An article by the NPD Group, a leading market research company, states that today, 91 percent of kids in the United States aged 2-17 play video games. This is an increase of 9 percentage points when compared to their 2009 survey results.  Is this a statistically significant increase?

I did some research and was able to find the methodology used for NPD’s results.  2011’s survey utilized 4,136 individuals ages 2-17, whereas 2009’s results cited "over 5,000 members."  Hmm. Well, I guess I have to assume that it’s a number close to 5,000, so we’ll round down to 5,000.  Because the articles provide the total participants and the end percentages, I needed to back-calculate to get the numerator of this fraction. The results are displayed below:

With this information in hand, I was ready to see if this represented a statistically significant increase. In Minitab, I chose Stat->Basic Statistics->2 Proportions. I filled out the dialog as shown below:

I got the following results:

Test and CI for Two Proportions

Sample     X     N  Sample p
1       4100  5000  0.820000
2       3764  4136  0.910058

Difference = p (1) - p (2)
Estimate for difference:  -0.0900580
95% CI for difference:  (-0.103821, -0.0762949)
Test for difference = 0 (vs not = 0):  Z = -12.82  P-Value = 0.000

Fisher's exact test: P-Value = 0.000

Because our P-value is below our alpha (set at 0.05), we reject the null hypothesis that both proportions are equal. In other words, there is a statistically significant difference in the results of this year's survey.

Let’s say that another study is done a couple of years down the road. What proportion would you need to achieve to call it statistically significant against a baseline proportion of 0.91? With those 2011 survey statistics in mind, let’s fill out the dialog box for  Power and Sample Size for 2 Proportions. I am assuming a sample size of 4136 again and a power of 0.95 (It's probably unrealistic that they’d use precisely the same sample size, but let’s assume for the sake of argument, anyway.)

I got the following results:

Power and Sample Size

Test for Two Proportions

Testing comparison p = baseline p (versus not =)
Calculating power for baseline p = 0.91
Alpha = 0.05

Sample
Size  Power  Comparison p
4136   0.95      0.931411
4136   0.95      0.886019

In order for the test to be significant, the comparison proportion would either have to be >= 0.93144 or <= 0.886019.

I found this statement from the 2009 survey results very interesting:

The decline in teen usage of video games is likely due to diversifying, maturing interests, which translates into stiffer competition for their mind and wallet share,” said Anita Frazier, industry analyst, The NPD Group. “In addition to competition from other areas of the entertainment space, more school work, activities, and parent-imposed time limits on gaming are factors which the data suggests may be contributing to this dip in older teen engagement.

What happened in 2 years? The article talks about how mobile devices may have had a stake in this, due to the sheer availability in both smartphones and content (apps).  Most notable in this 2011 survey is the percentage increase of kids gaming from ages 2-5: 12.68 percent.  This does make some sense to me though.  More smart phones = more video games to be played on smart phones. However, these games aren’t as complex as the ones you find on a console system like a Nintendo or Sony PlayStation.  These games might be at the level which a 2 to 5 year old child can comprehend, thus the spike in video game use for that age group.  This is all speculation, however.

Whether you're a gamer or not, I hope that you enjoyed this article!

Sources
(1) http://www.npdgroup.com/wps/portal/npd/us/news/pressreleases/pr_111011
(2) https://www.npd.com/lps/PDF_SpecialReports/Kids-and-Gaming-2009.pdf