An Economic Critique of Self-Publishing

Its positive and negative consequences

As most of you know, I am a self-published (also called “independent”) author and have spoken very positively about the benefits of decentralizing the publishing industry: lowering the barrier of entry for marginalized groups, and essentially giving every author a chance to share their stories with the world. However, the landscape of the publishing industry post self-publication technologies has its caveats, which are often overlooked. Another thing that has been swept under the rug is the use of AI in writing. Most authors I have talked to are vehemently opposed to AI, and while it does pose serious issues pertaining to IP and copyright, there are also many potential applications to improve the industry for all players. The following brief analysis is a university assignment that takes on this topic from an economics perspective — interesting enough, I believe, to be shared publicly here, as many of you readers are also authors. As always, discussion, questions, and comments are welcomed.

Introduction

Traditionally, publishers served as gatekeepers between authors and readers; because publishers were selective in choosing books to publish, they acted as quality control and created a bottleneck to ensure that only books of a certain standard made it on the market.

In the 2000s, self-publishing technologies like Amazon KDP, IngramSpark, and Lulu emerged, lowering barriers of entry for authors, and allowing them to bypass traditional publishers. Since then, the supply of books has increased, and the overall quality has gone down.

The current state of the fiction book market is a market failure because the lack of quality control for books leads to increased information asymmetry between authors and readers. However, using AI-curated recommendations, it is possible to keep the benefits of democratizing creative industries — allowing underrepresented and marginalized authors to enter — while solving the problem of information asymmetry by leveraging AI’s processing power to replace the role of publishers as quality control.

Analysis

The Third Law of Library Science states that “every book has a reader” (Ranganathan, 1948), suggesting that due to personal preferences, there is an audience for every type of book. Ideally, this would result in readers finding books they enjoy, ensuring that no books remain unsold, but there exists no effective curating system to achieve this. Thus, readers struggle to find the right books, a challenge that has been exacerbated by the recent surge in the number of books available on the market — almost 320 million books are thrown away every year (Jackus, 2023).

Information asymmetry contributes to this market’s allocative inefficiency. Generally, the higher a book’s quality — where quality is objectively measured by factors like grammatical accuracy, structural consistency, and logical plot progression — the more likely readers will enjoy it. Like a peach/lemon problem, authors know more about the quality of their books than the readers; if they follow best practices — editing multiple drafts, hiring an editor and beta readers — they can be confident that their books are of higher quality than authors who do not. But unlike the lemon/peach problem for used cars, books are an experience/hedonic good, meaning that readers purchase the good with time in addition to money. Due to this, the typical “death spiral” of prices does not happen; as data shows, the majority of books have low prices but still struggle to attract readers (Fiol, 2021).

The traditional solution was to use publishers as quality-control moderators. With only publisher-screened books on the market, readers still have a risk of picking up a book they personally dislike, but they can be assured of the book’s general quality. This phenomenon is known as the Mahogany Desk Syndrome: a stamp of approval from a publisher signals to readers that the book has been deemed worth selling, so it might be worth reading (Lilly, 2017, as cited in Fiol, 2021).

However, with self-publishing comes the ability to bypass publishers and thus quality control. There are nearly 160 million books on the market as of 2023 with quality unknown to readers. (Giordano, 2023). Additionally, the large number of books makes the reader’s choice harder due to limited cognitive power; readers are increasingly doubtful that they will make the right one. They perceive a high opportunity cost attached to any choice, leading to FOMO and a higher rate of regret post-decision (Pilat & Krastev, 2024). People may fall into the status quo bias; to avoid the risk of regret, readers stick to books/authors they are familiar with. Without the benefit of the doubt, new and rising authors suffer, and well-known authors stay in the mainstream (Samuelson & Zeckhauser, 1988).

Additionally, because people often consider choice in complex situations with the most available information, they are susceptible to framing — in this case, “judging a book by its cover” (Thomas & Millar, 2011). As authors become aware of this tendency, profit-maximizing strategies can lead to competition in non-quality aspects, such as presenting a book with a nice cover. This makes it more difficult for readers as framing also proves to be an unreliable indicator of quality.

The cumulative effect of this results in readers being discouraged from buying and reading books, even though there are books out there of excellent quality and suit their tastes perfectly — they simply just have no way of telling a good quality book from a bad one, or a peach from a lemon, resulting in market inefficiency (Hviid et al., 2019). This also diminishes consumer surplus. When readers cannot confidently assess a book’s quality, they may be less willing to pay for it, leading to lower overall satisfaction and reduced consumer surplus.

This market inefficiency affects individuals (authors and readers) as well as society: because books are an elastic good, movies and short-film apps like TikTok are convenient substitutes, tempting individuals and groups to stop buying books altogether. An annual survey conducted by the National Endowment for the Arts found that the percentage of US adults reading fiction fell from 45.2% in 2012 to 37.6% in 2022 (PublishersWeekly, 2023). Problematically, US youth are also entering college overwhelmed by reading and wholly unprepared to read books, as noted by Professor Nicholas Dames of Columbia University along with around 33 other professors (Horowitch, 2024).

Information asymmetry also negatively impacts publishers and authors through unsold books. The average return rate for unsold printed books is 30% (Warner, 2016), representing a loss of potential producer surplus. Publishers and authors incur costs without achieving sales, which decreases the overall welfare of producers in the industry.

On a higher level, the US book industry was worth over 28.1 billion USD in 2023 (Watson, 2024); continuous market failure with readers taking their money elsewhere may potentially cause the industry to crash, which negatively impacts other related industries such as distribution and retail. Ceteris paribus, GDP will fall as book consumption/sales decline.

This market inefficiency ultimately harms overall welfare, as both readers and authors experience losses. By addressing information asymmetry and aligning book availability with reader preferences, there is potential to increase consumer and producer surplus, thereby improving the welfare of the entire industry.

Proposal

As analyzed, human quality control cannot keep up with the pace of new books — but AI, with its fast processing speed, can (Dodda & Navin, 2023).

Existing tools such as Readow.AI can recommend books, but run on biased algorithms, leading to biased results. Their main metric under consideration is book ratings/reviews because AI is good at comparing and analyzing objective criteria; however, this is not a holistic measurement of a book’s quality and can be easily manipulated (and bought!). Book reviews are highly subjective, often reflecting how a book resonated emotionally with an individual reader. Moreover, focusing solely on ratings and reviews can overlook books with few or no reads, which doesn’t necessarily mean they are poor in quality — they might simply be recently published. Secondly, AI-powered algorithms often contain availability bias because AI models train on the most popular or available data, leaving new and emerging books underrepresented (Lee, Resnick, & Barton, 2023). Because our preferences are subjective, popular books are not always good recommendations; many great unconventional books are overshadowed. Thus, we see existing solutions are inadequate to solve this market failure.

We propose a private-party solution: a digital platform that uses AI to holistically analyze a book and curate personal recommendations for readers, mitigating algorithmic biases that existing solutions have.

This algorithm will not only rely on reviews; it is possible for AI to do a quick, holistic analysis of a book, examining its grammar, tropes, reading level, and character development, which form a better objective indicator of quality. This gives all books — regardless of popularity — an equal chance to be discovered by readers. Marlowe.ai (which I have used before on my previous books) is one such analysis tool, demonstrating that this solution is practical and feasible. On this proposed platform, writers will upload their manuscripts for AI analysis, and readers can input a few titles they enjoy. The AI then identifies common themes and recommends similar books to readers.

To mitigate availability bias, the algorithm will personalize data for each user, ensuring that future recommendations rely on their past experiences and reviews, rather than general market trends. By incentivizing readers to leave reviews and feedback on how they felt about a book, the algorithm can avoid relying on generalized or popular data to drive personal recommendations. As the recommendations gets more personal and accurate by mitigation of algorithmic bias, the amount of options needed to present to the reader becomes less which can decrease the cognitive load.

This solution benefits both authors and readers by replacing the role of a publisher as a means of quality control. This approach not only bridges information asymmetry but also highlights the benefits of decentralizing the book industry — giving each book an equal chance to be discovered and matching every book to its right reader.

Sources

Dodda, S., Kamuni, N., Sai, V., Vuppalapati, M., & Vemasani, P. (2023). AI-driven Personalized Recommendations: Algorithms and Evaluation. Tuijin Jishu/Journal of Propulsion Technology, 44(6), 1001–4055. https://www.researchgate.net/publication/379664167_AI-driven_Personalized_Recommendations_Algorithms_and_Evaluation

Fiol, M. G. (2021). An Analysis of the Self-publishing Book Industry. Universitat de les Illes Balears. https://dspace.uib.es/xmlui/bitstream/handle/11201/158989/Gallent_Fiol_Margalida.pdf?sequence=1

Giordano, V. (2024, October 18). How Many Books Are In The World? (2024) - ISBNDB Blog. ISBNDB Blog. https://isbndb.com/blog/how-many-books-are-in-the-world/#:~:text=The%20organization%20estimates%20that%202.2

Horowitch, R. (2024, October 2). The elite college students who can’t read books. The Atlantic. https://www.theatlantic.com/magazine/archive/2024/11/the-elite-college-students-who-cant-read-books/679945/

Hviid, M., Izquierdo-Sanchez, S., & Jacques, S. (2019). From Publishers to Self-Publishing: Disruptive Effects in the Book Industry. International Journal of the Economics of Business, 26(3), 355–381. https://doi.org/10.1080/13571516.2019.1611198

Jackus, D. (2023, January 3). Increasing Recycling - Dream Books Co. and Nordsense. Greener and Smarter Waste - Nordsense. https://nordsense.com/increasing-recycling-and-keeping-books-out-of-landfills-with-dream-books-co/

NEA Survey Finds Decline in Adult Reading. (n.d.). PublishersWeekly.com. https://www.publishersweekly.com/pw/newsbrief/index.html?record=4377

Pilat, D., & Krastev, S. (2022). Choice Overload Bias. The Decision Lab. https://thedecisionlab.com/biases/choice-overload-bias

Ranganathan, S. R. (1948). The Five Laws of Library Science. http://arizona.openrepository.com/arizona/bitstream/10150/105454/3/PrefM.pdf

Samuelson, W., & Zeckhauser, R. (1988). Status quo bias in decision making. Journal of Risk and Uncertainty, 1(1), 7–59. https://doi.org/10.1007/bf00055564

Thomas, A. K., & Millar, P. R. (2011). Reducing the Framing Effect in Older and Younger Adults by Encouraging Analytic Processing. The Journals of Gerontology Series B, 67B(2), 139–149. https://doi.org/10.1093/geronb/gbr076

Warner, B. (2016, October 31). Returns 101: What New Authors Need to Know. PublishersWeekly.com. https://www.publishersweekly.com/pw/by-topic/authors/pw-select/article/71886-returns-101-what-new-authors-need-to-know.html

Watson, A. (2024, May 16). U.S. book market - statistics & facts. www.statista.com; Statista. https://www.statista.com/topics/1177/book-market/