How to Write the
“Haidt’s Data: Solid Truth
or Selective Storytelling?” Essay
This essay is not a debate about social media. It is an evaluation of Haidt as a writer and researcher — his data choices, his rhetorical moves, and where his argument holds up versus where it stretches. Here is how to approach it without getting that confused.
📝 Need expert help with this analytical essay? Our writers understand Haidt’s argument inside and out.
Get Expert Help →What the Prompt Is Actually Asking — and What It Isn’t
Read the instructions one more time. The prompt says: “This essay is NOT to argue for or against social media, screen time, or parenting styles.” That line eliminates about 80% of what students instinctively write when they read “Haidt.” You are not being asked whether phones are bad for teenagers. You are not writing a position paper on social media policy.
The actual target is Haidt the researcher and writer. The prompt asks you to be an evaluator, not a debater. Specifically: does his data genuinely support his conclusions, or does he use numbers strategically to steer the reader toward a conclusion the numbers don’t fully justify? When does reading him feel like reading science, and when does it feel like reading advocacy?
This is called a rhetorical analysis or argumentative evaluation — a type of essay that treats a text as an object of study. Your thesis will make a claim about Haidt’s method, not about his topic. There is an important difference. A thesis like “Social media harms teenagers” is a position essay. A thesis like “Haidt’s use of objective clinical data strengthens his core claim, but his selective aggregation of self-report studies and his framing of correlation as causation reveal a researcher who is also an advocate” is an analytical evaluation.
The Most Common Way Students Lose Marks on This Essay
Writing a standard “here’s why social media is bad” or “here’s why Haidt is right” essay. Your instructor has seen that a hundred times. The prompt is deliberately designed to test whether you can step back from the content of an argument and assess its construction. Stay focused on the quality of the evidence and reasoning, not whether you personally agree with the conclusion.
Haidt’s Core Argument and the Data He Uses to Build It
You cannot evaluate an argument you have only half read. Before writing a word, you need a clear map of what Haidt actually claims and what evidence he deploys at each stage. His argument is not simple — it stacks several layers.
Layer One: A real mental health crisis happened, starting around 2012
Haidt’s starting point is empirical: rates of adolescent anxiety, depression, self-harm, and suicide rose sharply and suddenly around 2012–2015 in the US, UK, Canada, Australia, and Nordic countries. He calls this “A Tidal Wave.” His evidence at this layer includes CDC emergency room data on self-harm, government suicide statistics, large annual surveys like Monitoring the Future, and international data from PISA and the Health Behavior in School-Age Children survey. This is arguably the strongest layer of his argument — the data he is drawing from is government-collected, large-scale, and objective. The ER visits for self-harm tripled for preteen girls (ages 10–14) between 2010 and 2020. These are not self-reports; they are hospital records.
Evaluating this layer: is the crisis real? Most researchers — including Haidt’s critics — accept that something happened. The debate is about what caused it.
Layer Two: The timing matches the smartphone/social media transition
Haidt argues that the inflection point — roughly 2012 to 2015 — corresponds to when smartphones became widespread among teenagers and social media platforms like Instagram (acquired by Facebook in 2012) became ubiquitous. He calls this the “Great Rewiring.” He presents a figure showing smartphone adoption rates alongside mental health decline curves, arguing the timing is not coincidental. He explicitly distinguishes this from the first wave of internet (1990s PCs), noting that the earlier wave did not produce a mental health decline — but the smartphone wave did.
Evaluating this layer: is the correlation real? Yes, broadly. Is that enough to establish causation? No — and Haidt knows this, which is why he then builds layer three.
Layer Three: The relationship is causal, not just correlational
This is where Haidt makes his most contested move. He argues that the relationship between social media use and mental health decline is not merely associational — it is causal. His evidence here shifts from large-scale epidemiological data to a mix of: longitudinal studies (the UK Millennium Cohort Study showing girls using five or more hours of social media daily were three times as likely to be depressed); dose-response relationships; international quasi-experiments (a 2022 study on fiber-optic rollout in Spain finding high-speed internet access increased adolescent mental health hospitalizations for girls); and mechanistic pathways (how social media affects sleep, social comparison, and attention).
Evaluating this layer: this is where the real analytical work happens. Each piece of evidence has strengths and limitations, and your essay should engage with that complexity.
Layer Four: The harm is gendered — girls are affected more than boys
Haidt dedicates significant space to arguing that social media harms girls via social comparison, appearance-based metrics, and the nature of platforms like Instagram and TikTok — while boys are harmed differently, primarily through video game and pornography use. He explains this through the psychological framework of agency versus communion, and uses UK Millennium Cohort Study data, TikTok algorithm research (by the Center for Countering Digital Hate), and psychogenic illness spread patterns.
Evaluating this layer: the gendered data is real but also where some of Haidt’s sub-group selection is most visible. When critics argue he is “selective,” they often point to how gender-disaggregated analysis lets him find large effects by focusing on the subset most affected.
Between 2010 and 2015, the social lives of American teens moved largely onto smartphones with continuous access to social media, online video games, and other internet-based activities. This Great Rewiring of Childhood, I argue, is the single largest reason for the tidal wave of adolescent mental illness that began in the early 2010s.
— Jonathan Haidt, The Anxious Generation (2024), p. 45Where Haidt’s Data Is Genuinely Strong
A good analytical essay does not just criticize. It gives credit where credit is due — then examines where the argument strains. Your reader (and your instructor) will trust your critical analysis more if you first demonstrate that you understand what Haidt gets right.
Objective Clinical Data on Self-Harm and Suicide
CDC emergency room data on nonfatal self-injury and National Center for Health Statistics suicide data are not self-reports. They are objective clinical counts. The tripling of self-harm ER visits for girls aged 10–14 between 2010 and 2020, and the 167% increase in the suicide rate for that group between 2010 and 2021, are hard to dismiss as artifacts of changing diagnostic culture. This is Haidt’s strongest empirical foundation.
International Replication Across Countries
Haidt does not rely solely on US data. He presents equivalent trend lines from the UK, Canada, Australia, and Nordic countries — all showing similar inflection points around 2012–2015. The synchronicity of a mental health shift across different healthcare systems, economic conditions, and cultures is a meaningful observation that common-cause arguments (recessions, political events) struggle to explain.
The Two-Wave Argument for Smartphones Specifically
Haidt’s observation that the first wave of internet (1990s–2000s, PCs) did not produce a mental health decline — while the second wave (2010s, smartphones and social media) did — is a meaningful analytical distinction. Millennial teens, who grew up with the early internet, were slightly happier than Gen X. Gen Z, who went through puberty with smartphones, were not. This controls for “internet in general” as the cause and points more specifically to smartphone-based social media.
Dose-Response Relationships and Mechanistic Evidence
In epidemiology, a dose-response relationship — more exposure, more harm — is considered stronger evidence for causality than simple correlation. Haidt presents UK Millennium Cohort Study data showing that girls using social media zero hours daily were least depressed, with depression rates rising in proportion to hours of use (three times higher at five or more hours). Sleep disruption research further identifies a physiological mechanism: 36 correlational studies confirm associations between social media use and sleep disruption, and experimental sleep restriction studies confirm the causal direction from phones to sleep loss.
Where Haidt’s Data Raises Legitimate Questions
This is the heart of the essay. “Selective storytelling” does not mean Haidt is dishonest — it means he is a researcher with a thesis, and like all researchers with theses, he makes choices about which data to foreground, how to aggregate it, and which counterarguments to address (and which to minimize). Your job is to see those choices clearly.
The Self-Report Problem
A large portion of Haidt’s mental health trend data — particularly the college mental health survey data and the annual surveys showing rising anxiety and depression — is based on self-reports. Self-report data is subject to multiple interpretations: prevalence may genuinely have increased, or young people may have become more willing to disclose symptoms, or the cultural normalization of mental health diagnosis may have changed how people categorize mild distress. Haidt addresses this objection in chapter 1 by pairing self-reports with objective clinical data (ER visits). But critics argue that the weight of his presentation still relies heavily on self-report trends for much of the international and longitudinal analysis.
Correlation Presented as Near-Causation
Haidt acknowledges the correlation-versus-causation problem directly — he devotes a section in chapter 6 to it. But his rhetorical strategy after that acknowledgment is to marshal enough correlational evidence that it begins to feel like causation by accumulation. Researchers Amy Orben and Andrew Przybylski, writing in Nature Human Behaviour (2019), applied a different statistical method to the same underlying datasets — specification curve analysis — and found that the association between social media use and adolescent well-being had effect sizes equivalent to wearing glasses or eating potatoes. Haidt disagrees with their approach. But your essay should flag that two serious researchers looking at the same data reached very different conclusions — and that methodological choices drive those differences.
Selective Sub-Group Aggregation
Haidt’s critics — most prominently Candice Odgers, a developmental psychologist at UC Irvine — have argued that he finds large effects partly by disaggregating data in ways that surface the subgroups where effects are largest (adolescent girls, heavy users, visually oriented platforms) while downplaying or averaging away subgroups where effects are small or absent (boys overall, moderate users, non-visual platforms). When you aggregate across all teens and all platform types, effect sizes shrink substantially. Haidt’s response is that averaging away the most affected group is analytically unsound — but this debate about the right level of aggregation is a genuine methodological dispute, not one that resolves cleanly in either direction.
Alternative Explanations Not Fully Ruled Out
Haidt works to eliminate competing explanations — the financial crisis, climate anxiety, school shootings — by arguing that none of them explain the timing (starting around 2012), the gender gap, or the international pattern. His rebuttals are strong in some cases but incomplete. He does not fully engage with research on rising academic pressure and college admissions competition, changes in sleep norms independent of phones, or the possibility that economic inequality and housing instability — rising throughout the 2010s — contributed to adolescent distress through pathways unrelated to smartphones.
Key External Source: Orben & Przybylski in Nature Human Behaviour
The most widely cited academic challenge to Haidt’s data interpretation is Amy Orben and Andrew Przybylski’s 2019 study, “The association between adolescent well-being and digital technology use” (Nature Human Behaviour, 2019). Using specification curve analysis on three large datasets, they found effect sizes for screen time on well-being were comparable to other trivial daily activities. Haidt and others have challenged their methodology in turn. For your essay, this debate is analytically productive — two credible researchers, same datasets, opposite conclusions. That is a data interpretation argument worth developing in a body paragraph.
Haidt’s Rhetorical Techniques — and Why They Matter for Your Essay
The prompt specifies that you should assess Haidt’s “use of data” and his “rhetorical persuasion.” Those are two different things, and your essay should treat them as such. Data is what the numbers say. Rhetoric is how those numbers are framed, sequenced, and emotionally delivered.
The Mars Analogy (Opening Chapter)
The book opens with a thought experiment: would you send your child to grow up on Mars, where conditions are untested and developmental risks are unknown? The answer is obviously no — which is exactly the emotional frame Haidt wants before presenting any data. Analyze this move: it pre-loads the reader with a “protective parent” frame that makes every subsequent data point feel morally urgent rather than scientifically uncertain. The analogy bypasses skepticism before it can form.
Anecdotal Framing Around Hard Data
Each chapter opens with a named child, a concerned parent, a specific story of gaming addiction or social media harm. These are followed by aggregate statistics. The rhetorical sequence matters: the anecdote makes the data feel personal and inevitable, and the data makes the anecdote feel representative. This is a standard journalistic technique — but it works by making every data point feel like it is describing “your child,” regardless of what the base rates actually suggest.
Visual Graphs and the Shaded “Great Rewiring” Zone
Haidt explicitly describes adding a shaded region to every graph to make it “easy for you to judge whether or not something changed between 2010 and 2015.” This is a visual framing choice. The shaded zone directs the reader’s attention to the post-2012 upturn, visually minimizing any pre-2012 trends or alternative inflection points. The choice of where to draw “baseline” years, and which years to shade, is not neutral — it is a design decision that supports his interpretation.
Preemptive Rebuttals — Selective Coverage
Haidt addresses some critics directly (the psychiatrist skeptical of self-reports, those who blame the financial crisis) and dismisses them with evidence. But he does not engage substantively with Orben and Przybylski’s specification curve analysis, or with Candice Odgers’ critique of his aggregation methods, in the main text. These appear in endnotes. The rhetorical effect is a book that appears to have answered its critics — when in fact it has answered the critics it chose to answer.
Analytical vs. Evaluative Language — Know the Difference
In your essay, be precise about what you are doing. Analytical language describes a technique: “Haidt opens the book with a thought experiment that positions smartphone culture as analogous to Mars colonization.” Evaluative language assesses its effect: “This framing pre-empts skepticism by activating parental protective instincts before any data is presented, making the subsequent statistics feel more alarming than they might without that emotional priming.” Use both — describe the technique, then evaluate its effect on how a reader receives the data.
How to Build a Thesis for This Specific Prompt
Your thesis is the most important sentence in the essay. It needs to make a specific claim about the quality of Haidt’s argument — not just observe that his argument has both strengths and weaknesses (that is not a thesis, it is a truism). The thesis should stake out a position on the balance between solid evidence and rhetorical construction, and it should be specific enough to tell your reader what three to four analytical claims you will make in the body paragraphs.
| Thesis Direction | Example Thesis Statement | Body Paragraphs It Sets Up |
|---|---|---|
| Balanced — Strong objective data, weaker causal claims | “Haidt builds his case on a foundation of objective clinical data that is difficult to dismiss, but the weight he places on correlational studies — and the rhetorical techniques he uses to present them — transforms a compelling public health observation into an argument that outpaces the evidence.” | 1. Strength of CDC/suicide data. 2. Correlation-as-causation problem. 3. Rhetorical framing (Mars, anecdotes, shading). 4. Sub-group selection issue. |
| More critical — Rhetoric dominates | “While Haidt grounds his argument in genuinely alarming clinical data, his rhetorical choices — selective aggregation, emotional analogies, and preemptive rebuttals that avoid his strongest critics — reveal a researcher writing advocacy more than analysis.” | 1. What the objective data does support. 2. The Orben/Przybylski debate and aggregation methods. 3. Rhetorical techniques (anecdotes, Mars framing). 4. Who Haidt doesn’t respond to. |
| More sympathetic — Data is solid, gaps are expected | “Haidt’s data, taken as a whole, presents the strongest available evidence for a phone-based cause of the adolescent mental health crisis, and while his rhetorical presentation sometimes blurs the line between correlation and causation, the breadth of convergent evidence he assembles makes the selective storytelling critique difficult to sustain.” | 1. Objective data strength. 2. Two-wave argument as controls for “internet in general.” 3. Where critics’ alternative explanations fall short. 4. Rhetoric as audience-appropriate rather than distorting. |
Your Thesis Does Not Need to Be “Balanced” — It Needs to Be Arguable
Students often feel they need to say “Haidt is partly right and partly wrong.” That is not a requirement. A strong analytical essay can take a clear position — that Haidt is primarily a solid researcher, or primarily an advocate — as long as it engages seriously with the evidence for the other side. A specific, arguable claim is stronger than a carefully balanced hedge. Pick a direction and defend it with evidence from the text and your secondary sources.
Essay Structure, Paragraph by Paragraph
3–4 pages double-spaced in MLA format is roughly 850–1,100 words of body content. That is four to five substantial body paragraphs — enough to develop three or four analytical claims with textual evidence and analysis. Do not try to cover everything. Cover fewer things better.
Every Body Paragraph Needs All Three: Evidence, Analysis, Connection to Thesis
A paragraph that only summarizes Haidt’s data will lose marks. A paragraph that only criticizes without textual evidence will also lose marks. The structure that works: Topic sentence → specific textual evidence from Haidt (with page citation in MLA format) → a secondary source that confirms, complicates, or challenges that evidence → your own analytical comment connecting the paragraph to your thesis. That four-step rhythm is what analytical body paragraphs look like at college level.
Finding 5 Sources Published Within the Past 5 Years
The assignment requires a minimum of five sources published within the past five years (2021–2026). That means no older sources, even highly relevant ones. Haidt’s book (2024) counts as one — you need four more. Here is where to look and what qualifies.
Haidt, Jonathan. The Anxious Generation. Penguin Press, 2024. Primary
Your primary text. This counts as one of your five sources. Every claim you make about Haidt’s argument must be supported by a specific page citation from this text. Do not paraphrase from memory — quote or closely paraphrase with page numbers.
Peer-reviewed responses to Haidt in psychology/psychiatry journals (2022–2026) Academic
Search Google Scholar or your library database for: “Haidt Anxious Generation critique,” “adolescent social media mental health causation 2022,” “screen time well-being effect size.” Journals to target: JAMA Pediatrics, Psychological Science, Nature Human Behaviour, Journal of Adolescent Health. Odgers, Twenge, Orben, or Przybylski are all names worth searching.
Government/institutional mental health data reports (2021–2026) Data Source
CDC reports on adolescent mental health, the Surgeon General’s Advisory on Social Media and Youth Mental Health (2023), or NHS/Public Health England mental health statistics. These let you independently verify or contextualize the data Haidt presents, which demonstrates you are evaluating the data rather than just accepting his framing of it.
Rhetorical analysis scholarship on science communication and advocacy (2021–2026) Academic
For the rhetorical analysis section, a secondary source on how scientists communicate risk to the public — or on the specific rhetorical techniques in popular science books — strengthens your analytical framework. Search: “science communication rhetorical strategies,” “popular science advocacy rhetoric,” “researcher as public intellectual.” Journals like Rhetoric of Science or book chapters on science writing are relevant.
Review articles on adolescent digital technology and well-being (2022–2026) Academic
A systematic review or meta-analysis gives you empirical breadth. Search PubMed or PsycINFO for: “systematic review social media adolescent mental health 2022 2023 2024.” These papers often synthesize the same studies Haidt cites and come to their own conclusions about what the evidence supports — which is analytically useful for assessing whether Haidt’s interpretation is consensus or contested.
The Accuracy Requirement — Do Not Cite What You Have Not Read
The assignment specifically flags: “Make sure the references you cite are relevant to the paragraph! Accuracy!!!” This means your instructor will check that sources are (a) real, (b) actually say what you claim they say, and (c) logically connected to the paragraph they appear in. Do not cite Orben and Przybylski in a paragraph about rhetorical technique. Do not cite a CDC report in a paragraph about Haidt’s narrative framing. Match each source to the specific claim it supports. Read at least the abstract and key sections of every source you cite — do not cite based on title alone.
MLA 9 Formatting — The Practical Checklist
MLA 9 updated several conventions from MLA 8. Make sure you are using the current version, not the older one. Here are the requirements that trip students up most often.
| Formatting Element | MLA 9 Requirement | Common Mistake |
|---|---|---|
| Header (top of page 1) | Your name, instructor name, course name, date — each on its own line, left-aligned, double-spaced | Putting this in a header box, or using a title page instead |
| Title | Centered, plain text — no bold, no underline, no quotation marks. Capitalize major words. | Bolding the title or putting it in quotes |
| Margins and spacing | 1-inch margins on all sides, double-spaced throughout (including Works Cited), 12pt Times New Roman or similar | Forgetting to double-space the Works Cited page, or using extra spacing between paragraphs |
| In-text citations — book | (Haidt 34) — author last name and page number, no comma | Using (Haidt, 2024, p. 34) which is APA, not MLA |
| In-text citations — article with no page numbers | (Orben and Przybylski) — author names only when no page numbers | Making up page numbers, or using paragraph numbers when they are not provided |
| Works Cited — book entry | Haidt, Jonathan. The Anxious Generation: How the Great Rewiring of Childhood Is Causing an Epidemic of Mental Illness. Penguin Press, 2024. | Not italicizing the title, including “New York” as publisher location (not required in MLA 9) |
| Works Cited — journal article | Last, First, and First Last. “Article Title.” Journal Name, vol. X, no. X, year, pp. XX–XX. DOI or URL. | Not including volume/issue numbers, omitting DOI for online articles |
| Works Cited — hanging indent | First line flush left, subsequent lines indented 0.5 inches | Indenting the first line (that is the opposite of correct) |
| Italics vs. quotes | Book/journal titles in italics; article/chapter titles in quotation marks | Quoting the journal title instead of italicizing it, or italicizing the article title |
Use Purdue OWL as Your MLA 9 Reference — It Is Free and Authoritative
The Purdue Online Writing Lab (OWL) MLA formatting guide is the most reliable free resource for MLA 9 citation examples. It is regularly updated and covers edge cases (government documents, online-only articles, books with editors, etc.). Do not trust citation generators alone — they make errors in MLA 9, especially with DOIs and online source formatting. Use OWL to verify any citation format you are unsure about.
FAQs: Writing the Haidt Data Analysis Essay
The Short Version Before You Start Writing
Forget the social media debate. Your job is to read Haidt as a researcher who is also an advocate, and to assess how well those two roles stay in balance. When does the evidence actually carry his conclusion? When does the framing do work that the data alone cannot?
Start by mapping his argument in four layers — the crisis, the timing, the causation claim, the gender gap — and identify the type of evidence he uses at each layer. Then figure out which layers are solid and which are contested. Then look at the rhetorical techniques and ask what each one does to a reader’s interpretation of the data.
Write a thesis that takes a position on that balance. Use body paragraphs that combine specific textual evidence (with MLA page citations) and secondary sources (verify the accuracy of every claim you attribute to a source). Keep the Works Cited page clean, double-spaced, and in proper MLA 9 format.
For professional help with essay writing, analytical writing support, proofreading, or essay tutoring, the team at Smart Academic Writing is available for any stage of the process.