crime
Policing Deterrence Neighborhoods, Hot spots patrol, reactive patrol
strategy streets
Community relations Neighborhoods Community policing,
neighborhood watch
Problem-oriented Neighborhood, Problem-oriented policing,
streets
Zero-tolerance Neighborhoods, Broken windows policing
streets
Rehabilitat Community-based Persons Characteristics of probationers,
ion recidivism
Formal Parole and probation
organizations management
Institutional based Persons Recidivism of offenders, job burn
out of prison staff, morale
Formal Prison population and violent
organizations disorders, prison morale
 17

Judiciary Restorative justice Persons Mediation, conciliation,

arbitration of offenders and
victims
Formal Barangay justice system
Organization
Retributive justice Persons Judge’s decisions, plea
bargaining between the accused
and offended party
Organizations Speed of trial of different courts

What is my problem statement?

From your topic, you can then formulate your research problem. This step is the most
important step in the research process because a problem that is clearly stated is already
half-solved. Your research problem will then dictate your research design – who or what
to study, when, where, and how (Kothari, 2004). Without a research problem, you will
certainly have no direction and you may tend to collect all sorts of data without purpose
(Booth, Colomb & Williams, 2003).

We will illustrate in the next subsections how to frame our research problems in
descriptive and explanatory research.

A. Exploratory/Descriptive research

Suppose your area of interest is in criminal etiology, your frame of reference is

environmental criminology, and your topic is patterns of crime and you choose motor
vehicle theft as your crime of choice, then your research problem would pertain to the
things we do not know about motor vehicle theft, say, in Negros Oriental. These things
that you do not know (research problems) are sometimes stated as questions (Fraenkel &
Wallen, 2006). When we say, “Things we do not know,” we mean information which
have not been researched and published in a research journal. Here are the steps is
defining your research problem.

Step 1. Ask as many questions as you can.

The first thing we must do is to ask questions whose answers are not yet known. Now,
assume that we have arrived with the following questions on the things that we do not
know about motor vehicle theft in the province of Negros Oriental (our study area):

1. Where do motor vehicle (MV) thefts occur frequently in Negros Oriental?

2. When do MV theft occur frequently?
3. What are the usual hot products of MV theft?
4. How do offenders steal MV theft?
5. What are the personal characteristics of MV thieves?
6. What are the motives of MV thieves?
7. How do the victims contribute to MV theft commission?
8. What are the policing strategies employed by law enforcement
agencies to curb MV thefts?
 18

9. Does poverty increase MV thefts in a city?

10. Does collective efficacy decrease MV theft?

These are just few of the many (and limitless) questions that we can ask ourselves about
MV thefts in Negros Oriental. Because of space limitations, we will just deal with these
10 questions for now.

Step 2. Eliminate unrelated/unrealistic/impractical questions

The next thing we should do now is to select questions according to some criteria, as
follows:

 Think of your overall purpose. This time we are defining a research problem for a
descriptive study. We want questions whose answers describe MV theft. Thus,
questions that ask explanatory answers must be eliminated. So we will eliminate
Questions # 7, 9, and 10. These questions are for explanatory research.

 Think of your frame of reference. Your frame of reference is environmental

criminology. Environmental criminology is a perspective that assumes that
offenders are rational; opportunity facilitates crimes; crime occurs when a
motivated offender converges with a suitable target in the absence of a capable
guardian; and places and times where this convergence usually occurs tend to be
the hot spots and hot times of crimes. With this, we can eliminate Questions 9 and
10. These questions can be answered using the social structure perspective (social
disorganization theory).

 Think of your topic. Your topic is patterns of crime; this time, patterns of MV
theft. Thus, we can eliminate Questions 8, 9 and 10. These questions do not ask
about patterns of crime. They are unrelated to our topic.

 Think of the available and accessible data on MV theft. Of course, for a beginner
like you, the easiest source is the law enforcement agency (i.e. Philippine
National Police – Highway Patrol Group) that handles MV theft cases in the
province. Suppose that you have asked permission from that agency and they
provide you their data. However, in their database only the (a) location of the
theft, (b) the year, month, day, time of the theft, (c) type, model, and make of
vehicle stolen are recorded. This is not surprising because most databases were
not built for research purposes. With these data, what remains are Questions # 1,
2, and 3. You cannot answer the other questions using the current data unless you
conduct surveys and qualitative research methods.

Step 3. Write your problem statement.

Always remember that a research problem is something that we do

not know yet or about which little is still known. You are left with three
 19

questions – Questions 1, 2 and 3. In journal articles, research problems

are often stated as purpose statement and specific questions follow this
statement like this one:

The purpose of this study is to explore the patterns of motor vehicle (MV) theft across
space, time, and targets. Specifically, this study will answer these questions: (a) What are
the hot spots and hot times of MV thefts in Negros Oriental? and (b) What are the hot
products of MV theft in the province?

Problem statement
The purpose of this study is to explore the patterns of motor vehicle (MV) theft across
space, time, and targets.
Research questions
Specifically, this study will answer these questions: (a) What are the hot spots and hot
times of MV thefts in Negros Oriental? and (b) What are the hot products of MV theft in
the province?

Area of Frame of reference Unit of analysis Topic for investigation

interest
Criminal Environmental Person Decision-making of offenders
etiology criminology, Rational Streets/specific Hot spots/hot times of crime,
choice perspective address hot products
Criminal sociology Neighborhoods Social disorganization,
collective efficacy, fear of
crime, social disorder and
crime, and crime rates
Cities Social disorganization,
collective efficacy, and crime
rates
Groups (Gangs) Subculture, social learning
Persons Strain, social control,
parenting and crime

Now, review the how we came up with our problem statement on motor vehicle theft in
Negros Oriental. Table above shows the topic for investigation matrix for motor vehicle
theft.

Examples of exploratory/descriptive problem statement

Here are some examples of problem statement of published research articles in reputable
criminology and criminal justice journals. Note that they vary in style but still contain the
problem statements.

1. Sherman, Gartin & Buerger (1989: 33)

Hot spots of predatory crime: Routine activities and the criminology of place

“This article examines that premise. Using street addresses and intersections as an
operational definition of urban places, we assess police call data as a measure of place
 20

crime in Minneapolis. We describe the distribution of crimes by place and test for the
randomness of that distribution. We then consider the implications of the results for
further development of a routine activities criminology of place.”

2. Weisburd et al. (2004: 286)

Trajectories of crime at places: A longitudinal study of street segments in the city of
Seattle

“In this paper we use official crime data to examine the distribution of crime at street
segments in Seattle, Washington, over a 14-year time period”

B. Explanatory research

This time, let us turn our attention to problem statements of explanatory research.
Assume that you are interested in explaining police misconduct. Your broad question is:
Why do some police officers commit police misconduct or deviance? You might have a
few hunches on the answers to this question, and might think that termination of these
officers is the easiest and best solution. But as we know, prevention is better than cure.
Thus, it is important to determine in advance who are at risk of committing police
deviance. With this, police managers can plan for police misconduct prevention
programs. This is an interesting but complex question. As you read the literature
especially previous research on police misconduct, you will discover that previous
researchers used different frames of reference or perspectives in explaining why some are
involved in police deviance. Using one frame of reference, however, can provide focus
on your investigation, (and of course can lighten your load). See Table 3 for some frames
of reference on our current area of interest – police misconduct. (Note that this is not an
exhaustive list).

Suppose you choose the individual level frame of reference. Implications of this choice
are: You will study persons as units of analysis (police officers), and you will explain
police misconduct using police officer’s characteristics and attitudes. Suppose again that
after reading the literature, you discover that previous studies already focused on some
offender characteristics (e.g. age, education, training, and experience) and if you do the
same, your topic might be overdone. However, you have found that no study has yet used
self-control to explain why some officers commit more police deviance compared with
their counterparts. This is what Donner and Jennings (2014) successfully did: They used
self-control to explain different forms of police deviance for the first time.

Area of Frame of reference Unit of Topic for investigation

interest analysis
Police Organizational level Formal Supervision, management
misconduct organizations and police deviance
Situational level Persons Victim characteristics, victim-
officer relationship and
deviance
Individual level Persons Officer characteristics: age,
training, education,
experience, peers, self-control
 21

Neighborhood level Neighborhoods Social disorganization,

poverty, transition,
population heterogeneity,
collective efficacy
Topic for investigation matrix for police misconduct

Formulation of problem statement

In explanatory research, the researcher goes beyond describing a phenomenon, say crime,
and proceeds to explain it. For example, you can explain why some children commit
crimes through differential association. You can state that: Some children commit more
crimes because they have more delinquent peers. With this explanation, you are actually
talking about the relationship between the number of delinquent peers and the number of
crimes committed by children. Thus, in formulating research problems in explanatory
research, we ask questions entailing relationships. Some examples are presented below:

Do children who have more delinquent peers commit more crimes than children who
have fewer or no delinquent peers?
How do officer and neighborhood characteristics influence the relationships between
situational characteristics and police behavior? (Sun, Payne & Wu, 2008: 25)
Do male and female police officers differ in their levels of job satisfaction?

Since Donner and Jennings (2014) chose self-control to explain police deviance, they
needed to focus on the relationship between self- control and police deviance in
formulating their problem. They had only one research question as shown below:

“Acknowledging these issues, the current study contributes to the literature by exploring
the potential relationship between low self-control and police misconduct. Specifically,
the main objective of this study is to investigate whether a police officer’s level of self-
control is significantly related to his/her involvement in police misconduct. To this end,
and building upon the existing literature linking self-control to occupational deviance
(e.g., Gibson & Wright, 2001; Langton et al., 2006), the current study uses a sample of
1,935 police officers from the Philadelphia Police Department (PPD; Greene & Piquero,
2004) to examine one central research question: Does low self-control influence
individual-level police misconduct?” (Donner & Jennings, 2014: 204-205)
Problem statement
Specifically, the main objective of this study is to investigate whether a police officer’s
level of self-control is significantly related to his/her involvement in police misconduct.

Research question

Does low self-control influence individual level police misconduct?

Examples of explanatory research problem statement

You will learn faster when you study how researchers formulate their explanatory
problem statements. Here are some examples:
 22

1. Manis, Archbold & Hassell (2008: 510)

Exploring the impact of police officer education level on allegations of police misconduct

“The study presented in this paper responds to that call for additional research by
providing a contemporary look at the relationship between police officers’ education
levels and complaints of police misconduct. This study is unique because it utilizes both
formal and informal complaints filed against police officers working in a Midwestern
police agency. In addition, this study also examines the impact of the type of college
major (Criminal Justice (CJ) major compared with non-CJ majors) on complaints filed
against patrol officers.”

2. Hancock & Raeside (2009: 100)

Modeling factors central to recidivism: An investigation of sentence management in
Scottish Prison Service
“In this article, we focus on the role of prisoner management, particularly the processes
associated with sentence management (SM), in the reduction of recidivism.”

From research question to research hypothesis

Often, in problem statements of explanatory studies, research questions are stated as

hypothesis (Pyrczak & Bruce, 1992). According to Fraenkel and Wallen (2006: 46), “A
hypothesis is, simply put, a prediction of the possible outcomes of a study.” In other
words, a hypothesis is the tentative answer to a research question in an explanatory
research. It states the expected relationship between two variables (Dantzker & Hunter,
2006), for example delinquent peers and delinquent acts.

For instance, if your research question is:

Do children who have more delinquent peers commit more delinquent acts than children
who have fewer or no delinquent peers?

Your research hypothesis might be:

Children who have more delinquent peers tend to commit more delinquent acts than
children who have no or fewer delinquent peers.

Obviously, a hypothesis is just our tentative answer to our research question. It is this
statement that we want to provide evidence for their support using our research article.
Looking back to our earlier examples in criminal prosecution: It is like stating that Juan
dele Cruz might have committed the crime of murder against Juana dela Cruz. Then the
prosecutor will provide evidence to show some support for this claim.

There are, however, three types of hypothesis, namely: research hypothesis, null
hypothesis, and rival hypothesis (Dantzker & Hunter, 2006). A research hypothesis is
the tentative answer to the research question stating the relationship between two
 23

variables. Null hypothesis states that there is no relationship between two variables. This
hypothesis is important in statistical inference. Although usually researchers state the
research hypothesis instead of the null hypothesis in journal articles, it is the null
hypothesis that researchers try to disprove using empirical data. We can only say that our
data supports our research hypothesis when our data rejects the null hypothesis. (We will
further deal with this more in our lessons on statistics). Finally, a rival hypothesis is
arrived when the data reject both the research hypothesis and the null hypothesis and
provides an alternative answer to the research question.

MEASUREMENT OF VARIABLES

What are variables?

Fraenkel and Wallen (2006: 40) posit: “A variable is a concept – a noun that stands for
variation within a class of objects . . .” Gender, self-control, and crime are variables
because they vary across subcategories. These subcategories are called attributes.

Variables Attributes
Sex Female, Male
Age 0, 1, 2, 3, 4 years
College year level st nd rd th
IQ 15, 30, 65, 90, 100, 115, 120, 160, 195 . . .
Self-control level High 24 . . . . . . . . . . . . . 96 Low (reverse coded)
Number of crimes committed None, 1 crime, 2 crimes, 3 crimes . . .

On the table above in terms of rank: female and male cannot be ranked (no sex is higher
than the other), while 1 year old and 5 years old children can be ranked (a 5 year old child
is older than the 1 year old). In terms of distance: female and male have no distance
(since they have no rank), while a 5 year old child is 4 years older than a 1 year old (we
can measure the distance which is 4 years). Because of these differences, variables can
then be divided into types (Babbie, 2001; Bachman & Schutt, 2006; Dantzker & Hunter,
2006).

Nominal variables comprise variables whose attributes differ buthave no rank order, e.g.
gender (female, male), nationality (Filipino, Australian, Canadian), court decision (guilty,
not guilty), types of crime (murder, arson, rape), and crime commission (committed, not
committed). Male and female persons differ but have no rank (no one is quantitatively
higher than the other in terms of gender).

Ordinal variables comprise variables whose attributes differ, have ranks but have no
distance, e.g. college year level (1st, 2nd, 3rd, 4th year), crimes committed (None, 1-3
crimes, 4-5 crimes, 6-10 crimes), and social status (poor, middle income, rich family). In
terms of college year level, a 1st year student is different from a 4th year student, the 4th
year student has higher year level compared with the 1st year student; however, we
cannot really say how many more subjects this 4th year student has taken compared with
the 1st year student by just basing on the year level data.

Interval variables comprise variables whose attributes differ, have ranks and distance,
 24

but have no absolute zero point. Examples of these are IQ (20, 40, 65, 90, 100, 125),
temperature (-20ºC, -5ºC, - 1ºC, 10ºC, 20 ºC), and self-control (24, 25, 26 . . . 94, 95, 96).
In terms of IQ, a person with 50 IQ differs from a person with 100 IQ, 100 IQ is higher
than 50 IQ, 100 IQ is 50 more than 50 IQ, but we cannot say that 100 IQ is twice as large
as 50 IQ because there is no absolute zero. IQ, temperature, and self-control have no zero
points. The value of zero in these variables have no meaning; for example, 0 ºC does not
really mean there is no hotness or coldness; there is no 0 IQ; and it is impossible to find a
person with 0 self-control who cannot control himself at all times.

Ratio variables comprise variables whose attributes differ, have ranks, distance, and
absolute zero point. Examples are age (0, 1, 5, 10, 30 years old), number of crimes
committed (0 crime, 1 crime, 3 crimes), monthly income (None, P1, P100, P1,000,
P10,000, P100,000). These variables possess all the qualities of the three preceding
variables plus an absolute zero point. In terms of age, a 1 year old child is different from
a 5 year old child, the 5 year old child is older than the 1 year old, the 5 year old child is 4
years older than the 1 year old. Thus, they have a ratio: the 1 year old child is 5 years as
young as the 5 year old child (Ratio = 1:5).

Notice that these variables increase from variables with little

information (nominal) to the ones with the most information (ratio).

Variables, research questions and hypothesis

We will appreciate variables more if we look back on research questions, for example:
Does low self-control influence individual-level police misconduct? You might
remember that this was the central research question of Donner and Jennings (2014: 205).
Clearly, the variables in their research question are self-control and police misconduct
(see Table 6). Police officers differ in terms of self-control levels and police misconduct.
Some have low self-control; some have high. Some committed more misconduct, while
others did fewer or none at all. They wanted to know whether those with low self-control
commit more or less police misconduct. If so, they could say that low self-control
influences police deviant behavior. But what did they mean about self- control? How
about police misconduct? How could they measure self- control and police misconduct?
How would they find out if a police officer has high or low level of self-control and
police misconduct?

An illustration of the relationship between a research problem, question, hypothesis, and

variables in explanatory research.

Problem The main objective of this study is to investigate whether a police

statement officer’s level of self-control is significantly related to his/her
involvement in police misconduct.
Research Does low self-control influence individual-level police misconduct?
question
Hypothesis Police officers with low self-control levels commit more police
misconduct than police officers with high self-control levels.
 25

Variables Independent variable Dependent variable

Self-control Police misconduct

In the table above, we can see another typology of variables – independent and dependent
variables. This is the case when we deal with relationships between variables. One is the
independent variable; the other is the dependent variable. The independent variable is
the variable that is expected to affect the dependent variable, while the latter is the
variable that depends on or is affected by the independent variable (Babbie, 2001;
Fraenkel & Wallen, 2006; Maxfield & Babbie, 2005). In other words, we treat the
independent variable as the cause, while dependent variable is the effect. In our present
example, we hypothesized that self-control level is the cause, while police misconduct as
the effect. In sum, in explanatory research, we deal with cause and effect relationships.

How will I measure intangible variables in criminology?

At the outset of this lesson, we talked about the difference between natural sciences (e.g.
physics) and social sciences (e.g. criminology) in terms of the measurement of their
variables of interest. In natural sciences, the objects of study (e.g. chemicals, speed,
stones, and animals) are tangible, measurable, and predictable; but in criminology, we
might be puzzled by how to measure self-control, attitudes, job satisfaction, strain,
aggression, stress, which are not tangible – we cannot observe them directly. But this
could not preclude scientific studies in criminology.

The good news is that criminologists have devised indices and scales to measure their
independent variables and dependent variables.
This is hinged on the assumption that anything that exists can be measured directly or
indirectly (Babbie, 2001). Directly, we can determine the sex (and other directly
observable variables) of a person by just looking at his appearance or asking him to fill
out a paper asking his sex. Indirectly, we can measure intangible attitudes or traits by
observing the behavior of the person or asking him to fill out a set of questions or items
that tap the indicators of those attitudes or traits. For example, when you go to a party,
how would you know that a particular person is friendly? It is easy: just observe him. If
you see him hanging around a lot of friends, making jokes around a lot of people,
greeting a lot of people, not making trouble with other people, and acting as host of other
observable behaviors, you might say that he/she is friendly. You do not really measure
his/her being friendly using a measuring stick or a weighing scale, but you just infer this
characteristic from his/her behaviors.

Another way of determining whether a person is friendly or not is by letting him/her fill
out a questionnaire with items tapping friendly behaviors to which he/she would respond.
If most of his/her responses indicate being friendly, then it is safe to assume that he/she
friendly based on his/her answers.
 26

There are variables that are easily measurable by asking a question, for example age
(How old are you?). In contrast, using one question or item to measure an attitude or trait
can cause bias to our measurements. For example, it is a highly inaccurate measure when
we ask only one question to individual whether he frequently greets other persons or not,
and then conclude that his one affirmative or negative answer makes him/her as friendly
or not. We need to choose several items that fully and accurately represent a particular
attitude. These items that tap the indicators of a particular attitude or trait are chosen to
devise an index or a scale.

Although sometimes they are used interchangeably, an index is different from a scale
(Babbie, 2001). An index is a composite measure of a variable that simply assigns one
point to each item in the index and sum these points afterwards. If there are five items in
an index, and the respondent chooses the 4 items, the score of this respondent is 4. On the
other hand, a scale is a composite measure of a variable that places different weights on
the responses depending on their relative importance in measuring a variable. For
example, if there are five items in a scale, and the respondent chooses the 2
items/responses that weigh 3 points each, the score of this respondent is 6 points. None of
these measures, however, is superior to the other; deciding which is more appropriate
depends on the characteristics of the variable and its indicators.

Let us go back to our current variable of interest – self-control. After reading the
literature, we can find that researchers use indices and scales to measure this intangible
variable. Table 7 and 8 show the two types of measures used by previous researchers to
measure self-control.

Measuring self-control using a scale

The most commonly used scale to measure self-control in criminological studies is the
scale created by Grasmick et al. (1993). They used 24 items to measure self-control
following Gottfredson and Hirschi’s (1990) definition. These items were subdivided into
6 dimensions, and each dimension was measured by four items (see Table 7). Their scale
is an example of a Likert scale (one of the most popular scaling techniques in
criminology). There are, however, other useful scaling techniques for the social sciences.
See Babbie (2001) for an excellent overview of these techniques.

Self-control attitudinal scale

Self-control Attitudinal Scale
(Grasmick et al., 1993)
Dimensions Attitudinal Indicators Strongly Agree Disagree Strongly
agree somewhat somewhat disagree
(4) (3) (2) (1)
Impulsivity I don’t devote much thought ✓
and effort to preparing for the
future.
 27

I often do whatever brings me ✓

pleasure here and now, even
at the cost of some distant
goal.
I’m more concerned about ✓
what happens to me in the
short run than in the long
run.
I much prefer doing things ✓
that pay off right away rather
than in the future.
Simple tasks I frequently try to avoid things ✓
that I know will be difficult.
When things get complicated, ✓
I tend to quit or withdraw.
The things in life that are ✓
easiest to do bring me the
most pleasure.
I dislike really hard tasks that ✓

stretch my abilities to the

limit.
Risk seeking I like to test myself every now ✓
and then by doing something
a little risky.
Sometimes I will take a risk ✓
just for the fun of it.
I sometimes find it exciting to ✓
do things for which I might get
in trouble.
Excitement and adventure are ✓
more important to me than
security.
Physical If I had a choice, I would ✓
activities almost always rather do
something physical than
something mental.
I almost always feel better ✓
when I am on the move than
when I am sitting and
thinking.
I like to get out and do things ✓
more than I like to read or
contemplate ideas.
I seem to have more energy ✓
and a greater need for activity
than most other people my
age.
Self-centered I try to look out for myself ✓
first, even if it means making
things difficult for other
people.
 28

I’m not very sympathetic to ✓

other people when they are
having problems.
If things I do upset people, it’s ✓
their problem, not mine.
I will try to get the things I ✓
want even when I know it’s
causing problems for other
people.
Temper I lose my temper pretty easily. ✓
Often, when I’m angry at ✓
people I feel more like hurting
them than talking to them
about
When whyI amIreally
am angry.
angry, other ✓
people better stay away from
me.
When I have a serious ✓
disagreement with someone,
it’s usually hard for me to talk
about it without getting upset.
Table above not only presents self-control indicators but also check (✓) marks indicating
the degree of agreement with each of the 24 items. Assume that we ask a respondent to
fill out this scale. When the respondent checks “Strongly agree” to the item, “I lose my
temper pretty easily,” we can say that he /she indicates that he/she loses his temper pretty
easily. His/her score for this item is 4, indicating low self-control. Note that in this scale,
a high score indicates low self-control.

To create the overall score for this respondent on this scale we can perform different
techniques, for example summative technique, use of z- scores, or weighting each item
using their factor loadings in a factor analysis. For ease of illustration, let us use the
summative technique; the other two are statistically complex, and space limitation
restricts us to perform these techniques. In the summative technique, you only need to
add all the scores. So, adding all the scores in this example results to the overall score of
83. Note that using this technique a respondent’s possible score for this scale ranges from
24 to 96. If he/she checks all strongly disagree (1) responses, his/her overall score is 24;
if all are strongly agree (4) responses, the overall score is 96.

Measuring self-control using an index

Let us now illustrate the use of an index to measure a variable.

The data Donner and Jennings (2014) used in their study did not allow them to use the
Grasmick self-control scale to measure self-control of police officers. Instead, they used
an alternative: self-control behavioral index.

An example of self-control behavioral index.

Self-control Behavioral Index
(Donner & Jennings, 2014)
 29

Indicators No Yes
(0) (1)
Pennsylvania driver license ever been suspended ✓
Ever involved in a motor vehicle accident ✓
Received any traffic ticket in the past 5 years ✓
Ever been dismissed or fired from a job ✓
Ever behind on bills ✓
Ever divorced or separated ✓
Ever used marijuana ✓
Ever received a “D.I.” (Deception Indicated) on a polygraph ✓
examination
Ever received an unexcused absence from the police academy ✓

Donner and Jennings (2014) used 9 items to construct their self- control index. Note that
these items are behavioral indicators. Thus, their study implies that we can measure self-
control by observing or asking the police officers’ current and present behaviors. If they
acted in one of the nine items, that is an indicator of low self-control. This is similar to
our earlier example on measuring friendliness by observing behaviors.
Again, we will assume that a police officer has done some of the items in the index as
indicated in Table 8. He/she is scored 1 if he experienced one item; 9, if all items. Thus,
the possible score for this index ranges from 0 to 9. With the current example, this
hypothetical respondent scores 6 on this index.

How can I measure crime?

We have already addressed how to measure intangible variables.
Now, we will learn how to measure crime – the most important variable in criminology
and criminal justice. Over the years, criminologists have used several techniques to
measure crime and each has its own purpose and, of course, weaknesses (Maxfield &
Babbie, 2005; Sullivan & McGloin, 2014). In this subsection, we will focus on the ways
to measure crime. At the outset, the researcher must consider the following before
choosing a particular measure of crime (Maxfield & Babbie, 2001):

Purpose of the measurement

Consider whether your measurement is for monitoring crime levels, agency
accountability, or research.

Type of crime
You should decide what crime to include in your measure (Violent crime or property
crime? If it is a property crime, what property crime? Determine what crime is involved
(e.g. robbery, theft, motor vehicle theft).

Units of analysis
Finally, you should consider your unit of analysis (e.g. persons, organizations, groups, or
social artifacts). Once you have considered all these three important preferences, you can
then decide how to measure crime. Perhaps, when you are asked how many crimes are
there in your city/municipality, you would point your city’s police station as the source.
But, this is not the sole source of information on the number of crimes.
 30

Crimes can be measured in several ways – through reported crimes, victimization

surveys, and self-report surveys of offending (Maxfield & Babbie, 2005). Each of these
measures have its own limitations and appropriate use.

Reported crimes

Reported crimes are often used when researchers study places as

units of analysis – streets, neighborhoods, cities, provinces, or countries.

In the Philippines, the most commonly used information about the

frequency of crimes are reported crimes. These data come from two main
sources: citizen reports and police observations. They are collected and
stored by several law enforcement agencies in the country. But most of
these data can be found in the National Crime Reporting System (NCRS)
of the Philippine National Police (PNP). The NCRS is an incident-based
reporting system. That is, each record contains one incident that may
have one or more offender, victim, and crime or offense. Before using
these data, the analyst must first disaggregate the different crimes
committed in one incident.

One major limitation of this measure, however, is what criminologists call as the “dark
figure.” Dark figure pertains to citizen’s unreported crimes or crimes on which the police
failed to observe (Winslow & Zhang, 2008). Citizens tend to ignore light felonies and
offenses. Thus, most of these crimes are not reported to the police. Also, police officers
are not omnipresent; thus, they fail to observe all crimes, especially those highly
clandestine crimes and offenses such as drug crimes. Another limitation is the quality of
the data recorded. There had been no strict and clear protocols in the PNP on what
incidents were to be recorded in the NCRS. Moreover, the PNP itself discovered that
some chiefs of police distort their crime rates just to portray good image that less crimes
occur in their jurisdictions (Elona, 2013). Fortunately, as of the time of this writing, the
PNP is auditing and upgrading its NCRS in all police stations around the country to
improve the recording of crime incidents.

The strength of the PNP NCRS, however, lies in some reported serious crimes, for
example homicide and murder for crimes against persons and motor vehicle theft for
crimes against property. A homicide is likely reported to the police because of its
inherent seriousness. While, citizens cannot afford to just keep quiet when their prized
motor vehicles are stolen.

Reported crimes are often used by the authorities to describe the crime situation of the
country and of other geographical units from Philippine regions to barangays. They do
this through crime rates. A crime rate is a relative measure of crime risk per a
geographic units’ total population. It is computed using the following formula
(Champion, 1993):
 31

Crime rate = Number of crimes

Population × 100,000.

Suppose that Dumaguete City, with an approximate population of

125,000, is our geographical unit of interest, and you wish to compute its homicide rate in
2014. In that year, there were 90 homicide offenses recorded in the Dumaguete City PNP
NCRS. The computation of its homicide rate in 2014 is as follows:

Homicide rate = 90 Homicide cases/125,000 x 100,000

Homicide rate = 0.00072 × 100,000

Homicide rate = 72 Homicide cases per 100,000 population

Interpretation:
In 2014, there were 72 homicide cases per 100,000 persons. In other
words, approximately 72 persons per 100,000 population were at risk of
being killed through homicide.

Note that the figure, “100,000”, is constant. This means “per 100,000 thousand
population”. You can, however, use other figures like 1,000, 10,000, or 1,000,000. The
choice among these constant figures depends on the frequency of the crime and the size
of population. If the crime is rare, like homicide in Dumaguete City, it is appropriate to
use 100,000; if the number of homicide cases is large, we could use 1,000. However, the
figure, “100,000”, is the most frequently used constant to calculate crime rates. We use
these constants in computing crime rates to reduce the bias associated with the number of
population. For instance, City A recorded 100 homicide; while City B, 130. You might
then conclude that A is more violent than B when not taking into account each city’s
population which is: City A = 80,000 population and
City B = 200,000 population. If you load these figures into our formula, this will be the
crime rates for each city: City A = 125 per 100,000 persons and City B = 65 per 100,000
persons. It turns out that City A is in fact more violent than B.

Victimization surveys

We pointed out earlier that reported crimes undercount less serious offenses that citizens
are reluctant to report and those offenses that the police fails to observe. Victimization
surveys were devised to overcome these limitations. These surveys take a sample of the
population and ask this sample of individuals their crime victimization experiences. With
this, policy makers are guided with a more comprehensive measure of crime. Also,
victimization surveys are often used when criminologists study persons (victims) as units
of analysis to test theories of victimization. However, there are times when victimization
experiences are aggregated to larger units like cities and countries (see e.g., Van Dijk,
Van Kesteren & Smit, 2007). In the Philippines, however, this is seldom done, and there
is no current national crime victimization survey in the country. Some of the most
popular large scale victimization surveys in other countries include the National Crime
Victimization Survey (NCVS) in the US (Maxfield & Babbie, 2005) and the International
 32

Crime Victimization Survey (ICVS) in some industrialized countries in Europe, America,

Asia, Australia, and Africa (Van Dijk, Van Kesteren & Smit, 2007). These surveys were
aimed at monitoring victimization rates in several countries as an alternative to crime
rates. Aside from this, some researchers also conduct small-scale surveys in cities and
schools for monitoring and research purposes, for example Mustaine and Tewksbury
(1998) who surveyed some US college students found that victimization risk was related
to the characteristics of neighborhoods where they went and what they did there.

Victimization surveys are also limited in some aspects (Maxfield & Babbie, 2005).
Victims of consummated homicide and murder could not be included in the survey. (Of
course, they are already dead!). Likewise, victimless crimes (e.g. drug crimes,
prostitution, and gambling) could not be captured by these surveys. You cannot ask a
person if he has been a victim of victimless crime, say, prostitution.

Sample of questions:

The following questions were taken from the ICVS 2004-2005 questionnaire (Van Dijk,
Van Kesteren & Smit, 2007). This is an example on how to ask respondents about their
victimization experiences.

Q100. First of all, you mentioned the theft of a car. When did
this happen? Was this .

This year
Last year, in 2004
Before then
Don’t know/can’t remember

Q101. How often did it happen in 2004?

Once
Twice
Three times
Four times
Five times or more
Don’t know

Self-report surveys of offending

To overcome the limitations associated with reported crimes and victimization surveys,
criminologists devised self-report surveys of offending. These surveys ask a sample from
a population to report their past offending behaviors through filling out a questionnaire.
This measure is commonly used by criminologists who study persons (offenders) as units
of analysis. According to Maxfield and Babbie (2005:159), these surveys “are probably
the best method for trying to measure certain crimes that are poorly represented by other
techniques.” In addition, self-report survey is the most common method to collect data
 33

used by criminologists to explain criminal behavior (O’Brien, 1985, quoted by Maxfield

& Babbie, 2005). This is partly because most theories in crime explain variation of
criminal behaviors across persons as units of analysis.

Remember that reported crimes undercount less serious crimes and police unobserved
offenses. While, victimization surveys exclude victimless crimes, self-report survey is the
method that could overcome these limitations. However, the nature of this method raises
some important concerns (Maxfield & Babbie, 2005). It could be that offenders are
reluctant to report their past crimes or offenses because of social desirability issues. Or,
offenders may suffer from recall issues, especially the chronic offenders who have
committed large number of offenses.

These issues seem apparent from the survey of the Dangerous Drugs Board (DDB; 2008).
The DDB conducted a national survey to measure the extent of drug abuse in the
Philippines in 2008. Surprisingly, there were no respondent in Region I, IV-B, and XII
who said that he/she was a current drug user at the time of the survey. There are three
possible reasons for this: (a) there were really no current drug users in those regions; (b)
the research method was defective, thus, the surveyors did not find respondents who were
current users, and (c) the users did not tell the truth for social desirability reasons. But, as
it stands, there were those who responded that they had tried using drugs
– 21 respondents in Region I, 46 in Region IV-B, and 21 in Region XII. It is quite
possible that some of these respondents were reluctant to report that they were current
drug users but just reported that they had tried using drugs and had stopped.

Sample questions:

Asking respondents to report their past offending behaviors is similar to asking them their
past victimization experiences. The following questions were taken from 2011 Global
School-Based Student Health Survey questionnaire.

Q26. During the past 12 months, how many times did you actually attempt suicide?

0 times
1 time
2 or 3 times
4 or 5 times
6 or more times

Q29. During the past 30 days, on how many days did you smoke cigarettes?

0 days
1 or 2 days
to 5 days
6 to 9 days
10 to 19 days
20 to 29 days
 34

All 30 days

How do I construct my questionnaire?

Knowing how to measure your key variables is an important step in the research process.
At this time, we will combine what we have just learned. (A curious reader is referred to
Patten, 1998, for an excellent practical guide for questionnaire construction).

Let us assume that we are to conduct a survey on adolescent delinquency of college

students, and we want to explain delinquency using the self-control theory of Gottfredson
and Hirschi (1990). We do this by having age, gender, and household monthly income as
control variables. That is, when we regress delinquency on self-control, we hold these
three variables constant. You should note that this is a basic and simple additive model of
explain delinquency. We can add other variables of interest in criminology such as
parental supervision, number of delinquent peers, attachment and commitment to
conventional people and activities. We can also use some mediating and moderating
variables.

SURVEY SAMPLING

How can we describe a population?

Before we answer this question let us first discuss how we describe an individual. For
instance, you are to describe PO3 Juana dela Cruz.
How will you describe her? Perhaps you will say: PO3 dela Cruz weighs
50 kg; she stands at approximately 150 cm tall; she has three children; she has arrested an
average of 6 suspects in the last three years; she has high self-control; and she is against
the imposition of death penalty to serious offenders. My next question is: How did you
know about these descriptions? Perhaps, you have looked into her Personal Data Sheet,
or interviewed her, or let her answer a questionnaire. Describing one person, therefore, is
somewhat easy.

But, how about if you are to describe not one but all 130,000 police officers, or all
100,000,000 million Filipinos in the country, or all 10,000 university students of Negros
Oriental in State University Main Campus I. If you were the government of the
Philippines (with hundreds of millions of funds!) intended for the decennial census, it
will not certainly bother you. Unfortunately, we do not have the luxury of funds the
government has. The good news, however, is that past researchers have devised and
established a method that can be used to describe a population by just collecting
information from a subset of that population. This method is called survey

A survey is a method of collecting data from a subset of a population to describe such

population. In the Philippines, we often hear from the news the results of national surveys
by the Social Weather Station (SWS), Pulse Asia, and the National Statistics Office.
Popular topics include trust ratings of government officials, polls and exit polls surveys
during elections, and family-related topics in the National Household Survey. But, how
 35

can, for example, these agencies precisely describe the trust ratings of government
officials by just surveying about a thousand persons out of the 100,000 people in the
Philippines? They can do this by using probability sampling techniques. But, before
delving on these techniques, let us first discuss how researchers calculate the sufficient
sample sizes for their surveys.

How to calculate the sample size?

This subsection introduces two common methods of determining the sample size –
interval estimation and power analysis. Interval estimation is usually used if one is
interested to describe the population of interest according to its characteristics like
average age, number of family numbers, percentage of the population who are in favor or
against death penalty, average fear of crime, or percentage of the population that have
ever used dangerous drugs. In estimating the sufficient number of units in a sample, the
researcher specifies the number of population, confidence level, and confidence interval.
(We will delve on these concepts later).

Another method in determining the sample size is one that has gained much interest. This
is the use of power analysis (Cohen, 1992).
Power analysis is usually used when one considers and attempts to avoid the errors in
hypothesis testing (i.e. Type I and Type II errors). To elaborate on these errors, let us go
back to our previous example of explaining youth delinquency. Remember that we
treated self-control, age, gender, and family income as independent variables and
delinquency as dependent variable. We wished to determine the effects of these four
independent variables on the dependent variable. Type I error occurs if, for instance, we
have statistically found that self-control or the other variables have affected delinquency
when in fact they have not. On the other hand, Type II error occurs if we have found
that self-control or the other variables have not affected delinquency when in fact they
have. To avoid these errors in hypothesis testing, we must consider some parameters such
as significance level, power of the hypothesis test, effect size, and number of independent
variables.

Sample Size Determination

The calculation of the sufficient sample size is somewhat mathematically complex. The
good news is: There are point-and-click program and software that we can use to perform
this calculation. We will demonstrate how to use some of these programs in this
subsection.

Interval estimation

For interval estimation, several programs are available online, for example the Sample
Size Calculator by Creative Research Systems (2015). We will demonstrate how to use
this program in the following tutorial. Assume that you are to conduct a survey on police
officers’ attitudes toward some aspects of the administration and supervision of the
Philippine National Police by the National Police Commission. Specifically, you wish to
 36

determine the percentage of the police officers in, say, Region VII whether they are in
favor or not of the deputation of local chief executives (i.e. mayors and governors) to
administer and supervise local police stations. So, the result would be in terms of
percentage (e.g. 40%, 50%, 60%, or 70% against or in favor). In short, we want to
estimate the percentage or proportion of police officers who are in in Region VII through
a survey of a sample of these officers. Suppose that there are 10,000 police officers in the
said region. Using the calculator to determine the sample size out of 10,000 police
officers is relatively easy. Here are the steps:

Step 1. Just type in your web browser (e.g. Google or Yahoo) “sample size
calculator,” and click the Sample Size Calculator. For quick access, you
can, however, visit this address:
http://www.surveysystem.com/sscalc.htm#two

You will then be directed to a web page containing the interface shown below.

Step 2. The next thing you will do is to specify the parameters the calculator needs – the
confidence level, confidence interval, and the number of the total population. For
example: Confidence level = 99%, Confidence interval = 2, and Population = 10,000.
Then, click Calculate. You will then obtain the sample size needed as shown below.

For instance, after the randomized survey of 2,938 police officers you obtained these
results: 45% in favor and 55% against. You will then arrive at the following
interpretation.
 37

Interpretation: We are 99% confident that our randomized survey of the

2,938 police officers captures the true percentage of officers who are in favor or against
the deputation of local chief executives to administer and supervise local police stations.
The true population proportion of police officer who are pro-deputation is within plus or
minus 2% from 45%. That is, the true percentage of all 10,000 police officers who are in
favor of the deputation is from 43% to 47% based on the results of our survey.

Note: Confidence level means that if we repeat choosing 100 samples,

99% of these samples contain the true percentage value of the population. Note that you
can also use other online sample calculator such as that of the Australian National
Statistical Service (2015).

What are the types of sampling?

There are two types of sampling which you can use in selecting the sample in your
research project – probability and non-probability sampling (Babbie, 2001; Kothari,
2004; Maxfield & Babbie, 2005). Probability sampling is a sampling design that gives
equal probability or chance to each item or individual in the population of being selected
in the study. On the other hand, non-probability sampling is a sampling design that
gives unequal chance to each item or individual in the population of being selected in the
study. Each of these types, however, have their own strengths and weaknesses.
Probability sampling is ideal when the researcher wishes to generalize his findings to a
population of interest. Often, it is used for quantitative surveys. Whereas, non-
probability sampling is ideal when the researcher wishes to study in- depth a particular
phenomenon. Qualitative studies take advantage of the ease of using non-probability
sampling.

What are probability sampling methods?

In probability sampling, all elements (or items or units) in the population are afforded
with equal probability of being selected for a study, say a survey. This method is
important in order to generate a sample that is representative of the population of interest.
Going back to our previous example of determining the perceptions of 10,000 police
officers, we can ensure that the perceptions of the sample of, say, only 800, 1000, or 2000
officers, approximate the perceptions of all the 10,000 police officers, if we have
generated a representative sample from the population. That is, the characteristics of our
sample approximate the characteristics of the population. Because of several constraints
in generating a representative sample, there exists several methods in conducting
probability sampling such as simple random sampling, systematic sampling, stratified
sampling, multistage cluster sampling (Babbie, 2001; Kish, 1965). Note, however, that
these methods can be combined depending on the circumstances (Babbie, 2001). Note
again that different sampling situations demand different sampling methods.

Simple random sampling

 38

From its name, simple random sampling is the simplest probability sampling method. It is
relatively easy. One only needs to have a population frame, which is the complete list or
at least the most complete list of the desired population, say, police officers in Region,
and a random numbers table to facilitate the random selection of police officers to
whom a survey is conducted. We will illustrate this sampling method in the following
steps:

Step 1. Assume that we need to survey 800 police officers out of all
10,000 police officers in Region VII. If we use simple random sampling,
we need to have the (alphabetical and numbered) list of all police officers
in the region and a random numbers table. To acquire the list, you can ask the PNP
Region VII Headquarters. To acquire a random numbers table, you can check research
and statistics methods books or download from the internet the RAND Corporation’s
table of random numbers.

Step 2. We will randomize the selection of the first random number. Find any Philippine
paper money bill in your wallet. Mine is a PhP 20.00 bill with serial number, AY680511.
The first number in this serial number is 6. So we will start with the 6th column with the
1st random number of 54876 (see Figure 5). Remember that we need to choose 800
officers out of 10,000. Since the total number of police officers is 10,000, we will use all
5 digits of all numbers in the 6th column as reference. (Note that if the total number of
population is smaller than 10,000, say 1,000, we will only use the 1st four digits of 54876
which is 5487.) Going back to our example, the first random number (i.e. 54876) meets
our 5-digit criterion; however, there is no 54876th respondent in our list of officers. There
are only up to 10,000th officer, so we will skip this number and proceed to the one below
it which is 24,037. Still this number does not correspond to any of our respondent. The
next number is 02560. Now, if we would go to our list of police officers, we could find a
2,560th officer out of 10,000. So we pick this officer as our first respondent. Finding
another respondent follows the same process until we arrive with a sample of 800
respondents. However, if you reach the bottom of the random numbers table and you
have not yet completed the required 800 respondents, you can go back to the 8th column
corresponding to the 2nd number of the serial number (i.e. AY680511) of our PhP 20.00
money bill, and follow the same selection process.
 39

Systematic sampling

Systematic sampling is a popular alternative to simple random sampling. It affords the

ease and efficiency in selecting the elements of the sample, and is sometimes called as
“pseudo-random” selection (Kish, 1965: 113). It consists of these steps:

(1) conducting a random start, and

(2) selecting every kth element of the population after Step 1.

These steps also afford every element in the population to have an equal probability of
being selected for the study. One important tool in this selection process is the sample
interval (Babbie, 2001; Kish, 1965). The selection interval is the distance of next
element from the previously selected element; and thus, determines what elements are
chosen systematically. We will illustrate the steps in systematic sampling below:

Step 1. Determine the total number of population elements. In this example, we assume
that we have a 10,000 total population of police offices in Region VII.

Step 2. Determine the total number of sample elements. We will assume that we need, for
this example, 1,000 police officers.

Step 3. Find the selection interval by using 1,000 to divide 10,000, as shown below:
Selection interval = 10,000/1,000 = 10
 40

As shown, our selection interval is 10. Thus, every 10th element of the population will be
selected.

Step 4. Create a random start among the first 10 police officers. To do this, you can
follow the process we used in our simple random sample illustration. If we have
randomly selected the 2nd column of the random numbers table above, the number 05
(i.e. the 5th police officer in our population frame) will be our first element in the sample.

Step 5. Select the next police officer by just adding 10 to 5 (see Figure 6).
The sum would be 15; thus, the 15th officer in our list will be the next sample element.
Do this process of addition until you have already selected 1,000 officers.

As shown above, systematic sampling is somewhat simpler than the simple random
sample. However, systematic sampling is vulnerable to systematic arrangement of the
population elements. For instance, when one’s selection interval is 10, the random start
element is 1, and every 10th element in the list of police officers is systematically
assigned as a squad leader, then the respondents of the survey would all be squad leaders
excluding other officers from other ranks. This then creates bias in your survey results.
Nevertheless, if the population elements are not systematically arranged, systematic
sampling is an ideal alternative to simple random sample (Kish, 1965); and no evidence
to support which of the two is better exists (Babbie, 2001).

Stratified sampling

There are times when simple random sampling or systematic sampling may not be
enough to achieve representativeness of the sample, especially when the population is
naturally divided into subpopulations or strata. For example, police officers in the
Philippines are naturally composed of different subpopulations in terms of ranks (i.e.
Police Officer level, Senior Police Officer level, Police Inspector level, Police
Superintendent level, Police Director level), gender (i.e. male or female), work stations
(i.e. municipal police station, city police station/office, provincial police office, regional
police office, national headquarters), to name a few. To achieve variance between and
homogeneity within strata, the researcher can use stratified random sampling which
 41

follows the following steps: (1) divide the population into strata, and (2) choose sample
elements from each stratum using either simple random sampling or systematic sampling.

Note that the size of the sample elements in a stratum must be relatively proportional to
the size of the population of elements in that stratum. For example, among all police
officers in the Philippines, Senior Police Officers constitute 30% of all officers; thus, in
the sample, there must also be 30% Senior Police Officers. We will illustrate a simple
process of arriving at a stratified sample in the following steps:

Step 1. Assume again that we want to conduct a survey of police officers in Region VII
with 10,000 police officers, comprising 7,000 Police Officers
(70%), 2,500 Senior Police Officers (25%), 450 Police Inspectors (4.5%), and 50 Police
Superintendents (0.5%). Then, we have arrived with a desired sample of 1,000 officers.

Step 2. Let us make sure that the proportion of ranks of the1,000 sample of police
officers must be relatively proportionate to the proportion of ranks of the population of
10,000 police officers in the region. To do this, we will perform simple mathematics by
multiplying 1,000 with the different percentages of ranks in the populations. For instance,
Senior Police Officers comprise 25% of all 10,000 police officers; thus, we shall multiply
1,000 with 25% or 0.25, and derive 250Senior Police Officers, as illustrated below:

0.70 = 700 Police Officers

0.25 = 250 Senior Police Officers
0.045 = 45 Police Inspectors
0.005 = 5 Police Superintendents

Step 3. Perform either simple random sampling or systematic sampling to choose, say,
250 Senior Police Officers from all 2,500 Senior Police Officers in the region. For ease
and simplicity, you may choose systematic sampling. Just arrange the list of all police
officers in the region by rank so that, for example, the first in the list are all Police
Officers (arranged alphabetically or shuffled randomly), followed by Senior Police
Officers and so on. Then use the sample interval of 10 (10,000 ÷ 1,000), and choose a
random start among the first 10 officers in the list by using the random numbers table.

Multistage cluster sampling

What if there is no available list of the population elements, say, all names of people
living in a city, province, region, or country, what methods exist to help the researcher?
Or what if using simple random sampling, systematic sampling, and stratified sampling
methods are too expensive and complicated for a researcher? To answer these concerns,
researchers devised a less accurate but a more flexible and efficient way of probability
sampling compared with other probability sampling we have discussed so far.
Researchers usually use multistage cluster sampling to arrive at a representative sample
of a population which have not been compiled in a list, for example all residents in
Dumaguete City (Babbie, 2001; Kish, 1965). Previous probability sampling methods we
have discussed are conducted in one-stage only (i.e. we only select sample units once). In
 42

multistage cluster sampling, we select sample units twice, thrice, or more at different
stages or units of analysis.

Going back to our lessons on units of analysis, a researcher has the freedom what unit of
analysis he/she is to study, whether at the regions, provinces, municipalities/cities,
barangays, barangay puroks, dwellings/houses, residents (see Figure 7). In multistage
cluster sampling, he/chooses sampling units at different stages. For example, if he/she
wishes to survey some attitudes of all residents in the Province of Negros Oriental, he/she
is to:

Step 1 Divide Negros Oriental into clusters, for example municipalities/cities. Use either
of the probability sampling methods we have discussed (whichever is most ideal) to
select the cities/municipalities to survey.

Step 3. Divide these cities/municipalities into clusters of barangays, and use again either
of the probability sampling methods we have discussed (whichever is most ideal) to
select what barangays to survey.

Step 4. Divide these barangays into barangay puroks (villages) and select
what villages to study.

Step 5. List the houses/dwellings in the selected puroks/villages, and use simple random
sampling or systematic sampling to select the dwellings.

Step 6. Upon the survey, randomly select the occupant of the house/dwelling who is
going to be the respondent of the study.

What are nonprobability sampling methods?

There are times, however, that probability sampling methods are inefficient in answering
some research problems, especially research problems of qualitative research methods we
have discussed in Lesson 3. In this subsection, we shall learn nonprobability sampling
methods that are used by qualitative researchers. These methods, however, do not afford
equal chance to all elements of the population of being selected in the study. Thus,
research results of studies using these methods cannot be generalized to the population of
interest. But, nevertheless, the results provide more in-depth examination of the
phenomenon studied compared with quantitative studies that use probability sampling
methods. Nonprobability sampling methods include convenience sampling, purposive
sampling, snowball sampling, and quota sampling (Babbie, 2001; Cohen, Manion &
Morrison, 2007).

Convenience sampling

In convenience sampling, the researcher selects respondents who are nearest to him or
available to him; and, thus, affords him with the ease of selecting respondents but at the
risk of not achieving representativeness of the sample. Results from studies that use
 43

convenience sampling cannot be generalized to the whole population of interest. One

example is when we wish to determine the perception of the students on a particular city
ordinance, but we only interview those students who are currently available at the time of
our survey, say, those who are present at the school canteen. Another example is when
we want to study the perceptions of police officers to any policing topic and we only
survey those officers who are around when we arrive at the police station in the morning.
The former example exclude those students who are not in the canteen where we conduct
our interviews. The latter example excludes police officers who are in the other shifts, for
example, night shifts. Results of these studies are biased and questionable if we attempt
to generalize all students or all police officers. Nevertheless, if our purpose is not to
generalize but to answer some qualitative research questions or to provide a pilot study
then convenience sampling may be the most efficient.

Purposive sampling

Often, some researchers select respondents based on their judgment on the purpose of
their study. They use purposive sampling when they want to study those respondents who
are the most knowledgeable or those who have more experiences with a particular
phenomenon of interest. For example, researchers who want to determine the negative
experiences of battered wives may interview those wives listed in the police station as
victims of gendered violence. The results of this method, however, exclude the
experiences of battered wives who did not report their victimization to the police station.

Snowball sampling

Sometimes, however, target respondents are hard to access using the nonprobability
sampling methods described above. This is especially true when the phenomenon of
interest is highly sensitive and secretive, for example, drug abuse or drug pushing. Even
though some drug users and traffickers are incarcerated, research results based on
information from these respondents are limited since there are still those who are
unincarcerated who may have different answers to our research questions. However,
population frame of all current unincarcerated drug traffickers or drug abusers is
unavailable. But it is important for us to learn some things about these individuals for
theoretical and policy- making purposes.

One remedy to this problem is the use of snowball sampling. With this sampling method,
the researcher first approaches a known respondent who has experienced the researcher’s
phenomenon of interest, say, drug trafficking. The researcher then builds rapport with
this drug trafficker and make him as the study’s first informant. After collecting
information from this respondent, the researcher then asks him to point out another drug
trafficker he knows. The researcher then interviews this second drug trafficker and asks
him another drug trafficker to interview. The process continues until the researcher
arrives at his desired number of respondents or no other trafficker is pointed out.

Quota sampling
 44

This sampling method is considered as the nonprobability sampling’s equivalent of

probability sampling’s stratified sampling method (Bailey, 1978 cited in Cohen, Manion
& Morrison, 2007). The researcher attempts to obtain samples representing the
proportion of population elements in strata. One major difference, however, between
quota sampling and stratified sampling is the method of sample selection. Quota
sampling does not use probability sampling method (e.g. simple random sampling and
systematic sampling) in selecting sample elements; instead it uses convenience sampling
or purposive sampling. Thus, to perform quota sampling, just follow Step 1 and Step 2 of
the
steps (except Step 3) in stratified sampling method illustrated above.
Then, perform convenience sampling or purposive sampling.

DATA ANALYSIS

Statistical software

Numerous statistical programs (both free and proprietary) are available that could make
our calculations easier and faster. Of these programs, the most easy-to-use are point-and-
click programs that work like Microsoft Word and Excel include SPSS, Minitab, and
some free software which are available for download from the internet, for example the
program Tanagra. Perhaps, the most popular statistical software in the social and
behavioral sciences is the SPSS. However, there are some programs that are less complex
for a beginner like you.

1. The t-test
Data analysis tutorial 5: Calculating the difference between two means.
For example, we want to determine who between male and female US students, engage
more frequently in binge drinking.
The research questions can be phrased as:
 What is the difference between male and female students in terms
of engagement in binge drinking?

2. The Analysis of Variance (ANOVA)

Calculating the difference of means among at least three groups. With t-test, we tested the
difference between two groups. Now, what if we want to test the differences among, say,
three or four groups? This time we will use Analysis of Variance or ANOVA (an
extension of t-test) to do such comparisons. We will use it to determine whether different
year levels have different binge drinking levels. That is, who have higher binge drinking
levels, 1st year US students, 2nd year, 3rd year, or 4th year? The research question can be
phrased as follows:
 What is the difference among college year levels in terms of
engagement in binge drinking?

3. Chi-square
Determining the association between two nominal variables. In t-test and ANOVA, we
 45

measured the relationship between a nominal variable (e.g. sex) and a continuous one
(i.e. binge drinking means). This time we will examine whether two nominal variables
are associated. Specifically, we will determine the relationship between sex
(Male/Female) and membership in fraternities/sororities (Member/Not member). That is,
who between male students and female students
have higher proportion of membership in fraternities/sororities?
 What is the association between sex and membership in fraternities?

4. Pearson correlation
Calculating the relationship between two continuous variables. What if we want to
determine the relationship between two continuous variables? We cannot use the
preceding
bivariate statistics. We will use instead Pearson correlation which calculates the extent
and direction of the relationship between two continuous variables. Here is the research
question we will attempt to answer this time.
 What is the relationship between favorable attitudes toward parties
and engagement in binge drinking?
 Is there a significant relationship between favorable attitudes
toward parties and engagement in binge drinking?

5. Multiple linear regression

Determining whether a group of independent variables (nominal/interval/ratio) predict or
explain a dependent variable (ratio/interval). Here is the research question we will
attempt to answer using multiple linear regression:

 Can binge drinking be significantly predicted using age, gender, membership in

frat/sor, and attitudes toward athletics, religion, and parties?

6. Binary logistic regression

Data analysis tutorial # 10. Determining the relationship between a group of

predictors/independent variables (nominal, ordinal, ratio and interval) and a nominal
(binary) dependent variable. This time we will attempt to answer this research question:

 Can membership in frat/sor be significantly predicted using age, gender, and attitudes
toward athletics, religion, and parties?

REFERENCING

Books

Books/EBooks
One author/editor (editor identifiable by surname followed by Ed.).
− Paraphrasing – (Author’s last name, year)
e.g.
Average count of words read in one minute ranges between 250 and 300 words (Kellogg,
 46

2007).
− Quotation – (Author’s last name, year, page number)
e.g.
“Normal reading speed is about 250 to 300 words per minute” (Kellogg, 2007, p. 243).

Two authors/editors (editor identifiable by surname followed by Eds.)

− Paraphrasing – (Author’s last name & Author’s last name, year).
− Quotation – (Author’s last name & Author’s last name, year, page number).

Three + authors/editors (editor identifiable by surname followed by Eds.).

− Paraphrasing – (Author’s last name, Author’s last name & Author’s last name,
year).
− Quotation – (Author’s last name, Author’s last name & Author’s last name,
year, page number)

Article/chapter in an edited book

− Paraphrasing - (Article/chapter author/authors last name/s, year)
e.g.
An essential and determining factor in modern day market is environmental protection
(Fischer,
2017).
− Quotation – (Article/chapter author/authors last name/s, year, page number)
e.g.
“Environmental protection has become an essential issue and a determining factor in
today’s
market” (Fischer, 2017, p.75).

Books with no author

− Paraphrasing - (Book title {shortened}, year).
e.g.
Springhouse Drug Guide (2007) recommends investigating the patents full medical
history.
− Quotation – (Book title {shortened}, year, page number).

Ex.

“When taking a history, investigate the patient’s allergies, use of drugs… medical
history, lifestyle
and beliefs and socioeconomic status” (Springhouse Drug Guide, 2007, p. 1).

Journals

One author
− Paraphrasing – (Author’s last name, year)
e.g.
Instances of climate variation are documented throughout history (Brander, 2010).
 47

− Quotation – (Author’s last name, year, page number)

e.g.
Climate change and climate variability have occurred throughout history and natural
systems have
developed a capacity to adapt, which will help them to mitigate the impact of future
changes
(Brander, 2010, p. 390).

Government Documents

APA handles government publications as books, technical reports or brochures

− Paraphrasing – (Author’s last name or Name of government organization,
year)
e.g.
In June 2017, the Philippine peso weakened below PUP/US$50.00 mark (World Bank,
2017).
− Quotation – (Author’s last name or Name of government organization, year,
page number)
e.g.
“The Philippine peso weakened in June and once again crossed the PUP/US$50.00 mark”
(World
Bank, 2017, p. 1).

Magazines and Newspapers

− Paraphrasing – (Author’s last name, year)

e.g.
Master weavers may take on apprentices as an alternative source of income (Aquino
Tugade, 2016).
− Quotation – (Author’s last name, year, page number)
e.g.
“Although it is comforting to think that cultural transmission is the primary reason for
teaching, the
monetary incentive motivates most masters to take a new weaver under their wing”
(Aquino
Tugade, 2016, p. 88).

Webpages

Basic webpages
− Paraphrasing – (Author’s last name, year).
− Quotation – (Author’s last name, year and paragraph number if available).

Webpage with no author

− Paraphrasing - (Webpage title {shortened}, year)
e.g.
 48

(LGBTQ asylum seekers face prejudice, 2017).

Quotation - (Webpage title {shortened}, year, page or paragraph number if
available ).

Multimedia

Films and recordings

− Paraphrasing – (Producers last name & Directors last name, year)
In four million years, the earth has not experienced average global temperatures that
scientists
predict may occur by the end of this century (Dicaprio & Stevens, 2016).
− Quotation – (Producers last name & Directors last name, year)
e.g.
“We are on average moving towards four degree warming this century. And we haven’t
been in a
four degree warmer world for the past four million years” (Dicaprio & Stevens, 2016).

Youtube
− Paraphrasing – (Authors last name {or Screen Name}, year)
e.g.
In developing countries, the annual rate of deaths in mothers and newborns caused by
limited access
to basic hygiene exceeds one million persons (Ted, 2017).
− Quotation – (Authors last name {or Screen Name}, year)
e.g.
“Over one million mothers and babies die every single year in the developing world, only
because of
lack of access to basic cleanliness while giving birth to their babies” (Ted, 2017).

References

Braga, A., Papachristos, A. & Hureau, D. (2012). The effects of hot spots
policing on crime: An update systematic review and meta-analysis.
Justice Quarterly, 1-31.
Booth, W., Colomb, G. & Williams, J. (2003). The craft of research.
Chicago: The University of Chicago.
Dantzker, M.L. & Hunter, R. (2006). Research methods for criminology
and criminal justice. Sudbury, MA: Jones and Bartlett Publishers, Inc.

Donner, C. & Jennings, W. G. (2014). Low self-control and police

deviance: Applying Gottfredson and Hirschi’s general theory to officer
misconduct. Police Quarterly, 17(3): 203-225.

Fraenkel, J. & Wallen, N. (2006). How to design and evaluate research in

education. New York, NY: McGraw-Hill.
 49

Gottfredson, M.R. & Hirschi, T. (1990). A general theory of crime.

Stanford, CA: Stanford University Press.

Kelling, G.L., Pate, A., Diekman, D. & Brown, C. (1974). The Kansas City
Preventive Patrol Experiment: Summary Version. Washington, D.C.: Police
Foundation
Maxfield, M. & Babbie, E. (2005). Research methods for criminal justice
and criminology. Belmont, CA: Wadsworth.

Agnew, R. (1992). Foundation for a general strain theory of

crime and delinquency. Criminology, 30, 47–88.

Akers, R., Krohn, M., Lanza-Kaduce, L. & Radosevich, M. (1979). Social

learning and deviant behavior: A specific test of a general theory.
American Sociological Review, 44(4): 636-655.

Babbie, E. (2001). The practice of social research. Belmont, CA:

Wadsworth

Babu, B.V. & Kar, S.K. (2010). Domestic violence in Eastern India:
Factors associated with victimization and perpetration. Public Health,
124: 136-148.

Champion, D. (1993). Research methods for criminal justice and

criminology. Englewood Cliffs, NJ: Prentice Hall.

Dangerous Drugs Board. (2008). Study on the Current Nature and Extent
of Drug Abuse in the Philippines. Retrieved on February 12, 2013, from
http://www.ddb.gov.ph/joomla/images/2008%20National%20Househol
d%20Survey.pdf

Dantzker, M.L. & Hunter, R. (2006). Research methods for criminology

and criminal justice. Sudbury, MA: Jones and Bartlett Publishers, Inc.

Donner, C. & Jennings, W. G. (2014). Low self-control and police

deviance: Applying Gottfredson and Hirschi’s general theory to officer
misconduct. Police Quarterly, 17(3): 203-225.

Fraenkel, J. & Wallen, N. (2006). How to design and evaluate research in

education. New York, NY: McGraw-Hill.

Gottfredson, M.R. & Hirschi, T. (1990). A general theory of crime.

Stanford, CA: Stanford University Press.

Kelling, G.L., Pate, A., Diekman, D. & Brown, C. (1974). The Kansas City
Preventive Patrol Experiment: Summary Version. Washington, D.C.: Police
 50

Foundation
Maxfield, M. & Babbie, E. (2005). Research methods for criminal justice
and criminology. Belmont, CA: Wadsworth.

Ysmael, F. (2009). “Lumen”: A case study on domestic violence. JPAIR

Multidisciplinary Research, 3(1):58-72.