EC3004 Dissertation
BA Financial Economics
Student Number 149034559
13 March 2017
A Report on
Media Influences on Social Outcomes: The Impact of MTV’s 16 and Pregnant on Teen Childbearing
By Melissa S Kearney and Phillip B Levine.
Abstract:
This report aims to emphasise on the ideas and findings from Kearney and Levine’s article. It will state the empirical methodology used, identify the key findings and how these findings show the effect of the media on social outcomes.
Introduction
The impact of the media, the content it makes available to its viewers, and the potential effect that this content could have on the mindset of the viewers has been a preexistent discussion amongst policy makers.
Teenage pregnancy especially, has been a social problem which policymakers have been trying to decrease the rate of. This is due to children being born to young mothers without tertiary education and lower earning potential tend to be relatively lacking economically, as well as socially, as compared to those born to older parents (Kearney and Levine, 2012a).
In the US, although, there has been a drop in the amount of teenage pregnancies observed, relative to other developed economies, it is still the highest. This drop, according to Kearney and Levine, can be seen in the years from 1991 to 2008, and with an even larger drop from 2008 to 2012. In 2009, MTV premiered their franchise ’16 and Pregnant’ and this coincidently, this was within the period in which the reduction in teenage childbearing rates in the US was highest.
The show ’16 and Pregnant’ was released to show viewers the life struggles teens face having to deal with puberty, growing up as well as becoming parents. In the report, Kearney and Levine believe that the show has been somewhat effective in reducing the rate of conception in teenagers. The show has brought on controversial criticism by others on the effect it has had on teenage childbearing, with some claiming the struggles depicted on screen by the teen mums has discouraged teenagers to get pregnant, whilst some others believe that the fame and popularity proves an incentive to other teenagers to want to get pregnant (Jonsson, 2010)
In their paper, Kearney and Levine sought to find out if being exposed to the content in the ’16 and Pregnant’ franchise affect the views of the teenagers towards teenage pregnancies. The phrase ’16 and Pregnant’ is used by them in this article as an umbrella term for the show and all of its spinoffs. Also, they studied whether the show peaked the interest of teenagers in contraceptive use or abortion, and how much, if any, of an effect, the show’s introduction had on the rate at which teenagers got pregnant.
The following sections of this report will thus show, and detail the various sets of data used, the experimental methodology that Kearney and Levine used to obtain their results, summarise their findings and in conclusion make a general summary of the ideas and results contained in the article.
Method
A. DATA
K-L made use of data from varied sources within a period from 2009-2012; Nielsen ratings which are the primary rating system for television shows to view how highly rated the show was in various DMAs (Designated Market Areas) which are different locations which receive signals from a particular television tower in a market), Google trends and Twitter record used to search for the frequency in which people searched for the phrase ’16 and pregnant’ and also after the show aired how often people searched for abortion and contraceptive terms, and Vital Statistics natal data, which gives data on births in the US, .
Limitations existed in the use of these various data sources. With the Nielsen ratings, the improvement in technology had reduced the ability of the company to monitor television viewings, as alternative means such as online streaming have reduced the number of people watching live showings. The Google trends and twitter records did not contain information about demographics of the individuals, and their locations was difficult to determine.
B. EMPIRICAL METHODOLOGY:
1. Analysis of Teen Births
Kearney and Levine made use of the Ordinary Least Squares regression model to explain the relationship between teenage birth rates and ratings of the show from when it was introduced, taking into account the geographic variations from the different DMAs
1) ln(Bjt) = β0 + β1Rate16Pj ×pXXXost + β2Ujy + Xjyγ + θt + δjs + εjt
where t represents yearly quarters, j represents DMAs and s represents different seasons in a year. Ln(B) depicts the natural logarithm of teenage births. Fixed effects were included for t and j x s, in form of θt, and δjs respectively, to control for the variability in birth rate seasonality.
The main independent variables in this model are Rate16P, which is used to represent ratings given to the show by their sample age group, and post, which denotes yearly quarters observed after the show was introduced in 2009. Ujy signifies average unemployment rate, and Xjy represents different ethnicities.
There was the potential issue of endogeneity of between the rating variable and birth rates, and their possible correlation with unobserved factors. K-L believed this could result in there being a bias in their results, and hence made use of an IV regression model to correct for this. They included exogenous variable, MTV0809, to account for ratings of MTV shows from 2008 to 2009. This variable helps to
Following the two-stage regressions done in the IV framework, they reduced the equation to equation 2, gotten from equation 4 (Kearney and Levine, 2015), with the variables still as described above.
2) ln(Bjt) = β0 + β1 MTV0809j × post + β2Ujy +Xjyγ + θt + δjs + εjt
To test the validity of their IV, K-L observed a linear relationship existing between MTV Ratings from 2008 – 2009, and 16 and Pregnant ratings. This proved that a monotonic relationship existed between both variables, verifying their monotonicity assumption. Also, from regressing both variables against each other, they obtained a t-statistic of 8.0, with a correlation coefficient of about 0.5, and the f-statistic on their omitted instrument of 48.1, showing the strength of the IV used.
To conclude their validity testing, K-L made use of an event study, to view how much of a change was experienced in the behaviour of teenagers following the premiering of the 16 and Pregnant franchise. Supposing changes in behaviour were observed from before the show was introduced, then it might disprove the causal relationship between changes in behaviour and the introduction of the show.
2. Analysis of Google Search and Twitter Activity
K-L made use of the Google trends and twitter searches to search for spikes in google searches and tweets about the show, and also about contraceptives or abortion whenever new episodes are aired. This was to confirm the causal effect of the introduction of the show on behavioural outcomes of teenagers. By examining high-frequency data, which is weekly data for google trends, and daily data for twitter searches, they were able to formally present their analysis using the Ordinary Least Squares regression model:
searchBCt = β0 + β1 NewReleaset + Xt γ + εt
In which searchBC represents a search for birth control, NewRelease represents periods when new episodes are released, and X depicts monthly fixed effects and quadratic trends.
From their Google trends data, they consider the time period between the years 2009 and 2012, with the focus on differences between weeks where episodes are aired, and not aired. Seasonal fixed effects are used to control for differences due to changing seasons. With twitter data, the focus is placed on those weeks when a season is airing. They examined the data, with an added lagged period of 24 hours after the release of an episode.
They further regressed searches about the show against searches for birth control, to test for if the increased exposure to the show’s content, led to increases of searches for terms related to birth control. This could provide evidence as to whether the show could alter factors leading to pregnancy.
In a brief analysis, K-L made use of low-frequency data with geographic variation to check how the show’s introduction altered the searches made about birth control in different states comparatively. A similar regression model was used, with variables only subjected to geographical changes, but with the addition of state and period fixed effects.
RESULTS
This section will discuss the results that Kearney and Levine found from their empirical studies. It is broken down into two sections to discuss the different results gotten from their analysis on tweets and searches, and their analysis on teen birth rates.
A) TEEN BIRTH RATES
K-L based this aspect of their analysis on the direct link between teenage pregnancy rates and the exposure of teens to 16 and Pregnant. From the regressions done, they were able to get the estimates shown below in table 1. An increase in the ratings for the show by a point would lead to the reduction in the birth rate by 1.02 percent.
From their instrumental variables model, the first stage regression is shown in the second column, and it shows a positive correlation between the MTV ratings from the year before the 16 and Pregnant show premiered and the show’s ratings. The IV estimates in the third column show a reduction in the teen birth rate by 2.37 percent when the show’s ratings increase by a point. All results are statistically significant.
Dependent Variable
OLS
Ln(B)
First Stage
Rate16P
IV
Ln(B)
Reduced form
Ln(B)
Rate16P
-1.020
(0.552)
-2.368
(0.942)
MTV0809
Unemployment Rate
-1.440
(0.401)
-1.487
(0.375)
-1.485
(0.432)
F-statistic on omitted instrument
48.1
TABLE 1, gotten from Kearney and Levine (2015)
They graphically plotted results from their event study analysis showing estimates from their IV regression model. Changes in birth rate with increases in MTV ratings were plotted against quarterly periods from 2005 to 2010 to measure the impact of the introduction of the 16 and pregnant show. They passed a best fitting line to make their results clearer. Before the show began, coefficients were positive but not statistically significant, but after the show started, coefficients became negative, and we can see a significant drop in the estimates in the quarter when the show was introduced seen on the graph as 2008: II – 2009: II. The graph supports a causal effect of the show’s introduction in teen birthrate changes.
Next, from the reduced form IV model, they performed regressions on ratings and unemployment rate against teenagers (15-19), as well as older women (20-34). Results they got for women between 20 – 24 were similar to those from 15 -19. As the age group got older, the point estimates they observed from the regressions declined, with that of the 30-34 years’ group being statistically insignificant.
The last three columns gave point estimates for teens (15-19) distinguishing them by race. A point increase in ratings led to a 2.4 percent decrease in white birth rates, a 0.14 percent decrease in Black birth rates, and a 3.782 percent decrease in Hispanic birth rates. Nevertheless, the results they obtained from the various ethnicities were statistically insignificant.
In general, they were able to deduce that the show 16 and Pregnant had a substantial effect on teenage pregnancy rates. Their results from their IV model regressions show that the 16 and Pregnant show had a significant effect on teenage birth rates and from their graphical presentations, the impact on birth rates from the show’s introduction can be visualised. Finally, the show was seen to have a greater effect on its youth audience, who were the target audience, than on older women. K-L’s results show evidence of a causal effect of the show on teenage childbearing
From Table 1, they obtained from their IV model that a point increase in ratings for the 16 and Pregnant show led to 2.47 percent decrease in teenage birth rates. They also noted* that ratings for the 16 and Pregnant show averaged about 1.8 points. By multiplying this figure by the 2.47 percent obtained in table 1, they were able to obtain that the 16 and franchise show contributed to a 4.27 percent reduction in teenage pregnancies between 2009 and 2010. As teenage pregnancy fell from 16.62 to 12.05 percent between fall of 2008 and fall of 2010, Kearney and Levine predicted that 24.3 percent of this drop was credited to the show.
The point estimates they obtained in table 1 also reveal the importance of unemployment towards teen birth rates. The results shown show a reduction in teenage births when there was an increase in unemployment rate, showing that employment plays a role in reducing the rate at which teens give birth. In the period of the Great Recession, at about 2008, unemployment rates rose greatly, showing why there was a large reduction in the rate of teenage births.
B) SEARCHES AND TWEETS
In their analysis of high-frequency data, they found spikes in tweets and searches as soon as a new episode is aired. These can be viewed graphically below, using weekly data from the start of 2009 and the end of 2012. Spikes can be seen in the highlighted regions of the graph which represent the weeks that the episodes aired.
They also plotted a graph from daily data, with similar results, only that the peak in daily searches and tweets occurred just a day after the show is aired. They reasoned this to be because the target audience watches the episodes at night, only to discuss it the following day with their friends.
In table 3, they examined the impact of a new release of an episode on searches and tweets relating to birth control and abortion. From the data on twitter, they obtained that abortion-related tweets rose by 14.2 percent on the day a new episode is released, and by a further 21.2 percent the day after. Also, tweets related to birth control rose by 12 percent on the day of release, and a further 23 percent the day after the episode is released. The other google searches relating to birth control and abortion were found to be statistically insignificant.
Finally, with Table 4, on the higher panel, they showed the results of their analysis of national level high frequency data with “regression models where the dependent variable reflects search/tweet activity for terms related to unprotected sexual activity and the key independent variable reflects search/tweet activity mentioning 16 and Pregnant “(Kearney and Levine, 2015 pg. 3624).
Panel A shows estimates of a statistically significant increase in Google searches relating to ‘how to get birth control’ when there is an increase in Google searches for ‘16 and Pregnant’. Also, from the twitter data on the right side if the panel, Kearney and Levine obtained that “the elasticity between tweets containing “birth control”/” abortion” and tweets containing 16 and Pregnant is 0.077/0.064” (Kearney and Levine, 2015 pg. 3625). These results were statistically significant
In the lower panel, which showed estimates from their analysis on the state level with lower frequency data they found that there were increases in searches for how to birth control, as well as searches on getting an abortion when searches for the 16 and Pregnant show increased. Twitter searches related to birth control also increased by 13.7 percent when there was an increase in tweets about the show.
All of these provide supportive evidence towards their belief that the introduction of the show provided incentive against childbearing, by providing a glimpse into the thoughts of teenagers through their tweets and google searches
Conclusion
Kearney and Levine’s sought mainly to find out the effect that the introduction of the popular MTV show ’16 and pregnant’ had on the teenage pregnancy rates in the United States. They found various results and provided evidence supporting how exposure to the show had played a part in reducing teenage pregnancy figures.
To avoid bias that could come about from possible endogeneity between ratings and the teen childbearing rates, Kearney and Levine made use of the IV regression model, and from their analysis deduced that the 16 and Pregnant franchise led to a 4.3% in teenage pregnancy between the summer of 2009 and the end of 2010. They also analysed data from google searches and twitter searches, which gave enough evidence to support the claim that the show affected teenagers’ mindsets about pregnancy, contraceptive use and abortion. Kearney and Levine acknowledged data limitations, preventing them from measuring the effect of contraceptives and abortions towards the reduced teen pregnancy rates.
Although not created as a social campaign to discourage teenage pregnancies, K-L believe this is how the show has worked itself out to be, from its showings of the various hardships faced by pregnant teenagers, and teenage parents. This in a sense can be viewed as a useful economic policy against teenage pregnancies.
Their article’s main purpose was to examine the effect of the media on social outcomes, and they were able to show this through MTV’s ability to draw the attention of teenagers, in such a way to affect their attitudes. Although general concern is for the possible negative effects the media could have, their article has shown that the media exposure can also yield positive outcomes. All in all, the authors have clearly shown that the media “has the potential to be a powerful driver of social outcomes” (Kearney and Levine, 2015 pg. 3626)
References
• Jonnson, Patrick. “A Force Behind The Lower Teen Birthrate: MTV’s ’16 And Pregnant'”. The Christian Science Monitor. Available at http://www.csmonitor.com/USA/Society/2010/1221/A-force-behind-the-lower-teen-birthrate-MTV-s-16-and-Pregnant. Web. 13 Mar. 2017.
• Kearney, Melissa S and Phillip B. Levine. 2015. “Media Influences on Social Outcomes: The Impact of MTV’s 16 and Pregnant on Teen Childbearing.” American Economic Review, 105(12): 3597-3632.
• Kearney, Melissa S and Phillip B. Levine. 2012. “Why Is the Teen Birth Rate in the United States So High and Why Does It Matter?” Journal of Economic Perspectives, 26(2): 141-63.