1. Unobtrusive measures are ways of studying social behavior without affecting that behavior. They most often involve observing or recording indicators of what happened based on evidence that is left behind after that social behavior has taken place. For instance, unobtrusive measurement can be done by observing or recording the number of clients who successfully completed a treatment program on the basis of what their case records or files indicate.
2. Case record studies are a common and important form of social work research. Many important studies have been done by primarily recording information from case records. Such studies include ones indicating whether an agency is providing the appropriate level of service for the cases assigned to it, whether particular programs are effective in providing treatment, whether some forms of treatment are more effective than others, whether some groups of clients are more likely to receive favorable treatment, etc. These and other topics can be studied by coding information in case records and interpreting the results. These studies are most often quantitative but they can also be qualitative.
3. Studying case records is one form of content analysis. Content analysis can be defined as a method for studying communications. The units of observation in content analysis are most often written documents but they can also include other records of communications such as video recordings. We usually examine these units of observation to learn something about the people who either created these documents or are referred to by them. Who authored the documents, say therapists, or who is referred to by the documents, e.g., clients, would be the units of analysis. For instance, we often study the content of case records to learn about the clients or service providers referred to in those records. The records are the units of observation and the clients are the units of analysis.
4. For instance, for a study of body language, we could tape therapy sessions and then observe say a random sample of them to count up the number of times certain therapists adopt a posture that we had defined as an empathetic position. In such a study, the therapist would be the unit of analysis while each session would the unit of observation. In content analysis, as in other studies, the unit of analysis is not always the same as the unit of observation. We might find that older therapists were more likely to practice empathetic body language and that when they did so clients were more likely to be responsive to questioning. The advantage of using video as a record of communication associated with body language is that such communication is not recorded in transcripts, case records or written documents generally.
5. Content analysis of case records can involve many things from studying the frequency of use of certain words, to how often clients are treated in a particular way, to counting up which types of clients tend to receive what types of treatment, to many other tallies. Content analysis can also take many other forms as the body language example suggests. All of these however involve a coding scheme. Coding is critical to content analysis. All coders must use the same criterion or operational definition for coding each item or variable you study. The coding instructions for all the items or variables are usually specified in a codebook or protocol. These instructions specify how to look for certain pieces of information and convert them to codes in numerical form that can later be tabulated usually using a computer. Say you wanted to audit welfare eligibility determinations to see if certain groups of clients were more likely to be inappropriately denied assistance when they are actually eligible. You would need to first define eligibility, select a representative sample of case records, and then code each case as to the clients background and whether that case was correctly designated as eligible or not. Then you could later enter all the data into a computer and tabulate whether certain groups of clients were more likely to be incorrectly denied welfare benefits.
6. For a content analysis to have any merit, coders must have a high level of intercoder reliability. Intercoder reliability is achieved when coders are found to be coding the records in the same way. This can be tested for by periodically having a second coder code the same records as another coder. For the selected test cases, each coder ideally should record the same results as the other coder for each case.
7. The issue of intercoder reliability highlights how content analysis involves making judgments as to how to read a record and convert what you read into codes or numbers. Sometimes this can be very obvious as in counting up the number of times in each record a certain word is used. For instance, a study of how often clients were categorized in their records as addicts would not be very complicated. This would be an example of coding the manifest content or surface content of the records. Yet, often content analysis requires coders to make more of a judgment or interpretation as when coders are asked to code if the case record indicates anything to the effect that the client was prematurely terminated from treatment. This would be an example of coding the more latent content or nonobvious content.
8. Emphasizing the manifest over the latent content often involves trading off validity for reliability. Coding the manifest content can be highly reliable but merely a superficial and not a very valid measure of how a client is treated. Coding the latent content can be very valid but not very reliable as each coder is expected to make subjective interpretive judgments.
9. Case record and other content analyses often involve sampling. For a quality control audit to estimate the error rate in the determination welfare eligibility and allocation of welfare benefits, a sample of case records is selected.
10. There are many other forms of unobtrusive measurement including analyzing existing statistics and reviewing historical records and documents. Often existing statistics are in aggregate form as in data on neighborhoods, cities, states and countries. Such aggregate data are also called ecological data. One major potential pitfall of using ecological data is the ecological fallacy. The ecological fallacy is where we use data on ecological units, like neighborhoods or cities, to make inferences about individuals. For instance, we might have census data indicating that neighborhoods with drug addicts have more murders. We want then want to infer that we have evidence for the hypothesis that drug addicts are more likely to commit violent crimes, like murder. Yet this could well be an instance of the ecological fallacy. It is distinctly possible that drug addicts hardly even commit murder and do so no more frequently than any one else; it just happens to be that they tend to live in neighborhoods that also have relatively high murder rates. Poor neighborhoods tend to have both a relatively higher number of murders and drug addicts, though the latter may not be committing the former. The only way we could find out evidence for our hypothesis would be to make our unit of analysis individuals not neighborhoods and then study to see if individuals who were addicted to drugs were more likely than other individuals to commit violent crimes, including murder.
11. Unobtrusive measurement generally has advantages and disadvantages. Advantages--the people being studied have already acted and we need not worry about research reactivity, such studies have low cost, and they can usually be redone if mistakes or omissions are later noted since the case records will most often remain available. Disadvantages include that such studies may only produce superficial evidence about what the documents refer to. Documents and records, textual or otherwise, contemporary or historical, are not a complete record of all that happened. What is off the record or not in the public transcript, or even what is between the lines, may be more important than what is recorded in text or image, documents or videos. And it is important to remember that examining records has its own ethical issues which should be addressed in each case, most especially issues of confidentiality and privacy.