The Core Principles Related To Public Relations Research

Public Relations research is highly important for PR professionals, not because they will often conduct research, but because they will need to make reliable and informed decisions about hiring research firms and evaluating how their research benefits the bottom line. In PR the bottom line refers to a company’s income after all expenses have been deducted from revenues. So, the main use for Public Relations research is to demonstrate to clients that what they have produced has impacted some public or audience. There are two ways of viewing research: formal and informal. Formal research is the systematic gathering, analyzing and evaluating of data through a certain methodology. Informal research is the observing of events, people or objects of interest as they happen to occur. Formal research can be quantitative or qualitative, while informal research is typically qualitative. We will go deeper into what these categories mean when we analyze categorical and continuous statistics, but first we will learn their definitions.

To understand how we conduct Public Relations research we must understand what research is. Research encompasses two methodological approaches to data, and data are the observations we make of the world around us through these methodologies. Quantitative research is objective with a heavy reliance on numbers. It is a systematic and controlled form of gathering data. On the other hand, qualitative research is subjective and less controlled. This type of research relies more on the subjective evaluations of the researcher.

As many of us know, one of the main tasks in conducting research is to ask questions. However, to focus on asking appropriate research questions we must distinguish between theoretical and applied research. Theoretical research seeks to provide the underlying framework for the study of PR. A theoretical researcher defines which concepts can be used, how they relate to each other and under what conditions we can expect results. Applied research uses theory-driven research in real-world situations.

One category of research questions are questions of definition, which I will use to give an example of both types of research. Questions of definition are the most basic questions asked by PR researchers. They define what it is we are attempting to observe. The theoretical researcher will ask whether a particular concept or idea actually exists and how it can be potentially measured. The applied researcher will then take those conceptual definitions and develop a communications program around them. Questions of definition can be answered with both qualitative and quantitative methodology. Next are questions of fact, which seek to compare across or between groups. They often deal with quantity and are also referred to as empirical questions. Questions of fact cannot technically be looked at qualitatively but can be confirmed or denied based on observation. Questions of value ask “how well” or “how good” something is. These types of questions often rely on attitude measures, and they can be answered with both qualitative and quantitative research. Quantitatively we can find out if the audience believes something was done well, but we need the qualitative research to find out why and provide feedback. Lastly, questions of policy are almost always strategic, and they ask what should be done. Majority of the time these questions fall under applied research and carry legal implications. Oddly enough, these questions are rarely asked by researchers but instead by theorists in the academic world or by executives in the business world. An example of a policy question might be: Should we do X because of Y?

Now that we know a bit of background about the requirements for asking appropriate research questions, let’s move on to the difference between categorical and continuous statistics. However, statistical analysis is secondary to the collection of data. So, we will discuss measurement and collection of data first. To determine the impact of a Public Relations campaign we must be able to measure it, and in PR research we have certain standards in terms of measurement type and creation. By definition measurement is the process by which we observe things seen and unseen. It normally includes a comparison of some sort. Formal and informal measurement exists, but for the purpose of this discussion we will focus on formal measurement. Formal measurement requires that as precisely as possible we state how we measure what we observe. It helps us decide exactly how to approach problems by stating what level of measurement we are using when observing, and it provides standards of analysis and interpretation.

Observations for formal measurement can be made at four levels: nominal, ordinal, interval or ratio. Nominal and ordinal fall under categorical measurement. Categorical measurement is defined as occurring when our observations are placed into classes to be counted. Interval and ratio fall under continuous measurement, which is defined as occurring when our observations are placed on some continuum. In short, categorical data is pretty simple and language-based, while continuous data is more complex and assumes that observations are equidistant apart. Continuous data can be reduced to categorical, but categorical data can never be changed to continuous. An important note to remember is that Public Relations research uses both classes and makes observations at all four levels.

So, let’s go further into depth on each of these four levels. Starting with nominal, it is the most basic measurement we can make placing data into classes, with distinction made only as to differences. For example, it may include sex, race or size. With no saying that one is better than another. Ordinal measurement places the same simple observations into some type of order. For example, a campaign asks survey respondents what brand of car they had purchased over a 5-year period. If we are just interested in the brand, not comparing them, then we have nominal data. However, if we also ask respondents to order their preference using a list of brands then we have ordinal data. We must note that categories are exclusive – you cannot be in two categories at the same time. The key to understanding ordinal measurement is to know that it represents an order. Moving on to interval measurement, we say that is assumes that the distances between observations are equal all along the continuum. If we were to ask how old someone is, we know the distance between ages 1 and 2 is 1 unit and between ages 35 and 70 is 35 units. And finally, ratio measurement adds the requirement of an absolute zero point to the continuum. Meaning not only are the classes equidistant on the continuum, but the continuum now has an absolute zero point. For example, the number of days a newspaper is read each week can be 0 to 7. Each of the four levels of measurement has its advantages and disadvantages. Categorical observations are very simple to make but they can be hard to interpret, while continuous observations are harder to make but easier to explain. Everyone should note that percentage is not a continuous output. Percentage is based on categorical data only. Lastly, remember that all measures should be assessed for their validity and reliability, and good measures are both reliable and valid. It is important that we know how to correctly gather and measure data because if it is not done properly the statistical analysis will be full of error. Data are only as good as their definition and the method by which they have been collected. A statistical approach to data is defined as the frequencies, means and percentages used to asses a campaign or program. So, the difference between categorical and continuous statistics is that categorical statistics assume that the data are simply observed, and these observations are labeled as “variables.”

On the other hand, continuous statistics looks for differences in the measured variable as defined by some other variable. We will discuss what these variables are later on. When speaking in terms of descriptive statistics we would say continuous data or categorical data. However, continuous data can also be known as parametric and categorical data as non-parametric. This is because statistical theory redefines data according to how they are distributed. Parametric data are interval in nature, have a tendency to group together and have a common mean from which they vary individually. Parametric data fall on the normal curve which I will mention again when I discuss univariate analyses. Because categorical data is found in categories it becomes non-parametric. We would use the terms parametric and non-parametric when discussing inferential statistics.

Now we’ll look at what univariate analyses are. Beginning with univariate continuous data, data are found along some continuum rather than in categories. In this case, the data must fall on what statisticians call the “normal curve.” All continuous data have their own “normal distribution” which brings us to the measure of central tendency. It is defined to be a statistic that describes the typical or average case in the distribution of a variable. Central tendencies include mean, median, mode, standard deviation and variance. Each of these univariate continuous analyses is important to the basic understanding of continuous data. The mean is often labeled as the average response found by summing up all the data and dividing it by the number of valid observations. However, the median comes in handy because the mean is sensitive to extreme scores. The median represents the midpoint, or the 50th percentile of the data set, when listed along the continuum. The mode is simply the most frequently repeated value in a data set. If there is more than one mode, the data set is considered bimodal or trimodal if there are three modes. The variance is what describes how the data is distributed around the mean. It represents the amount of dispersion around the mean and is obtained through a mathematical formula. Finally, the standard deviation tells us the range of scores that we would normally expect a certain percentage of the scores to fall on within the continuum. It is based on the normal curve

Now looking at univariate categorical analyses, they take the form of frequency counts and percentages. A frequency table is simply an analysis of each variable and its categories. This is the main difference between the measure of central tendency and frequency statistics. While central tendency would require the mean, mode, standard deviation, etc., frequency statistics are set up in tables. Each frequency table is divided into four parts. The first part would provide information on how many participants responded to the variable and how many did not. The second part would provide information about the variable such as the name, coding categories and labels. The third part consists of the actual counts for each category. And finally, the fourth part provides the actual “percent,” the “valid” percent and the “cumulative” percent. Some options for frequency statistics are to make bar charts, pie charts or histograms.

Now let’s talk about what content analysis is and why it’s important in Public Relations. Content analysis allows us to look at qualitative data in a quantitative manner. It helps us break up information from interviews, subject observations and focus groups. We then place the information into categories which can be counted and quantified. Examples of its usage have been to count the number of times a client’s name has been printed or to examine the impact of public relations messages. Although it may seem like content analysis is a quantitative methodology, it’s still taken from a qualitative method. Content analysis is essentially quantitative and qualitative, and it can be used to answer questions of fact. It is defined as a systematic, objective, and quantitative method for researching messages. Content analysis is best used to analyze documents, speeches, media releases, video content, scripts, interviews and focus groups. The most important concept for content analysis is that it must be done objectively. This means that a researcher should be able to give his or her content to another researcher and based off the same rules the new researcher would get the same results. Content analysis must also be systematic meaning the “rules” are consistently applied across all research. At the end, the qualitative research will be turned into numbers or data that can be quantified and compared.

Like all systems, content analysis has advantages and disadvantages. Its biggest advantage is that it can objectively and reliably describe a message, or group of messages, and its application to advanced statistical analyses. It also provides logical and statistical bases for understanding how messages are created. The only disadvantage of content analysis is that it requires messages to be recorded for analysis. Overall, content analysis is important in Public Relations research because it allows researchers to employ very sophisticated statistical analyses on qualitative data. It’s especially effective when used for examining public relations messages for their frequency, construction and content.

One of several steps involved in content analysis is coding. The coding process begins after the units of analysis have been identified, a category system is in place and the messages to be content analyzed are ready. Coding happens when the researcher identifies and places the messages in a chosen category system. Coding is the way you quantify messages when you distinguish between the units of analysis. Most coding schemes reflect simple differences that do not indicate ordering. Coding validity and coding reliability are extremely important to the process. Coding validity ensures that you are actually coding the messages the way they should be coded. Three things impact coding validity. The first is, what are the units of analysis and how are they defined? Defining units before coding them is crucial to validity. Second is, what is the category system? If the categories are not created meticulously and in the end are not complete nor exclusive, then they might not reflect the correct purpose of the analysis. Third, validity is often compromised by the process in which the actual messages are gathered for analysis. This means that a lack of sampling is an issue. Researchers must make sure they are sampling at appropriate times to reflect normalcy of the subject being observed.

Now that we know what coding validity is, let’s talk about coding reliability. Coding reliability can be measured, and it refers to the amount of error coders make when placing content into categories. Although we did not discuss different types of content analysis, I will mention that manifest analysis is more reliable than latent analysis. Even something as simple as word count can result in error and reduce reliability. All content analyses can be reviewed by two different coders, which is called intercoder reliability. The other option is for the same coder to review the material twice, which is called intracoder reliability. By coding the material twice, we can ensure reliability. For coding reliability, we look at a reliability coefficient. For example, a coefficient of 0.00 means that two completely different results were obtained. In this case, we have no reliability whatsoever. However, if we get a coefficient of 1.00 then that would mean the exact same results were received. You are considered to have “good” reliability when coders agree on coding content at least 90% of the time, or 0.90. There are two formulas used to compute reliability. So, we can see how important it is to make sure our coding has both validity and reliability. While coding, we should ensure that we get at least a 90% reliability coefficient to be able to consider our research “good”.

We began talking about continuous and categorical statistics, and I mentioned that later on we would discuss different kinds of variables. So, there are two different types of variables used in research. I already stated that categorical statistical analyses simply label the observations as variables. Continuous statistical analyses are the ones that use dependent and independent variables. The variable that is measured or collected and evaluated is called a dependent variable. The value of the dependent variable is determined by the level or class of the independent variable. An independent variable is defined as the variable against which the dependent variable is tested. The independent variable is usually the one that is manipulated or occurs naturally, given the occurrence of antecedents to the dependent variable. In most Public Relations research the dependent variable can be either continuous or categorical, and it’s labeled as the outcome variable. The independent variable is usually categorical, and it’s labeled as the predictor variable. Below is an example I have taken from Professor Barrie Gunther on how to differentiate between a dependent and independent variable. When a Public Relations campaign places a news story in a TV bulletin it expects some effect on audience recall. Putting the news story out for the public to see is going to affect the recall of its content for the viewer. So, the independent variable is the placing of the story and the dependent variable is the audiences recall. We can see how the independent variable affects the dependent variable and how the independent variable is the one researchers manipulate.

I hope that everyone was able to learn something from all of the information I have laid out. Don W. Stacks wrote an entire book containing this information, and much more, simply to prove that research does matter in Public Relations. We’ve discussed what research is, how we can measure it, how we can break it up into classes and how we can conduct it. It’s important for PR professionals to understand that these concepts can be extremely complex and should be studied in depth and with great attention.