[charts and tables have been removed for formatting reasons.]

Pornography’s Impact on Marriage:
An Econometric Approach

Introduction

The growth of the Internet has lowered the costs, increased the accessibility, and furthered the anonymity of one of males’ favorite vices—pornography. Although the discussion of pornography has long been considered taboo, its effects are now coming under greater scrutiny as both its use and acceptance are growing quickly. From a study by Adult Video News, the adult entertainment industry made approximately $12.6 billion in revenue in 2005. More than 70% of men from ages 18 to 34 visit a pornographic site at least once in a typical month (comScore). In addition, according to the American Academy of Matrimonial Lawyers, the Internet was a significant factor in two out of three divorces in 2003, with excessive use of online pornography contributing to over half of such cases. While some with more traditional or religious values are quick to oppose pornography, its use is becoming more accepted by society as a whole and received especially well by young adults (Carroll). And when you add to this the proponents of free speech who see any attempt to regulate pornography as a violation of one’s rights and those who believe pornography serves as an outlet and actually reduces the number of rapes and violent crimes, you’ve got yourself quite the debate. Because pornography consumption has recently been increasing at an unparalleled rate, researchers are starting to examine potential consequences of its use. This has resulted in the emergence of new datasets and studies that have enabled us to look at the effects of pornography on the user, surrounding family members and friends, and society as a whole.

This paper addresses the impact of pornography on marriage. Since pornography consumption is significantly higher for men than for women (though the gap seems to be shrinking), I specifically look at the effects of pornography on married men. If the results indicate that pornography significantly diminishes the quality of one’s marriage and leads to infidelity, then they could be used as evidence against its distribution and for its regulation.

Literature Review

Most of the research on pornography deals with either the causes of pornography use or its effects. In 2005, Buzzell found that males, young people, and residents of urban areas are significantly more likely to view pornography and that access to technology such as movie theaters, VCR’s, and the Internet plays a role in one’s pornography consumption. More recently, Carroll studied pornography acceptance and use among young adults and concluded that “roughly two thirds of young men and one half of young women agree that viewing pornography is acceptable, whereas nearly 9 out of 10 young men and nearly one third of young women reported using pornography.” He also found a link between pornography acceptance and use and young adults’ “risky sexual attitudes and behaviors, substance use patterns, and nonmarital cohabitation values.”

Padgett studied the effects of pornography use on one’s attitudes toward women and found the results to be inconclusive. Most of the studies involving pornography and attitudes toward women seem inconclusive. While some show that pornography use leads to a degrading view of women, others indicate that it actually helps the individual have a more favorable attitude toward women. In 1988, Demare performed a study on 222 undergraduate males and found that viewing violent pornography is positively correlated with the likelihood of rape or using sexual force. In 2004, Wongsurawat found a positive correlation between pornography use and rape and divorce from his OLS models, but after he employed instrumental variables the correlation between pornography and rape turned negative and the statistical significance for the relationship between pornography and the rate of divorce disappeared. In 2006, Kendall utilized the arrival of the Internet to study the link between pornography use and rape and concluded that they are actually substitutes.

In her 2006 review on the research done on “the impact of Internet pornography on marriage and the family,” Manning draws from previous studies to show that pornography consumption is associated with increased marital distress; decreased marital intimacy and sexual satisfaction; infidelity; and a devaluation of monogamy, marriage and child rearing. These conclusions seem to be pulled from a wide variety of studies, as not much research has been focused on the effects of pornography on marriage.

Models

In this study, I estimate four models relating to marital satisfaction and infidelity.

“Marital Trouble Model”

troublewithsp = β0 + β1xmovie + β2everdivorced + β3numkids + β4age + β5educ + β6black +

β7otherrace + β8protestant + β9catholic + β10jewish + β11otherreligion +

β12churchpermonth + β13incin1000s + β14hrsworked + β15agewed + β16agesq +

β17educsq

troublewithsp: whether the husband reported having “serious trouble” with his wife over the last

year (1) or not (0).

xmovie: whether the husband viewed an X-rated movie in the last year (1) or not (0).

everdivorced: whether the husband has ever been divorced (1) or not (0).

numkids: number of children the husband has ever had.

age: age of the husband.

educ: highest year of school the husband completed.

black: whether the husband is black (1) or not (0).

otherrace: whether the husband is some other race (1) or not (0).

protestant: whether the husband’s religious preference is Protestant (1) or not (0).

catholic: whether the husband’s religious preference is Catholic (1) or not (0).

jewish: whether the husband’s religious preference is Jewish (1) or not (0).

otherreligion: whether the husband has some other  religious preference (1) or not (0).

churchpermonth: number of times the husband attends religious services per month.

incin1000s: the husband’s family income in constant dollars (base year is 1986).

hrsworked: number of hours the husband worked last week.

agewed: the husband’s age when first married.

agesq: age squared.

educsq: educ squared.

This model is estimated using a probit. The baseline race group is white, and the baseline religious preference group is “no religion.” Since troublewithsp has recent values for only 1991 and 2004, I estimate this model separately for the two years. The different estimates may provide evidence for a change in the determinants of marital trouble between 1991 and 2004. Since pornography has become much cheaper and more accessible since 1991 (due to the growth of the Internet), the results of estimating this model for the two years may shed some light on the effect pornography has on marital strife.

“Attitude toward Adultery Model”

adulteryalwayswrong = β0 + β1xmovie + β2numkids + β3everdivorced + β4age + β5educ + β6black

+ β7otherrace + β8protestant + β9catholic + β10jewish + β11otherreligion + β12churchpermonth + β13incin1000s + β14hrsworked + β15agesq

+ β16educsq + controls for years

adulteryalwayswrong: whether the husband thinks adultery is “always wrong” (1) or “almost always wrong,” “sometimes wrong,” or “not wrong at all” (0). 76.4% of those surveyed responded “always wrong.”

controls for years: one dummy variable for each year with observations: 1993, 1994, 1996, 1998, 2000, 2002, 2004, 2006. The baseline year is 1991.

This model is estimated using a probit. See above for descriptions of the other variables.

“Number of Sexual Partners Model”

numsexualpartners = β0 + β1xmovie + β2everdivorced + β3age + β4educ + β5black

+ β6otherrace + β7protestant + β8catholic + β9jewish + β10otherreligion

+ β11churchpermonth + β12incin1000s + β13agewed + β14agesq + β15educsq

+ controls for years

numsexualpartners: number of females the husband has had sex with since his 18th birthday.

This model is estimated using OLS with the robust option to correct for heteroskedasticity. See above for descriptions of the other variables.

“Committed Adultery Model”

adultery = β0 + β1numsexualpartners + β2xmovie + β3numkids + β4everdivorced + β5age + β6educ

+ β7black + β8otherrace + β9protestant + β10catholic + β11jewish + β12otherreligion

+ β13churchpermonth + β14incin1000s + β15hrsworked + β16agewed + β17agesq

+ β18educsq + controls for years

adultery: whether the husband has ever had sex with someone other than his wife while married (1) or not (0).

This model is estimated using a probit. See above for descriptions of the other variables.

These four models estimate the effects of having watched an X-rated movie in the past year on variables relating to marital trouble, the husband’s attitude toward adultery, the number of sexual partners he has had since age 18, and whether he has ever committed adultery. The controls are selected to account for religiosity, marital and familial factors, education level, work and income related variables, and age—all of which are likely to contribute to the dependent variables of interest. agesq and educsq are included to see if the age and education of the husband have constant, increasing, or decreasing effects on the dependent variables. I also control for different effects of the independent variables in different years. xmovie is the main explanatory variable of interest in the four models, and though it measures the use of only X-rated movies, it is the best available variable for pornography use. Many who view pornography on the Internet or via some other medium are likely to also watch X-rated movies (often on the Internet itself), and thus this variable should capture much of the pornography consumption of married men. There is also some concern that the recent change in the “X” rating to apply almost entirely to pornographic movies will have a carryover effect on xmovie in the early 1990’s, as this change began in the late 1980’s and officially occurred in 1990.

numsexualpartners is also a key independent variable for the “committed adultery model.” If pornography use is shown to be correlated with an increased number of sexual partners and an increased number of sexual partners is shown to be linked with greater likelihood of having committed adultery, then pornography use may have both a direct and indirect (through numsexualpartners) effect on the likelihood of a husband having committed adultery.

Data

The dataset used for this study is the General Social Survey, which began in 1972 and completed its 26th round in 2006. The GSS is conducted every one or two years and contains a standard core of questions that remain constant from year to year and then topics of special interest that vary. Its website claims, “except for the U.S. Census, the GSS is the most frequently analyzed source of information in the social sciences” (http://www.norc.org/GSS+Website). The target sample size per year is 1,500, and the total number of observations since the survey’s inception is over 51,000. However, since some questions are asked not every year but only once or twice, it can be difficult to obtain a large sample size from the dataset for certain models. When obtaining the data, I drop all observations with missing values for the variables of interest.

Hypotheses

I test the following hypotheses: whether pornography use—as measured by having seen an X-rated movie in the past year—has any effect on (1) the probability a husband has had serious trouble with his wife in the past year, (2) the probability he believes adultery is always wrong, (3) the number of sexual partners he has had since age 18, and (4) the likelihood he has committed adultery. I also test (5) whether the number of sexual partners a husband has had since age 18 has any effect on the likelihood he has committed adultery. The null hypothesis for all five cases is that the marginal effect is zero, or, that there is no effect of pornography use or number of sexual partners (for the fifth hypothesis) on the dependent variable of interest. These will all be tested at the 5% level.

Empirical Results

“Marital Trouble Model”

Using a probit model, the results of the “Marital Trouble Model” for 1991 are presented in Tables 1 and 2, and the results for the “Marital Trouble Model” for 2004 are presented in Tables 3 and 4 (after eliminating independent variables with a z-statistic of 0.5 or less that were not considered important for the model). (The linear probability model alternatives for the 1991 and 2004 models are found in Appendices A and B, respectively.)

Table 1. “Marital Trouble Model” for 1991 Summary Statistics

Table 2. “Marital Trouble Model” for 1991 Probit Results

Table 3. “Marital Trouble Model” for 2004 Summary Statistics

Table 4. “Marital Trouble Model” for 2004 Probit Results

In the 1991 model, 31.3% of the married men surveyed viewed an X-rated movie in the past year and 4.82% reported having had serious trouble with their spouse in the past year, while in the 2004 model, 29.1% viewed an X-rated movie and 6.06% reported having had serious trouble with their spouse. The most telling difference between the two probit models is the large and statistically significant marginal effect for xmovie in the 2004 model as opposed to the smaller and statistically insignificant marginal effect in the 1991 model. According to the 2004 model, having watched an X-rated movie in the past year is associated with a 213% increase (12.9 percentage-point increase) in the probability that a husband has recently had serious trouble with his spouse. The 1991 model, however, provides inconclusive results for pornography’s effect on marital strife. There are at least two possible reasons for this difference: (1) the term “X-rated movie” may have still applied to some violent but non-pornographic movies in 1991, which would result in xmovie not capturing the true effect of pornography use, or (2) due to the spread of the Internet, pornography consumption may have been more widespread in general and more extreme among its users in 2004, which would explain why its use—as measured by xmovie—had a large and statistically significant effect in 2004 but not in 1991. Though the average of xmovie is slightly lower in 2004 than in 1991, real pornography use likely increased during this time period because of the Internet. The higher average in 1991 may be explained by the reclassification of the rating “X” and by the possible substitution of Internet pornography (that may not have been considered “X-rated movies” by its users) for the typical X-rated movie of 1991. Also, it is likely that those who did view pornography did it more intensively in 2004 than in 1991, as a consequence of the decreased costs—both monetary costs and costs relating to the potential embarrassment of being caught purchasing the pornographic material—and the increased accessibility of pornography.

In the model for 1991, none of the explanatory variables are significant at the 5% level, and only educ is significant at the 10%. This model estimates an 11.8% reduction in the likelihood that a husband has had serious trouble with his spouse in the past year for each additional year of education. Also, otherrace and jewish are dropped because there are not enough observations, and an f-test for joint significance of the religion variables does not show significance (p=.4704).

In the 2004 model, both xmovie and everdivorced are significant at the 5% level. As reported above, having watched an X-rated movie in the past year increases the probability that a husband has recently had serious trouble with his spouse by 213%, and having been divorced relates to a 122% increase in the probability that a husband has had recent serious trouble with his spouse. jewish is once again dropped because of an insufficient number of observations, and an f-test for joint significance of the religious groups does not conclude they are jointly significant (p=.5369).

“Attitude toward Adultery Model”

Using another probit model, the results of the “Attitude toward Adultery Model” are presented in Tables 5 and 6 (after eliminating independent variables with a z-statistic of 0.5 or less that were not considered important for the model). (The linear probability model alternative is found in Appendix C.)

Table 5. “Attitude toward Adultery Model” Summary Statistics

Table 6. “Attitude toward Adultery Model” Probit Results

As seen in Table 5, 80.0% of the married men in the sample believe adultery is always wrong. A larger sample size (compared to the previous models) helps with the statistical significance of many of the independent variables. xmovie has the greatest negative marginal effect in this model and is almost the most statistically significant. The marginal effect indicates that having watched an X-rated movie in the past year decreases the probability that a married man thinks adultery is always wrong by 12.6%. Other notable and statistically significant explanatory variables that decrease the probability that a married man thinks adultery is always wrong are educ and incin1000s. Education has a practically and statistically significant effect, as each additional year of education decreases the probability of thinking adultery is always wrong by 2.30 percentage-points, while the husband’s family income does not appear to have a practical effect. Being Protestant and attending religious services both increase the probability of thinking adultery is always wrong, and both portray practical and statistical significance. Having a Protestant religious preference is correlated with a 16.4% greater likelihood of believing adultery is always wrong, and each additional time a married man attends religious services per month leads to a 5.14% increase in the probability of believing adultery is always wrong. The positive marginal effects of the statistically significant year variables seem to indicate that more married men believed adultery is always wrong as time progressed from 1991 (the baseline year) to 2006.

F-tests for joint significance show that the year controls (p=.0138) and religion variables (p=.0003) are jointly significant at the 5%, while the race groups (p=.1160) are not. This implies that taken together, both the religion of the husband and the year in which he answered the survey have an effect on his attitude toward adultery, while his race has no such effect.

“Number of Sexual Partners Model”

Running an OLS regression with the robust feature to correct for heteroskedasticity (which was detected using the Breusch-Pagan / Cook-Weisberg test), the results of the “Number of Sexual Partners Model” are presented in Tables 7 and 8 (after eliminating independent variables with a t-statistic of 0.5 or less that were not considered important for the model). (The non-robust OLS alternative is found in Appendix D.)

Table 7. “Number of Sexual Partners Model” Summary Statistics

Table 8. “Number of Sexual Partners Model” OLS Regression Results

Approximately 8.43% of the variation in numsexualpartners is explained by the independent variables of this model. The average number of sexual partners a married man has had since age 18 is 11.07 in this sample, and xmovie, everdivorced, age, and black are all correlated with a greater number of sexual partners and significant at the 5% level. Having viewed an X-rated movie in the past year relates to a 3.90 increase in numsexualpartners, having been divorced relates to a significant 11.1 increase, and each additional year of the husband’s age is associated with a 0.596 increase in the number of sexual partners he has had since age 18. numsexualpartners increases at a decreasing rate as the husband ages, however, as seen in the negative and statistically significant coefficient on agesq. Using the estimates for age and agesq, the age at which numsexualpartners reaches its maximum is 52.4. educ and jewish are significant at the 10% level.

The only other statistically significant variable that has a negative effect on numsexualpartners is churchpermonth. For each additional time a married man attends religious services per month, the number of sexual partners he has had since age 18 goes down by 1.16. F-tests for joint significance of the religion variables (p=.0248) and the race variable (p=.0080) both show joint significance.

“Committed Adultery Model”

Using one final probit model, the results of the “Committed Adultery Model” are presented in Tables 9 and 10 (after eliminating independent variables with a z-statistic of 0.5 or less that were not considered important for the model). (The linear probability model alternative is found in Appendix E.)

Table 9. “Committed Adultery Model” Summary Statistics

Table 10. “Committed Adultery Model” Probit Results

13.8% of the sample reported having had committed adultery in their lifetime. The most statistically significant explanatory variables are xmovie and numsexpartners. Having viewed an X-rated movie in the last year corresponds with a 115% increase in the likelihood that the husband has committed adultery, and each additional sexual partner he has had since age 18 leads to a 2.09% increase in the likelihood that he has committed adultery. These are both significant at the 1% level, suggesting that pornography use may have both a direct and indirect (through numsexualpartners) effect on the likelihood of a husband having committed adultery.

Other variables that increase adultery at the 5% are age and otherreligion. The marginal effect of otherreligion is considerably large—being of a religion other than Protestant, Catholic, or Jewish is associated with a 348% increase in the probability that the married man has committed adultery over those who are of “no religion.” This result may be inaccurate due to the fact that only 19 “other religion” individuals were part of the sample. For each additional time a married man attends religious services per month, the likelihood that he has committed adultery in his lifetime decreases 15.0%. F-tests for joint significance indicate that the religious groups (p=.0076) are jointly significant, while the religious groups (p=.8080) are not.

Conclusions

The major findings of this study are that pornography use is associated with a 213% increase in the probability that a married man has had serious trouble with his spouse in the past year, a 13% decrease in the probability that he thinks adultery is always wrong, a 35% increase in the number of sexual partners he has had since age 18, and a 115% increase in the likelihood that he has committed adultery. The effects of these four consequences of pornography use on the user himself, his spouse and family, and society as a whole may include a lower level of marital satisfaction, an increased likelihood of divorce, a greater risk of obtaining or passing STD’s, unwanted pregnancies, and child abuse or neglect.

Though this work is far from comprehensive, it identifies key relationships between pornography consumption and marital strife and infidelity that can lead to future studies on the subject. Some major obstacles future researchers have to overcome are endogeneity problems (xmovie may be endogenous is some of my models, but I was unable to find an appropriate instrumental variable to correct for it) and the lack of datasets with more than a few pornography related variables and less chances for measurement error. As better datasets become available and more research is done on potential instrumental variables, additional effects of pornography use will likely be discovered—some of which may help change how pornography is distributed and regulated, how firms respond to the pornography problem in the workplace, and how husbands and wives react to the threat that pornography can pose to their marriage and family.

References

Buzzell, T. (2005). “Demographic characteristics of persons using pornography in three

technological contexts.” Sexuality & Culture. 9, 28-48.

Carroll, Jason; Laura Padilla-Walker, Larry Nelson, Chad Olson, Carolyn McNamara Barry, and

Stephanie Madsen (2008). “Generation XXX: Pornography Acceptance and Use Among

Emerging Adults.” Journal of Adolescent Research. 23; 6.

Demare, D.; Briere, J; and Lips, H. (1988). “Violent pornography and self-reported likelihood of

sexual aggression. Journal of Research in Personality.” 22, 140-153.

Kendall, Todd. (2006). “Pornography, Rape, and the Internet. Clemson University, Working

Paper.”

Manning, Jill (2006). “The Impact of Internet Pornography on Marriage and the Family: A

Review of the Research.” Sexual Addiction & Compulsivity. 13 (2-3), 131-165.

Padgett, Vernon; Brislin-Slutz, Jo Ann; and Neal, James A. (1989). “Pornography, Erotica, and

Attitudes toward Women: The Effects of Repeated Exposure.” The Journal of Sex

Research . 26, 479-491.

Wongsurawat, Winai. (2004) “Pornography and Social Ills: Evidence from the Early 1990s.

Journal of Applied Economics.” 9 (1): 185-213

ppendix A: “Marital Trouble Model” for 1991 LPM Results

Appendix B: “Marital Trouble Model” for 2004 LPM Results

Appendix C: “Attitude toward Adultery Model” LPM Results

Appendix D: “Number of Sexual Partners Model” Non-Robust OLS Regression Results

Tagged with:
 

Comments are closed.