Unit 4 notes - .......................... PDF

Preview

Summary

.............................

Description

Unit 4 The Logic of Causation Social science typically operates on the basis of a causal model that is probabilistic, rather than predicting. e.g., going to university, factors make attending university more or less likely High school students with University parents students w/o university parents

More likely to attend university university

High School

Less likely to attend

Does not mean either will or will not Causation can be defined as an action that brings about a change, an association, a relationship, or a correlation. But unlike in the natural sciences where it is easier to establish causality, the definition of causality in the social sciences is more complex. For example, the light comes on anytime you turn on the switch, with exception of when the bulb is dead or when the power is out. However, in the social sciences it is sometimes difficult to determine if A (the independent variable) is causing B (the dependent variable) and vice versa, or whether there are other factors influencing the change in B. Due to this potential problem, causal explanations in the social sciences are probabilistic. This means the likelihood of a certain event occurring when certain conditions are present, instead of the certainty that the event will occur. For example, research shows that children of divorce are more likely to divorce themselves. This does not mean that the entire population in this category will divorce. It just means that the likelihood of their divorcing is higher within that population. Next we shall examine two approaches to causation in the social sciences. Causation in idiographic and nomothetic models of explanation

The idiographic approach to causation involves listing all of the prior conditions that influence a given event. The focus is on looking for exhaustive explanations and specific reasons why an event may occur. Unlike nomothetic explanations, idiographic approaches do not look for general laws underlying events. For example, an idiographic approach to high unemployment rates in the United States since 2007 would focus on specific factors. In contrast, a nomothetic approach would involve selecting the most significant factors contributing to high unemployment. The goal would be to focus on the underlying factors that affect the majority or all of the unemployed. The nomothetic approach provides a partial explanation. This model is probabilistic, and the emphasis is on looking for general patterns. Due to the probabilistic nature of nomothetic explanations, researchers often use such terms as “likely” or “lead to.” Idiographic model aims at explanation through the enumeration of the many reasons that lie behind a particular event or action. Although in practice we never truly exhaust such reasons, the idiographic model is frequently employed in many different contexts. e.g., Traditional historians, for example tend to use the idiographic model, enumerating all the special causes of the French Revolution or World War II. Clinical psychologists Clinical psychologists may employ this model in seeking an explanation for the aberrant behaviour of a patient. A criminal court, in response to a plea of extenuating circumstances, may seek to examine all the various factors that played a role in the defendant’s behaviour. Social researchers sometimes use this model as well. While the idiographic model of explanation is often used in daily life and in social research other situations and purposes call for the nomothetic model of explanation. This model does not involve enumerating all the considerations that result in a particular action or event. Rather, it is designed to discover those considerations

that are most important in explaining general class of actions or events. e.g., suppose we wanted to find out why people voted for or against the 1995 referendum on sovereignty held in Quebec. Idiographic Approach: Each individual we spoke to could give a great number of reasons why s/he votes yes/no. Suppose someone gave us 99 different reasons for voting yes. We’d probably feel we had a pretty complete explanation for that person’s vote. In fact, if we found someone else with those same 99 reasons, we would feel pretty confident in predicting that that person also voted yes. The nomothetic model of explanation, involves the isolation of those relatively few considerations that will provide partial explanation for the voting behaviour of many or all people Nomothetic approach: Most of those sharing the attribute Anglophone probably voted no, while more of those sharing the attribute Francophone probably voted yes. This single linguistic consideration (mother-tongue) would not provide a complete explanation for all voting behaviour. Some Anglophones voted in favour of Quebec sovereignty and some francophones voted against it. The nomothetic model of explanation with the fewest number of causal variables to uncover general patterns of cause and effect. The nomothetic model of explanation is inevitably probabilistic in its approach to causation. Naming a few considerations seldom if ever provides a complete explanation for complex behaviours. The nomothetic model indicates a very high (or very low) probability that a given action will occur whenever a limited number of specified considerations to the larger number of specified considerations. Adding a larger number of specified considerations to the equation typically increases the degree of explanation, but

it also makes explanations more complex- so much so that they cease to be useful. e.g., everyone who believed the referendum was best for Quebec and Canada voted yes, but that would not be a very satisfying explanation. Phrases like “leads to”, “arise from”, and “predictor of” signal causal explanations. Notice how they are trying to determine which of a limited number of factors have an impact in order to provide useful general patterns of cause-effect relationships that hold for a wide range of cases. Approach to prejudice Ideographically: trying to understand why a given person is prejudiced, noting that a large number of idiosyncratic circumstances and experiences have contributed to their views. Nomothetic: they look for factors that affect levels of prejudice in general (e.g., those with more education are generally less prejudiced than are the less well educated. The idea of causation is present in both the idiographic and nomothetic modes of explanation. Both are legitimate and useful. Correlation: such that

An empirical relationship between two variables

a) changes in one are associated with changes in the other, or b) particular attributes of one variable are associated with particular attributes for the other. e.g., weight and height are said to be correlated because of the association between increases in height and increases weight. Correlation in and of itself does not constitute a causal relationship between two variables, but it is one criterion of causality. Criteria for Nomothetic Causality

Due to the difficulty involved in determining causal relationships in the social sciences, three requirements must be satisfied when determining causality. 1. Correlation: The variables must be correlated; that is, a change in one variable must produce an effect in the other. For example, studying and passing an exam. But the question is, What about those who study but don’t pass the exam? This means that there is no perfect correlation. 2. Time or Temporal Order: The cause must precede the effect in time. For example, does coming from the Horn of Africa make an athlete a good long-distance runner, or does the high altitude help with training, thus making for a good-long distance athlete? Establishing a time order can be complex, much as in the case of the chicken and the egg—which came first? 3. Nonspuriousness: The empirical relationship must not be the result of a third factor. For example, if a researcher conducts a study on the academic performance of students in working-class and upper-middle-class neighbourhoods and finds that students in working-class neighbourhoods perform poorly on standardized tests as compared to their more well-off counterparts, the researcher might conclude that students in richer neighbourhoods are smarter. But does living in a richer neighbourhood per se make a student smarter, or are there other factors at play? Is it possible that students in richer neighbourhoods have better schools, resources, and teachers? If so, then the established relationship between neighbourhood and performance on standardized tests is spurious, or not genuine, because there is a third factor—quality of schools or teachers—influencing the outcome. Much social research is ultimately geared to revealing the causes of social phenomena. The mere fact that variables are observed to have some relationship does not establish that the relationship is one of cause and effect.

E.g., People affiliated with the Liberal party – prefer strawberry ice cream. Any such observed association between these variables would be coincidence. E.g., People who own sailboats – more likely to own a large collection of recorded music. There is no reason why owning a sailboat would cause a person to own a large collection of recorded music. Perhaps, people who own sailboats may have higher incomes and thus also own large collections of recorded music. Science involves both observation and logic. In the above examples, logic suggests that the observed relationships, or associations, are not causal by nature. There are three criteria for nomothetic causal relationships in social research: 1. The variables must be correlated 2. The cause takes place before the effect 3. The variables are nonspurious. CorrelationTo say that a causal relationship exists, there must be an actual, observed relationship- or correlation- between two variables. A correlation exists to be related-that is, when one occurs or changes, so does the other. E.g., Exploding gunpowder causes bullets to leave muzzles of guns if, in observed reality, bullets did not come out after the gunpowder exploded (they came out even when it didn’t explode). In the probabilistic world of nomothetic explanations, there are a few perfect correlations-that is, as one variable takes on different values, so does some other variable, in perfect lockstep and without a single exception. Time order We can’t say that causal relationship exists unless the cause precedes the effect in time. It makes no sense to imagine something being caused by something else that happened later on.

e.g., a bullet leaving the muzzle of a gun does not cause the gunpowder to explode; it works the other way around. -owning a luxury car doesn’t cause one to earn enough money to afford one. Nonspurious the third requirement for a casual relationship is that the observed empirical correlation between two variables cannot be explained in terms of some third variable. e.g., there is a positive (also referred to as direct) correlation between ice cream sales and deaths due to drownings, and vise versa. There is obviously no direct link between ice cream and drowning. The third variable at work here is season or temperature. Most drowning deaths occur during summer-the peak period for ice cream sales The above is an example of spurious relationshipsrelationships- relationships that aren’t genuine. The observed association (correlation) between ice cream sold and drownings is real enough, but a causal linkage between the variables would be spurious. In reality, a third variable explains the observed association. E.g., another example of a spurious relationship is the positive correlation between shoe size and math ability among school children. Here, the third variable that explains the puzzling relationship is age. Older children have bigger feet and more highly developed math skills, on average, than do younger children. An example of negative (also referred to as inverse) relationship between the number of mules and the number of Ph.D.’s in towns and cities: the more mules, the fewer Ph.D’s and vide versa. What would be an additional variable to explain this relationship? The answer is rural versus urban settings. There are more mules (and fewer Ph.D.’s) in rural areas, whereas the opposite is true in cities.

Spurious Relationship A coincidental statistical correlation between two variables that is shown to be caused by some third variable. When social researchers say there is a casual relationship between, say, education in racial tolerance, they mean 1. There is a statistical correlation between the two variables 2. A person’s educational level occurred before their current level of tolerance of prejudice, and 3. There is no third variable that can explain away the observed correlation as spurious. False Criteria for Nomothetic Causality Since social science assumes that relationships among phenomena are probabilistic rather than certain, complete causation suggests false nomothetic causality. The existence of exceptional cases in nomothetic explanations does not nullify the existence of a causal relationship. For example, highly educated women often have fewer children; however, finding a highly educated woman with more than the average number of children for women in her category does not mean that the overall pattern does not exist. What’s important here is probability, or the likelihood of this pattern occurring. As well, causal explanations can exist even if they do not apply to a majority of cases. For example, research shows that people who cohabit before marriage are more likely to divorce than those who do not. The connection between cohabitation before marriage and divorce is causal so long as the population in this category is more likely to divorce as compared to those who do not cohabit before marriage. Instead of looking for complete causation or identifying exceptional cases, social researchers determine whether there is a necessary or a sufficient cause for an event to occur. A necessary cause is a condition that must be present for an event to occur, while a sufficient cause guarantees the occurrence of an event. It is important to specify some of the things social researchers no not mean when they speak of causal relationships. When they say one variable causes another, they do not necessarily mean to

suggest complete causation, to account for exceptional cases, or to claim that the causation exists in the majority of cases. Complete Causation An idiographic explanation of causation is relatively complete, a nomothetic explanation is probabilistic and usually incomplete. Exceptional Cases In nomothetic explanations, exceptions do not disconfirm a causal relationship. Majority Cases Causal relationships can be true even if they do not apply in a majority of cases. The social science view of causation may vary from what you are accustomed to, since people commonly use the term cause to mean something that completely causes another thing. The somewhat different standard used by social researchers can be seen more clearly in terms of necessary and sufficient causes. Necessary and Sufficient Causes Necessary cause represents a condition that must be present for the effect to follow. e.g., it is necessary for you to take university courses in order to get a degree-without the courses, the degree never happens. Simply taking the courses is not a sufficient cause of getting a degree. You need to take the right ones and pass them. Sufficient cause represents a condition that, if it is present, guarantees the effect in question. This does not mean that a sufficient cause is the only possible cause of a particular effect. e.g., skipping an exam in this course would be a sufficient cause for failing it, though students could fail it other ways as well. Therefore, a cause can be sufficient but not necessary. The discovery of a cause that is both necessary and sufficient is, of course, the most satisfying outcome in research. If juvenile delinquency were the effect under examination, it would be nice to discover a single condition that

1. Must be present for delinquency to develop and 2. Always resulted in delinquency. In such a case, you would surely feel that you knew precisely what caused juvenile delinquency. In the idiographic analysis of single cases, you may reach a depth of explanation for which it is reasonable to assume that things could not have turned out differently, suggesting you have determined the sufficient causes for a particular result. Obstacles in the Search for Nomothetic Causality There are four obstacles inherent to social scientific data that inhibit the search for causal relationships. 1. A hidden third factor may account for the variation in both the hypothesized cause and the effect variables. This obstacle is related to nonspuriousness, the third criterion for nomothetic causality. It is easy to control for a nonspurious relationship in an experimental situation because the effects of each variable can be tested. But it is not as easy when using a questionnaire. This problem can be minimized, however, when the same question is asked twice, but phrased differently. 2. Multivariable causation occurs when a number of factors may be acting together to cause the effect in the dependent variable. For example, a researcher might attribute Canada’s New Democratic Party’s (NDP) gains in the province of Quebec during the 2011 Federal Election to the personality of the late leader Jack Layton. However, there could be other factors involved, such as the party’s platform and decline in support for the Bloc Quebecois. The challenge is to find out how much each of these factors contributed to the effect (dependent variable), the NDP victory in Quebec. A researcher can hold some variables constant under experimental conditions (that is in a laboratory); however, all things cannot be equal (ceteris paribus) in everyday life. Hence a

researcher must collect data in more than one way to overcome or minimize the effect of this obstacle. 3. Another problem may be confusing the dependent and independent variables. In other words, a third factor may be causing both the dependent and independent variables. For example, to determine causality in the success (dependent variable) of the Jamaican sprinter Usain Bolt, we can look at the effect of the independent variable, training. However, it is possible that other factors, such as natural talent, the motivation to win, and the sports programs in his native Jamaica may be affecting both his training (independent variable) and his success (dependent variable) in the 100-metre race. 4. The interaction between dependent and independent variables may be an obstacle in determining causality. The dependent and independent variables may be influencing one another in terms of either positive or negative feedback. Here is a hypothetical situation: there are a lot of payday-loan stores in low-income neighbourhoods as compared to in high-income neighbourhoods. But does the location of the stores increase demand for their services in low-income neighbourhoods, or are the stores located there because the demand is high? A researcher can avoid this issue by focusing on whether or not there is an association or correlation between both variables. However, sometimes it is necessary to assign either a positive or a negative direction between the variables. To some degree, these difficulties can be managed by following the experimental methods outlined in Unit 11. It is worth noting that causality is applicable to the deductive approach when variables can be isolated and expected outcomes tested, but is rarely applied in the inductive approach. Social regularieties help researchers...