Gambling Behavior in the United States: A Data-Driven Learning Guide

Goal & Concept

Goal

The goal of this exercise is to explore gambling behavior and characteristics of gamblers in the United States. Frequencies, crosstabulation, and bar charts will be used. This exercise will also illustrate the role of confounding variables in statistical relationships.

Concept

Deviant behavior is socially and culturally defined. Behaviors considered deviant in one society may be viewed as quite harmless in another. A behavior may even be defined as deviant when one type of person does it, but not when another type of person does. In the United States, gambling is subject to similar socially constructed definitions of deviance. Gambling is closely regulated by the state and legalized forms of gambling are often considered socially acceptable, while illegal forms are not. Gambling is most often considered deviant when it is taken to extremes; compulsive or problem gambling, like substance abuse, is stigmatized in the United States.

Examples of possible research questions about gambling behavior:

How many people gamble?
What types of gambling are most popular?
Does gambling behavior differ by gender?
Is income level related to gambling behavior?
How many people are problem gamblers?
Is gambling related to substance use?
How are gambling and mental health related?

Dataset

The Gambling Impact and Behavior Study (1997-1999) investigates the gambling behavior and attitudes of adults and youth in the United States, and also estimates the effects of gambling facilities on a variety of local economic and social indicators. Respondents were randomly selected by a national random-digit dial (RDD) through a stratified design by state lottery status and distances to a major casino. The study includes three parts: the adult and youth questionnaire (Parts 1 and 2) covered areas such as demographic information, geographic region, gambling behavior and attitudes, motivations for gambling, gambling history, a problem-gambling diagnostic assessment, gambling treatment experience, family/marital status and issues, income and financial information, criminal activity, mental and general health, and substance use. The Community Database (Part 3) included the following: geographic locators (latitude, longitude), availability of gaming facilities, gaming spending estimates, employment patterns by industry, unemployment, bankruptcy, personal income, private and public earnings, government expenditures, income maintenance/AFDC, and vital statistics.

The study is representative of the civilian household population of the United States aged 16 and older, living in households with at least one working telephone line.

The Adult Survey will be used in this exercise.

This exercise will use the following variables:

Ever gambled casino (B1_)
Ever gambled on-track or off-track betting facility (B4_)
Ever gambled lottery (B7_)
Ever gambled bingo (B10_)
Ever gambled charity (B13_)
Ever gambled card room (B16_)
Ever gambled private (B19_)
Ever gambled store/bar/restaurant (B22_)
Ever gambled unlicensed (B25_)
Ever gambled internet (B28_)
Ever gambled Indian or tribal casino(B30_)
Past year gambled casino (B3_)
Past year gambled on-track or off-track betting facility (B6_)
Past year gambled lottery (B9_)
Past year gambled bingo (B12_)
Past year gambled charity (B15_)
Past year gambled card room(B18_)
Past year gambled private (B21_)
Past year gambled store/bar/restaurant (B24_)
Past year gambled unlicensed (B27_)
Past year gambled internet (B29_)
Past year gambled Indian or tribal casino (B31_)
Number of gambling problems (EVERPROB)
Gender (A1_)
Age Group (A2_R)
Race/ethnicity (RACETH)
Household Income (INCOME)

Application

Gambling Behavior

To measure whether respondents had ever gambled and whether they had gambled in the past year, we created two new variables, EVERGAMB and YEARGAMB. EVERGAMB was coded as "1" if respondents reported having ever done at least one of 11 different types of gambling (see variables B1_, B4_, B7_, B10_, B13_, B16_, B19_, B22_, B25_, B28_, and B30_), and "0" if they had not done any kind of gambling. YEARGAMB was coded in the same way by using variables measuring gambling activity in the past year (see variables B3_, B6_, B9_, B12_, B15_, B18_, B21_, B24_, B27_, B29_, and B31_).

To begin the exercise, determine the percentage of people who have ever gambled. Look at the frequency distribution of EVERGAMB. What percentage of sample respondents has ever gambled? Is it higher or lower than you expected?

Next, consider gambling behavior in the past year by looking at the frequency distribution of YEARGAMB. Compare the percentage of people who say they have not gambled in the past year to the percentage who have never gambled. Is there a difference?

Now consider the different types of gambling people may do. Limiting your sample to respondents who have ever gambled, what percentage have ever gambled in a non-tribal casino? How about in an Indian or tribal casino? An on-track or off-track betting facility? A lottery? How many have gambled on bingo or private games (dice, poker, pool, etc)? What percentage have ever gambled in a store/bar/restaurant (video poker, pull-tabs, etc), or in an unlicensed facility? Which type of gambling appears to be most popular?

Characteristics of Gamblers

Now that you have an idea of gambling behavior, think about the characteristics of gamblers themselves. First consider age. For ease of analysis, we recoded age (A2_R) into three categories: ages 18-29, 30-49, and ages 50 and older. We called the new variable, "AGE3."

Run a crosstab of EVERGAMB and AGE3. In which age group did the largest percentage of respondents report ever gambling?

Now look at gambling behavior in the past year alone, limiting the sample to only those who have ever gambled. What differences do you see in the table of YEARGAMB and AGE3 compared with the previous analysis? Is the age pattern different for gambling in the past year than for lifetime gambling?

Does gambling behavior differ by gender? Run a crosstab of EVERGAMB and A1_. Are men or women more likely to have ever gambled?

Now consider the relationship between gambling and income. Run a crosstab of EVERGAMB and INCOME. How does gambling behavior differ by income group?

Try the analysis again with YEARGAMB and INCOME. How might you explain the difference between the pattern shown in this analysis and the crosstab of lifetime gambling and income?

Next think about the relationship between gambling behavior and race/ethnicity. Run a crosstab of EVERGAMB and RACETH. Looking at the bar chart, what racial/ethnic group has the highest percentage of lifetime gambling? What racial/ethnic group has the lowest percentage?

Because racial/ethnic group is often closely related to income level, and you have seen in previous analyses that income level is related to gambling behavior, income group might be a confounding variable in the relationship between racial/ethnic group and gambling behavior. Rerun the crosstab of EVERGAMB and RACETH, limiting the sample to only those respondents in the three highest income groups (excluding those in the $24,000 and under group). Does the relationship between race/ethnicity and gambling behavior change when the lowest income group is excluded from the sample? Is income a confounding variable in the relationship between race/ethnicity and gambling behavior?

Problem Gambling

The variable "EVERPROB" was included in the study to count the number of gambling problems respondents report ever having. Gambling problems include items such as stealing money to gamble, being unable to stop thinking about gambling, lying about gambling losses, etc (for a full list see variables D1_ through D19_).

We recoded EVERPROB into two categories: ever had one or more gambling problem, or never had a gambling problem. We called the new variable "LIFEPROB".

Look at the frequency distribution of LIFEPROB. What percentage of gamblers have ever had a gambling problem?

Does the percentage reporting a gambling problem vary by age? Run a crosstab of LIFEPROB and AGE3. What do you see? Are the results as you expected?

Finally, look at the crosstab of LIFEPROB and INCOME. What does the bar chart show? Does problem gambling vary by income?

Note: The online data analysis system (DAS) used on this site uses a system called Survey Documentation and Analysis (SDA), developed and maintained by the Computer-assisted Survey Methods Program (CSM) at the University of California, Berkeley. Documentation for DAS/SDA can be found on their Web site.

Interpretation & Summary

Think about your answers to the application questions before you click through the interpretation guide for help in answering them.

Gambling Behavior

What percentage of sample respondents has ever gambled? Is it higher or lower than you expected?

Is the percentage of people who say they have not gambled in the past year different than the percentage who have never gambled at all?

Which type of gambling appears to be most popular?

Characteristics of Gamblers

In which age group did the largest percentage of respondents report ever gambling?

Is the age pattern different for gambling in the past year compared to lifetime gambling?

Does gambling behavior differ by gender?

Are men or women more likely to have ever gambled?

How does gambling behavior differ by income group?

How might you explain the difference between the pattern of past year gambling by income and the pattern of lifetime gambling by income?

Looking at the bar chart, what racial/ethnic group has the highest percentage of lifetime gambling? What racial/ethnic group has the lowest percentage?

Does the relationship between race/ethnicity and gambling behavior change when the lowest income group is excluded from the sample? Is income a confounding variable in the relationship between race/ethnicity and gambling behavior?

What percentage of gamblers have ever had a gambling problem?

Does the percentage reporting a gambling problem vary by age?

Does problem gambling vary by income?

Interpretation

Things to think about in interpreting the results:

It is important to look at the amount of missing data in each relationship and think about the ways in which that might affect the generalizability of the results.
The numbers in each cell of the crosstabulation tables show the percentage of people who fall into the overlapping categories, followed by the actual number of people that represents this sample. The coloring in the tables demonstrates how the observed number in a cell compares to the expected number if there were no association between the two variables. The accompanying bar charts display the patterns visually as well.
The use of column percentages, as shown in these tables, allows for the comparison of answers to the "outcome" of interest across values of the grouping variable. For example, of those in the highest income group, 17% reported ever having a gambling problem, compared with 11.5% of those in the second highest income group.
Weights (mathematical formulas) are often used to adjust the sample proportions, usually by race, sex, or age, to more closely match those of the general population. The analyses in this guide used weights to increase the generalizability of the findings, so the resulting tables are meant to reflect the relationships we would expect to see in the general population.

The analyses show the following:

About 88% of sample respondents report having ever gambled in their lifetime, and 69% report gambling in the year prior to the survey.
Of all the types of gambling, the highest percentage of respondents report having ever played the lottery (83.6%). The next most popular type of gambling was in a non-tribal casino (65%).
89% of respondents in the 30-49 age group report having ever gambled in their lifetime, compared with 84% of the younger group, and 83% of the older group. However, when only the past year's gambling behavior is considered, the percentage of gamblers is highest in the youngest group (78%), and lowest among those ages 50 and over (70%).
Men (88%) are slightly more likely than women (83%) to have ever gambled.
Individuals in the in the three highest income groups were roughly equally likely to have ever gambled, while a smaller percentage of individuals in the lowest income group (under $24,000) had ever gambled. The crosstab of past year gambling and income shows that individuals in the $50,000-99,000 income group were slightly more likely to have gambled in the past year than other individuals with incomes over $24,000. For all income groups, the percentage who reported gambling in the past year was about 10-15 percentage points less than the percentage who reported ever gambling. It is important to remember that household income was measured for the past year--the same timeframe as past year gambling behavior. Lifetime gambling behavior does not share the same timeframe with the income measure.
Whites report the highest percentage of lifetime gambling (88%), followed by Hispanics (83%), Other (81%), and Blacks (78%).
When the lowest income group is excluded from the sample, Whites and Hispanics have nearly the same rates of gambling (91% and 90%), as do Blacks and Other (83% each). Income appears to be a confounding variable in the relationship between race/ethnicity and gambling behavior.
12.3% of gamblers report ever having a gambling problem. Respondents in the 18-29 age group were most likely to report problem gambling (16%). Of the 50 and older age group, only 10% reported problem gambling.
Problem gambling is most common among the highest income group (those making $100,000 or more per year).

Summary

The goal of this exercise was to explore gambling behavior and characteristics of gamblers in the United States.

Taken together the results show that most Americans gamble at some point in their life, though gambling behavior varies by gender, age, income, and race/ethnic group. The most popular type of gambling is the lottery, followed by non-tribal casinos. The relationship between gambling behavior and race/ethnicity is confounded by income. Because racial minorities are overrepresented in the lowest income group, every minority group appears to gamble less than whites. However, when income is accounted for in the relationship between race/ethnicity and gambling, the results change. This exercise emphasized the importance of thinking carefully about how variables are measured, and to consider possible confounding variables that may influence your results.

Bibliography

We have compiled a list of references that might be useful to instructors and students wishing to further explore this topic. All were chosen because they relate to the topic of study, whether or not they use the specific dataset that was used in this exercise. Some relate directly to the concepts as defined by the exercise, others explore the topic more broadly either conceptually or empirically. For even more resources, try a key word search in the ICPSR Bibliography!

CITATION: Inter-university Consortium for Political and Social Research. Gambling Behavior in the United States: A Data-Driven Learning Guide. Ann Arbor, MI: Inter-university Consortium for Political and Social Research [distributor], 2009-04-16. Doi:10.3886/gambling

This work is licensed under a Creative Commons Attribution-Noncommercial 3.0 United States License.