Reading Survey Results

As part of my graduate studies in Business Analytics at the University of Arkansas (Fayetteville), I was given in a homework assignment in which I was given my second experience conducting a survey.  My first experience was almost a year ago when my boss tasked me with surveying and analyzing the retention metrics for a group of eighty four military members in our squadron.  I did my best with Adobe forms, excel spreadsheets, and a basic understanding of freshman statistics.  Needless to say, I handled this survey with a little more of a systematic approach.

The Sample Demographics:

I sent this survey to my fellow students (17), professor (1), coworkers (85), and Google Plus audience (10,487).  Given the normal Google Plus penetration rate of approximately 2%, that brings the people invited to survey to an estimated 313 people; of which, 43 (or 14%) responded.  The sample demographics were as follows:

  • 63% Male
  • 47% were under 35
  • 66% were college graduates
  • 44% of the college graduates were in STEM fields
  • About half (56%) of the respondents had children
  • Most (72%) were employed full time
Reading Survey 2.jpg
Reading Survey 3.jpg

The Reading Habits:

The results from the survey showed that lifetime reading habits were normally distributed with a positive skew as the result of 42% of respondents indicating that they had read more than 500 books in their lifetime.  Their reading habits were skewed heavily (47%) towards recreational reading with a standard deviation of 27.87.  Given that the responses for educational (34%; 23.16) and professional (16; 12.55) also had high deviations, this would indicate that some respondents only read educationally or recreationally, causing a high level of variation between responses.

There was more deviation in the results for last year’s reading habits, with 51% of respondents reading fewer than 10 books, and the motivation of reading being even more difficult to pin down without deeper analysis.

Reading Survey 1.jpg

When respondents were asked to rank their favorite genres, the results showed a high number of respondents (across all age groups) selecting Science Fiction and Historical Fiction as being the most enjoyable books to read.  Steampunk and Supernatural scored very low for all age brackets. 

  • Under 25 – Mystery, Young Adult, Mystery
  • 26-30 – Science Fiction, History, (Auto) Biography
  • 31-35 – Science Fiction, Historical Fiction, History
  • 36-40 – Historical Fiction, Educational, Science Fiction
  • 41-45 – Historical Fiction, (Epic) Fantasy, Mystery
  • 46-50 – Science Fiction, Mystery, Historical Fiction
  • 50+ - Science Fiction, Historical Fiction, History

 While there was some variation in favorite genres by age, there was substantially more variation when genre preference was broken out by gender:  Men preferred Science Fiction, Historical Fiction and History; whereas women preferred Mystery, Horror, and Science Fiction.  This trend held true throughout all age brackets.

While there was some variation in the ranking of genres, there was very little variation when users were asked to rank their favorite book formats.  Paperback ranked highest, followed closely by e-books, and more distantly by hardcover and audiobooks.  There was even less deviation when an overwhelming majority of users (71%) indicated that online shopping was their preferred method of shopping for books; and what little variation there was could be contributed to age (those under 35 were amiable to purchasing books from stores or websites; whereas those over 35 preferred brick and mortar outlets).

Word of mouth was the number one preferred method of receiving book recommendations for 72% of respondents. Amazon and GoodReads didn’t fare much better than third place for most individuals, and generic advertisements were barely considered by readers who cited reading more than 10 books in the past year.

 

The Inferential Analysis:

While using the Statistical Analysis System Enterprise Guide (SAS EG), I attempted to identify some trends worth noting that could predict the reading habits of individuals based on some identifying variables, but very few of these were statistically significant, and only one had an R-Square value of over 33%.

Interestingly, individuals without a Bachelor’s Degree read just under 21 books in the last year and approximately 400 books in their lifetime, whereas those with a Bachelor’s Degree (or higher) averaged 8 and 250 respectively.  This observation was surprising, as the intuitive assumption would be that those with degrees would have read more books in their lifetime than those without; the survey did not go into enough depth to determine why this trend existed, and the trend was not statistically sound enough to make any inferential assessment.  Other interesting, but not-fit-for-regression trends were: 

  • Women read, on average, 32% more within the last year and 26% more over the course of their lifetime.
  • Yearly reading remained relatively stable at 15 books a year across all age groups except 41-45 year olds who read fewer than 10 books a year.
  • Lifetime reading remained consistent at 201-300 books for all age brackets, except for 31-35 year olds and those over 50 who cited reading more than 500 books in their lifetime.  This spike in 31-35 year olds is likely a flaw in the sample selection (31-35 year olds frequent my class).
  • Individuals with children reported reading approximately 25% less over the course of the last year, and 18% less over the course of their lifetime.

The only indicator of how many books a person will read in a given year is their lifetime reading, with lifetime reading variability accounting for 44% of the variation in yearly reading.

Reading Survey 4.jpg

What this means is that bookstores should expect people with a history of reading to read more on a yearly basis than those without a history of reading.  Every other observation that showed a correlation with yearly reading habits was not concluded to have a causal relationship based on this non-random sample of the population.

 

Recommendations

Given the small (45) and non-random (three heterogeneous groups) sample that this survey was performed on, no actionable observations could be observed and no recommendations could be given.  This was an academic exercise meant to give me the experience of handling observational statistics from survey creation to report delivery.

However, if we were to assume that this sample was larger and random, and that the observations were actionable, we would come to the conclusion that the only variable that affects yearly reading would be historical reading, and even that appeared to be a little better than guessing.  There was some correlation between gender and family commitments that could be investigated more vigorously to determine if (through the averaging of multiple samples) a causal relationship exists.

This information would lead me to the assumption that marketers should not focus on improving reading habits through advertisements or reading campaigns; but rather focus on delivering recommendations based on genre with respect to gender and age groups.  Marketers should also focus on delivering more recommendations to purchase e-book formatted books to those under 35 to decrease publishing costs, and save advertising funds by not attempting to convert middle-aged readers to the digital format.

While the survey’s intent (to determine how to improve reading habits) was inconclusive, it did have a surprising kernel of truth that marketers should focus on:  word of mouth drives sales.  Amazon, Google, and GoodReads all have amazing algorithms that suggest book recommendations based on previous purchases, views, and wishlisted items.  These algorithms are increasingly complex, can be hauntingly accurate, and are a mathematical version of word of mouth; but they aren’t word of mouth, and they’re only considered slightly better than mass market advertising.

1 Hour Consultation: Data Analysis
45.00
Quantity:
Add To Cart

“But McCabe’s statistics show that only a piddling 10 percent of Amazon book choices are made because of its ‘bought this/also bought’ recommendation engine. Bestseller and top 100 lists influence 17 percent of book choices, with 12 percent down to promotions, deals, or low prices. Only 3 percent came through browsing categories. Planned search by author or topic, however, makes up a whopping 48 percent of all book choices” (Forbes, 2013).

Readers know what they want walking in (or logging on), and it’s based on what their friends are reading. The only way for companies to facilitate that is through social integration.  Combining the networks of Amazon and Facebook has been a hidden, and modestly implemented, Easter egg for the website for years; but if companies (including, but not limited to, Amazon) want to attempt to positively influence book sales, they need to look at driving (and controlling) word of mouth; and the easiest way to do that is through social integration.