The iPhone Was Not Yet the Ecosystem: Mechanism Audit, Calibration, and Falsification Tests for Myers and Hooper (2026)

[redacted for review]

June 2026

JEL codes: J13, O33, L86, D91

Keywords: fertility decline; iPhone; smartphones; feature phones; texting; social media; contraception; mechanism validity

Abstract

Myers and Hooper (2026) use AT&T's 2007-2011 U.S. iPhone exclusivity as a natural experiment and estimate large reductions in birth rates in counties with high AT&T 3G coverage. This comment accepts the usefulness of the reduced-form design but argues that the mechanism interpretation is underidentified. The paper's interaction-displacement channel was not smartphone-specific: ordinary cell phones, SMS, multimedia feature phones, PC social networking, mobile browsers, BlackBerry, Android, and Wi-Fi Apple devices already supported mediated teen interaction before a mature iPhone app ecosystem existed. This revised version adds three formal robustness modules: a first-stage calibration showing the treatment-on-treated magnitudes implied by plausible iPhone-adoption first stages; a dated ecosystem timeline that imposes a conception-to-birth lag on candidate mechanisms; and a mechanism-by-data matrix that specifies the public and restricted evidence needed to distinguish iPhone apps from SMS, PC social networking, contraception, pregnancy resolution, STI exposure, and local recession/urban trends. Existing public evidence favors a narrower interpretation: AT&T 3G coverage likely proxies a broader early-digital, urban, socioeconomic, and reproductive-control bundle. The iPhone may have been part of that bundle, but the public mechanism evidence does not yet support attributing one-third to one-half of the national fertility decline to an iPhone-specific behavioral mechanism.

1. Introduction

Myers and Hooper (2026) ask whether the iPhone functioned as a fertility-reducing technology. The empirical design exploits a historically useful fact: from June 2007 until the Verizon iPhone launch in February 2011, U.S. iPhone service was tied to AT&T. Counties with high AT&T 3G coverage therefore had better access to the iPhone during the first diffusion wave. The paper estimates birth-rate reductions of 4.5-8.0% among women aged 15-19 and 3.2-6.6% among women aged 20-24, and attributes 33-52% of the 2007-2011 decline in the general fertility rate to iPhone diffusion (Myers and Hooper 2026).

The design is creative. The weakness is not the institutional fact of the AT&T monopoly. The weakness is the interpretation of the treatment. The paper's empirical treatment is AT&T 3G coverage. Its title and decomposition are about the iPhone. Its mechanism story is broader still: reduced in-person social interaction, more pornography, lower sexual frequency, and possibly changes in reproductive information. Those are not the same object. A technology can be plausibly associated with all of these behaviors without the empirical design identifying that particular mechanism.

This comment extends the earlier calibration critique by auditing the mechanisms one by one. The aim is not to prove that phones cannot affect fertility. They can. The aim is to ask which phone-mediated mechanisms were actually available at scale from 2007 to 2011, which were specific to the iPhone, and which should leave observable traces outside county birth rates. The core finding is that the iPhone-specific mechanism burden is too heavy. The social-displacement mechanism existed, but it existed primarily through ordinary phones, SMS, PC social networking, BlackBerry, Android, mobile browsers, and Wi-Fi devices. The dating-app and modern image-feed channels mostly did not exist. The contraception channel has strong independent support but is not iPhone-specific. Pornography remains weakly measured as an iPhone mechanism.

The comment is structured as a mechanism audit. Section 2 sets up the causal chain and the falsification logic. Sections 3 through 10 examine social interaction, SMS and feature phones, mobile social networking, the app stack, dating apps, pornography, contraception, and outcome triangulation. Section 11 sets out specific tests that can be run using the original county panel and additional data. Section 12 concludes.

2. Mechanism identification requires more than a reduced-form treatment

The reduced-form estimate can be written as an effect of AT&T 3G coverage on birth rates. The mechanism interpretation requires a longer chain:

1. AT&T 3G coverage caused materially higher iPhone adoption among the relevant women, partners, and peer groups.

2. The marginal iPhone exposure changed a behavior that is fertility-relevant.

3. The behavior changed births rather than only reallocating births across ages or places.

4. The effect was large enough relative to the first stage to explain a substantial share of national fertility decline.

5. Competing channels correlated with AT&T coverage - feature phones, SMS, PC social networks, Android/BlackBerry, iPods, broadband, recession exposure, contraception, and media campaigns - do not explain the same patterns.

Each link implies different evidence. A social-displacement mechanism should predict reduced in-person interaction where iPhone exposure rises, after the relevant device and app timing. A contraception-information mechanism should predict reduced pregnancies and unintended births, possibly without a large decline in sexual activity. A pornography-substitution mechanism should predict lower partnered sex and perhaps lower STIs, but it requires device-specific evidence because online pornography was already available on PCs. A dating-app mechanism should not affect births until the apps exist and diffuse. A general digital-modernity channel should appear on non-iPhone platforms and Wi-Fi devices as well.

The target paper's mechanism evidence is weaker than its reduced-form evidence. The authors use national survey series and Google Trends to document broad declines in time with friends, some changes in sexual behavior, and rising pornography interest. Those series are useful background facts, but they are not linked to AT&T coverage at the county-year level. The paper is therefore vulnerable to mechanism slippage: a county-level AT&T treatment is used to support a national behavioral narrative that may have been driven by different technologies or non-technological changes.

Table 1. Mechanism audit summary

3. Social displacement: plausible, but not iPhone-specific

The social-displacement channel is the most intuitive mechanism in Myers and Hooper (2026): the iPhone may have absorbed attention, substituted mediated interaction for in-person time, and reduced opportunities for partnered sex. The paper's own national time-use evidence is consistent with a broad decline in time spent with friends. But a valid mechanism test must ask whether the marginal technology causing that displacement was the iPhone or a broader communication environment.

The existing evidence makes this distinction central. Pew reported that in 2009, 75% of U.S. teenagers aged 12-17 owned cell phones, 72% of all teens texted, and 54% texted daily (Lenhart 2010). Daily texting had risen rapidly from 38% in February 2008 to 54% in September 2009. Among teen texters, one-third sent more than 100 texts per day. This is already a high-frequency mediated-interaction environment before the mature app ecosystem. If texting reduces in-person contact, then ordinary cell phones are a first-order treatment.

This is not a minor control variable. For teenagers, SMS was closer to universal than the iPhone. The paper's estimated effects are especially large at ages 15-19, precisely the group for which early iPhone ownership was likely most limited relative to ordinary cell ownership. Pew later reported that teen smartphone ownership was only 23% in the 2011 teen survey, while cell-phone ownership was 77% (Lenhart 2012). Thus, the social-displacement mechanism was available to far more teenagers through non-smartphones than through iPhones.

The mechanism is also behaviorally direct. Texting is not merely a generic media channel. It coordinates meeting, flirting, conflict, avoidance, parental monitoring, and romantic negotiation. Pew's online-dating report later found that among adults with recent dating experience, 37% had asked someone out by text message on a cell phone and 17% had broken up with someone by text message, email, or online message (Smith and Duggan 2013). Those figures are later than the AT&T monopoly window and include adults, but they show that SMS itself is a relationship technology. The relevant treatment could be mediated communication, not smartphone ownership.

The falsification implication is straightforward. A paper claiming an iPhone-specific social-displacement mechanism should show that the treatment effect survives controls for local feature-phone/SMS intensity and that SMS-heavy non-iPhone places do not exhibit comparable fertility declines. If that test fails, the mechanism becomes a general mobile-communication story. That would still be interesting, but it would not support the decomposition attributing one-third to one-half of the national fertility decline to iPhone diffusion.

4. PC social networking and dumb-phone interaction were already mature enough to affect teens

A second reason the mechanism is not iPhone-specific is that social networking was already widespread on PCs and mobile browsers. Pew reported that 73% of online American teens used social-networking websites as of September 2009, compared with 55% in November 2006 and 65% in February 2008 (Pew Research Center 2010a). Online young adults were also heavy users. Thus the social graph existed before the modern smartphone-app environment.

This timing matters. If digital social life reduced in-person interaction, that process did not begin with the native App Store. It plausibly began with broadband, MySpace, Facebook, instant messaging, SMS, and camera phones. The iPhone improved mobility and interface quality, but it did not introduce the basic social-network mechanism. In an econometric design, the question is whether AT&T 3G coverage generated an incremental social shock large enough to produce the estimated fertility effects after accounting for the pre-existing PC/SMS social layer.

The iPod touch is an underused placebo for this claim. Apple introduced the iPod touch in September 2007 with Wi-Fi, Safari, YouTube, and the iPhone-style interface, but without an AT&T cellular contract (Apple 2007b). Apple reported App Store downloads jointly for iPhone and iPod touch users in 2008, and iPhone OS 3.0 served more than 30 million iPhone and iPod touch users worldwide by March 2009 (Apple 2008b; Apple 2009a). If apps, mobile web, YouTube, games, and portable screens are the mechanism, Wi-Fi Apple devices should matter even when AT&T cellular coverage does not. If only AT&T iPhone coverage matters, the paper needs to explain why cellular iPhone ownership, rather than the broader Apple mobile-media ecosystem, is the operative treatment.

This is not a semantic objection. The iPod touch and Wi-Fi devices break the link between AT&T coverage and Apple mobile-app exposure. The iPhone itself bundles three components: a phone, a mobile internet device, and an app/media platform. AT&T exclusivity identifies the cellular phone component most clearly. The paper's mechanism, however, often refers to the app/media/social components. A credible mechanism test should use iPod touch, home broadband, Wi-Fi density, and PC social-network penetration as placebo or competing treatments.

5. App ecosystem timing: the treatment window precedes much of the modern mechanism

The ecosystem-maturity critique becomes sharper when the birth lag is imposed. A birth observed in year t generally reflects conception roughly nine months earlier. Therefore, a mechanism must exist and diffuse before the relevant conception window, not merely before the birth year. A 2011 birth usually reflects 2010 or early-2011 behavior. A late-2011 app or service cannot explain most 2011 births.

The original iPhone went on sale in June 2007, but it did not have a native third-party App Store (Apple 2007a). The App Store launched in July 2008 with 500 apps (Apple 2018). Apple reported more than 10 million downloads in the first weekend, with apps ranging from games to location-based social networking, medical applications, and productivity tools, and later reported 100 million downloads by September 2008 (Apple 2008b; Apple 2008c). Push notifications appeared with iPhone OS 3.0 in 2009, when the App Store had more than 25,000 apps (Apple 2009a). Apple announced 100,000 apps in November 2009 (Apple 2009b). iMessage, Notification Center, and deep Twitter integration came with iOS 5 in 2011, at the end of the AT&T exclusivity window (Apple 2011a; Apple 2011c).

These dates imply a set of timing restrictions. Effects in 2007 and much of 2008 cannot be attributed to native apps. Effects in 2009 can be attributed to early apps, games, maps, mobile web, and Facebook mobile, but not to iMessage, Tinder, Snapchat Stories, Instagram Direct, or algorithmic feeds. Effects in 2010-2011 can include early Instagram only at the end of the period; they still cannot include Tinder or most modern social-app features.

Table 2. Ecosystem timing and earliest plausible birth effects

6. Mobile social networking: plausible, but cross-platform and not exclusively smartphone

Mobile social networking is a more plausible 2009-2011 mechanism than dating apps. It existed and it scaled. Comscore reported that 72.2 million Americans accessed social networking sites or blogs on a mobile device in August 2011, up 37% over the previous year; nearly 40 million did so almost daily (Comscore 2011c). But this evidence does not identify an iPhone channel. Comscore's mobile-social statistic covers mobile devices broadly. It includes browser and app access, smartphone and non-smartphone access, and multiple carriers.

The platform evidence also cuts against a uniquely iPhone interpretation. In December 2010, Comscore reported 63.2 million U.S. smartphone subscribers, with RIM at 31.6% of the smartphone platform market, Android at 28.7%, and Apple at 25.0% (Comscore 2011a). By June 2011, Android had 40.1% and Apple 26.6% (Comscore 2011b). Nielsen's 2010 app report similarly showed that popular smartphone apps were games, Facebook, Google Maps, and Weather Channel, with Facebook among the most popular apps across iPhone, Android, and BlackBerry users (Nielsen 2010). The relevant app categories were not Apple-only.

If mobile social networking is the causal channel, the test should not be AT&T iPhone coverage alone. It should compare iPhone, Android, BlackBerry, feature-phone mobile web, and PC social networking. The paper's Verizon/Sprint placebo tests are useful but not decisive because non-AT&T carriers had different device mixes and customer bases before 2011. The stronger test is a post-2011 handoff: after the Verizon iPhone launch, Verizon-strong counties should display a delayed fertility response if iPhone access is the causal technology (Apple 2011b). A further test is whether Android-heavy counties show similar effects by 2010-2011. If they do, the paper should be retitled around mobile internet rather than the iPhone. If they do not, the authors need to explain why an app ecosystem with cross-platform Facebook, maps, browsers, and games would produce an iPhone-only fertility effect.

7. Dating apps: the cleanest timing falsification

The mobile dating-app mechanism is the clearest case where ecosystem maturity falsifies the strongest interpretation. Tinder did not exist during the AT&T-exclusive window. Tinder's press materials describe a 2012 launch, and contemporaneous reporting in January 2013 described Tinder as a newly launched iOS dating app gaining traction on campuses (Tinder n.d.; Empson 2013). Hinge's own history states that Justin McLeod created Hinge in 2012 and launched a mobile version in 2013 (Hinge n.d.).

Even after launch, mobile dating-app penetration was initially modest. Pew's 2013 online-dating report found that 3% of all adults and 6% of smartphone owners had used a mobile dating app, with 5% of ages 18-24 and 11% of ages 25-34 reporting such use (Smith and Duggan 2013). The report's app-specific counts, based on a small dating-app-user subsample, included only one Tinder mention. These data are not precise measures of Tinder adoption, but they strongly indicate that swipe dating could not have driven births in 2007-2011.

The sign of the dating-app mechanism is also theoretically ambiguous. A dating app can reduce search costs and increase partnered sex, especially among adults, which could raise pregnancy risk absent contraception. It can also replace in-person courtship with screening, delay partnership, alter match quality, or increase contraception-aware casual sex. Because the direction is ambiguous and timing is late, dating apps should be treated as an out-of-window mechanism for this paper. If the authors wish to claim dating-app-mediated fertility effects, the appropriate outcome window is births from mid-2013 onward, not the AT&T monopoly period.

8. Image-based and ephemeral social media: early Instagram is not modern Instagram

The contemporary intuition that smartphones changed youth interaction is often based on Instagram, Snapchat, Stories, direct image messaging, algorithmic ranking, filters, and status-performance loops. Most of that stack is outside the paper's treatment window. Instagram launched in October 2010, meaning any effect on births within 2007-2011 would be limited to conceptions after late 2010 and would require very rapid adoption (Siegler 2010). Instagram Direct came in December 2013 (Crook 2013). Snapchat began as an early photo-messaging service in 2011 and launched Stories in 2013 (Snap Inc. 2017; Rodriguez 2013).

Thus the app history does not support a modern social-media narrative for 2007-2011 fertility declines. Early Instagram could contribute at the margin for late-2011 births, but it was not yet the mature product associated with algorithmic feeds, influencer culture, Stories, Direct messaging, or Android scale. Snapchat Stories and Instagram Direct are definitively post-window. A paper using 2007-2011 AT&T variation should not rely implicitly on behavioral mechanisms that became salient only after 2012.

A useful falsification is to estimate separate effects around the launch and diffusion of these products. If image-social mechanisms matter, later cohorts and later birth years should show event-study breaks around 2011-2014, with platform-specific timing. If the estimated fertility effect is already large before these products exist, then the effect must be due to a different mechanism: ordinary phone communication, mobile web, games, broader social networking, contraception, recession-era postponement, or selection.

9. Pornography: a possible mechanism with weak device-specific evidence

The pornography mechanism is possible but currently weak as an iPhone mechanism. Myers and Hooper (2026) use national evidence from Google Trends and the General Social Survey to show rising pornography interest and viewing. That evidence is consistent with increasing digital sexual substitution. It does not show that AT&T 3G coverage caused a county-level increase in mobile pornography use, nor that the increase occurred specifically on iPhones.

The identification problem is larger than measurement. Online pornography was widely accessible on PCs before the iPhone, and later mobile access was available through multiple smartphones, feature-phone browsers, Wi-Fi devices, and tablets. If pornography reduces partnered sex, then broadband, private computer access, home Wi-Fi, non-iPhone smartphones, and iPod touch/iPad access are all plausible channels. An AT&T/iPhone design identifies the pornography mechanism only if mobile porn consumption rose differentially in AT&T/iPhone counties and if that rise predicts sexual behavior, pregnancies, and births.

The sign is also not mechanically guaranteed. Pornography may substitute for partnered sex for some users. It may also complement sexual interest, alter partner search, or have heterogeneous effects by sex, age, relationship status, and baseline sexual activity. A mechanism test therefore needs more than a national upward trend in pornography search. It needs geography, platform, timing, and outcomes. The ideal data would be county- or media-market-level adult-content traffic by device class, operating system, and carrier; failing that, the authors could use proxies such as Google Trends with media-market aggregation, broadband exposure, smartphone-platform shares, and STI/pregnancy outcomes. Without such evidence, pornography remains a speculative behavioral channel.

10. Contraception and reproductive information: the strongest rival mechanism

The contraception channel is the strongest alternative explanation because it directly predicts fewer births without requiring a large decline in sexual activity. It is also supported by data from the exact period. Lindberg, Santelli, and Desai (2016) find that improved contraceptive use accounted for the entire decline in adolescent pregnancy risk from 2007 to 2012, while sexual activity in the last three months did not change significantly. The Guttmacher Institute summary reports that improved contraception accounted for the entire 28% decline in teen pregnancy risk during 2007-2012 (Guttmacher Institute 2016).

The later Journal of Adolescent Health update is equally important. Lindberg, Santelli, and Desai (2018) report that among women aged 15-19, use of one or more contraceptive methods at last sex rose from 78% to 88% between 2007 and 2014, dual-method use rose from 24% to 33%, and long-acting reversible contraception rose from 1% to 7%. They also report that the level of sexual activity did not change over time (Lindberg, Santelli, and Desai 2018; Guttmacher Institute 2018).

This evidence is damaging to a simple social-displacement explanation of teen births. If teen births fell because teenagers were much less sexually active, one would expect a substantial decline in contemporaneous sexual-activity measures. The CDC's YRBS also shows little movement over the core window: 47.8% of high-school students had ever had sexual intercourse and 35.0% were currently sexually active in 2007; in 2011 the corresponding figures were 47.4% and 33.7% (CDC 2008; CDC 2012a). The decline in current sexual activity is too small, by itself, to explain the large teen-birth decline.

A contraception-information mechanism could still involve phones. Young people may have used mobile web, social media, SMS, or apps to search for contraception, discuss pregnancy risk, coordinate clinic visits, or receive media messages. But that mechanism is not iPhone-specific. It operates through television, school, clinics, PC internet, Google search, Facebook, texting, and non-iPhone mobile devices. Kearney and Levine (2015) show that MTV's 16 and Pregnant reduced teen births by 4.3% in the 18 months after its initial airing and increased contraception/abortion information-seeking. That is the same order of magnitude as the paper's teen iPhone estimate, and it is explicitly a media/information mechanism rather than an iPhone mechanism.

The proper empirical response is not to treat contraception as background. It should be competed directly against the iPhone channel. A county-year model should include local reproductive-health infrastructure, Title X clinic access, Medicaid family-planning policy, school sex education, LARC diffusion, uninsured rates, Planned Parenthood presence, and media-market exposure to pregnancy-prevention programming. If the iPhone coefficient shrinks or is concentrated where contraception improved, the paper's interpretation changes: the iPhone may be a proxy or amplifier for reproductive information rather than a cause of reduced social interaction.

11. Outcome triangulation: births alone cannot distinguish mechanisms

Birth rates are downstream of several processes: sexual activity, contraception, fecundity, pregnancy intention, abortion, miscarriage, partnership, migration, and timing of births. A birth-only outcome cannot separate these channels. The mechanism tests should therefore triangulate across pregnancies, abortions, STIs, sexual behavior, and contraception.

The public aggregate evidence favors a pregnancy-prevention interpretation over a pure reduction-in-sex interpretation. CDC reported a 25% decline in the teen birth rate from 2007 to 2011, with especially large declines among Hispanic teenagers (CDC 2012b). Over a similar period, CDC abortion surveillance reports that from 2007 to 2011 the number of reported abortions decreased by 26,058 per year and the abortion rate decreased by 0.50 abortions per 1,000 women per year (CDC 2014). Falling births combined with falling abortions points toward fewer pregnancies rather than merely a shift from births to abortions. It does not by itself say whether fewer pregnancies came from less sex or better contraception. The contraception literature points strongly toward the latter for teenagers.

STI surveillance is a useful but imperfect falsification. If the dominant mechanism were sharply reduced sexual contact, one might expect reduced STI incidence among young people, all else equal. CDC's 2011 STD surveillance does not show a clean broad decline: chlamydia reporting reached 1,412,791 cases in 2011 with a rate of 457.6 per 100,000, while CDC emphasizes that changes in screening, testing, and reporting affect interpretation; gonorrhea rates declined over 2007-2011 but rose from 2010 to 2011 (CDC 2013). These data cannot falsify the paper on their own because STI surveillance is affected by testing intensity and case ascertainment. But they do show why births alone are too indirect. A mechanism paper should report whether AT&T/iPhone exposure predicts STIs, pregnancies, abortions, and contraceptive behavior in the same places and cohorts.

The most informative pattern would be as follows. If iPhones reduce in-person sex, treated counties should show lower sexual activity, fewer pregnancies, fewer births, and probably lower STIs. If iPhones improve contraception or reproductive planning, treated counties should show fewer pregnancies and births but little necessary decline in sexual activity or STIs. If iPhones alter pregnancy resolution, births and abortions should move in opposite directions. If the estimated effect is fertility postponement, younger-age births should fall while later-age births rise. These patterns are distinguishable, but not from births alone.

Table 3. Mechanism-specific outcome predictions

12. First-stage calibration remains the decisive test

The mechanism audit reinforces the calibration problem. In May 2011, Pew found that 35% of U.S. adults owned smartphones, with 49% ownership among ages 18-24 and 58% among ages 25-34 (Smith 2011). Pew's 2011 teen survey found that 23% of 12-17-year-olds had smartphones (Lenhart 2012). Comscore reported that Apple had 25.0% of the U.S. smartphone platform market in December 2010 (Comscore 2011a). A rough national teen iPhone penetration estimate around 2011 is therefore in the mid-single digits, although it was surely higher in affluent AT&T-covered places. Even for young adults, iPhone penetration was far from universal.

This does not make an effect impossible. It means the paper must report the implied treatment-on-treated effect. If the county-level first-stage difference in actual iPhone exposure among 15-19-year-olds is small, the implied direct fertility effect per marginal exposed user is very large. That can be reconciled only with strong peer spillovers, partner spillovers, or an interpretation in which AT&T coverage proxies broader county digital maturity. Mechanism evidence is therefore not ancillary. It is required for scale credibility.

The ideal first stage would measure iPhone adoption by county, year, age, and sex. If that is unavailable, the authors could use AT&T subscriber data, Apple activation data, Comscore/Nielsen/Scarborough market data, mobile web user-agent logs, app-download geography, or survey-based small-area estimates. The first-stage variable should be actual exposure to the behaviorally relevant technology, not merely network availability. The difference matters because many treated counties may have AT&T coverage but low adoption among teens, while many control counties may have high social-network/SMS exposure through other devices.

13. Proposed mechanism and falsification tests

The mechanism audit yields a concrete testing agenda. Some tests can be run with the original birth panel and public data. Others require restricted carrier, platform, or app data. The common principle is that each mechanism must be tied to timing, platform, geography, and intermediate outcomes.

13.1. Feature-phone and SMS falsification

Add local measures of cell-phone ownership, SMS intensity, QWERTY-phone penetration, youth texting, and pre-iPhone mobile communication. The strongest available public evidence is national, but commercial sources such as Nielsen, Comscore MobiLens, Scarborough, or carrier billing aggregates may support media-market or county-level proxies. If SMS intensity predicts fertility declines similarly to AT&T iPhone exposure, the mechanism is mediated mobile communication rather than the iPhone.

13.2. iPod touch, Wi-Fi, and broadband placebo tests

Test whether Apple mobile-media exposure without AT&T cellular service predicts similar fertility changes. The iPod touch had Wi-Fi, Safari, YouTube, and the App Store, and Apple bundled iPhone and iPod touch users in early App Store download reports (Apple 2007b; Apple 2008b). If app-mediated leisure, YouTube, games, and social browsing are the channel, then Wi-Fi/broadband and iPod touch diffusion should matter. If only AT&T iPhone coverage matters, the mechanism is likely cellular availability, selection, or the specific social status of the iPhone rather than apps per se.

13.3. Platform handoff tests: Verizon, Android, BlackBerry

Exploit the February 2011 Verizon iPhone launch and the rapid rise of Android. If iPhone access is causal, Verizon-strong counties should begin to exhibit a delayed fertility response after Verizon gets the iPhone, with the appropriate conception-to-birth lag. If mobile social/apps are causal, Android-heavy counties should show similar effects by 2010-2011. If BlackBerry/Facebook/texting channels matter, BlackBerry-heavy places should move as well. A null Verizon handoff or strong Android/BlackBerry effect would weaken the paper's iPhone-specific interpretation.

13.4. App-category timing tests

Estimate separate event-study breaks around the first iPhone, the iPhone 3G/App Store, push notifications, Instagram, Verizon iPhone, iMessage, Tinder, Instagram Direct, and Snapchat Stories. Mechanisms cannot precede their own diffusion. Dating-app effects should begin after 2012 plus gestation. Instagram Direct/Snapchat Stories effects should begin after 2013 plus gestation. If the estimated fertility response is already present before these product events, those mechanisms should be removed from the narrative.

13.5. Intermediate outcome tests

Run the same treatment design on intermediate outcomes: ATUS time with friends, time alone, time on computer/mobile devices, NSFG sexual frequency and contraception, YRBS sexual activity and contraception proxies, STI rates, abortion rates, and pregnancy rates. The current paper's national mechanism series are not enough. The correct question is whether the AT&T/iPhone treatment predicts the intermediate outcomes in the same county, age, sex, and year cells that drive the birth effects.

13.6. Race/ethnicity, nativity, and reproductive-health infrastructure

Decompose the estimates by race/ethnicity, nativity, age, marital status, and education. The teen-birth decline was especially large among Hispanic teenagers (CDC 2012b). If AT&T-treated counties experienced differential Hispanic composition, immigration, labor-market exposure, or access to reproductive-health services, the treatment may pick up compositional and institutional trends. Interact the treatment with Title X access, Medicaid family-planning waivers, LARC availability, Planned Parenthood presence, school sex-education policy, insurance coverage, and clinic density.

13.7. Fertility timing and completed fertility

Separate postponement from completed fertility. A reduction in births at ages 15-24 can be consistent with the same women having births later. If AT&T/iPhone exposure primarily shifts births to older ages, the paper is about timing, not necessarily lower completed fertility. Cohort follow-up through later ages, or at least age-profile accounting, is necessary before describing the effect as a large share of the national fertility decline rather than a change in birth timing.

Table 4. Data needed to adjudicate mechanisms

14. Empirical implementation: calibration, timing, and public-data tests

The critique above becomes more robust when its implications are turned into executable tests. This section adds four implementation components. First, a first-stage calibration table converts the paper's reduced-form birth effects into implied treatment-on-treated magnitudes under alternative assumptions about actual iPhone exposure. Second, a timeline figure imposes a nine-month conception-to-birth lag on product and app events. Third, a mechanism-test matrix states what would support or weaken each behavioral channel. Fourth, a public-data roadmap identifies auditable datasets that can be used to implement or approximate the tests.

14.1. First-stage calibration: how large must the direct iPhone effect be?

The target paper reports reduced-form birth-rate reductions of 4.5-8.0% for ages 15-19 and 3.2-6.6% for ages 20-24. Those effects are population effects. If the actual treatment is individual or peer-group iPhone exposure, the implied direct effect among marginally exposed users depends on the first-stage increase in exposure. The table below does not estimate the first stage; it shows the arithmetic burden placed on the mechanism. Under a no-spillover interpretation, the implied treatment-on-treated effect equals the reduced-form percentage reduction divided by the first-stage percentage-point increase in actual exposure. Spillovers can reduce the direct-effect burden, but then the paper must identify and measure those spillovers.

Table 5. Implied treatment-on-treated birth-rate reductions under alternative iPhone first stages

Note: RF = reduced-form percentage reduction reported by Myers and Hooper (2026). Entries equal RF divided by the assumed first-stage exposure increase. The table is a calibration, not an estimate of actual iPhone ownership. It highlights that small first stages imply very large direct effects or large unmeasured spillovers.

The national adoption benchmarks illustrate why this table matters. Pew reported that only 23% of U.S. teens aged 12-17 owned smartphones in the 2011 teen survey, while Comscore reported that Apple accounted for 25.0% of U.S. smartphone subscribers in December 2010. Multiplying those two national rates gives an approximate teen iPhone benchmark of 5.75%, not a county first stage. In contrast, ordinary cell-phone and SMS channels were already much larger: Pew reported 75% teen cell-phone ownership, 72% of all teens texting, and 54% daily texting in 2009. Figure 1 visualizes the scale contrast.

Figure 1. Teen digital interaction channels before 2011 compared with smartphone/iPhone benchmarks. Sources: Pew teen mobile reports and Comscore December 2010 smartphone-platform report. The final bar is a calculated national benchmark, not a measured county treatment effect.

14.2. Timing discipline: mechanisms cannot precede their own app stack

The product-timing test is simple. A mechanism cannot cause a birth outcome before the mechanism exists and before a conception-to-birth lag has elapsed. The App Store did not open until July 2008 and launched with 500 apps; Apple reported more than 3,000 apps in September 2008 and over 100,000 apps in November 2009. iOS 5/iMessage was announced in June 2011 for fall availability. Tinder launched only after the AT&T-exclusive window. Therefore, a modern app-social, iMessage, swipe-dating, or ephemeral-image mechanism predicts birth effects no earlier than its product date plus roughly nine months, and realistically later once adoption diffusion is allowed.

Figure 2. Product timing and earliest plausible birth effects under a nine-month conception-to-birth lag. The shaded region marks the AT&T-exclusive iPhone window used by the target paper. Events are dated from Apple, Tinder/Hinge/TechCrunch, Snap SEC, and contemporaneous social-app sources.

14.3. Mechanism-test matrix with pass/fail predictions

The central empirical improvement is to state, for each mechanism, what evidence would support the iPhone-specific interpretation and what evidence would move the estimate toward a broader digital-bundle or reproductive-control interpretation. This avoids treating all mechanism evidence as merely 'consistent with' the paper.

Table 6. Mechanism-specific empirical predictions

14.4. Public-data roadmap

Several of the proposed tests can be partially implemented with public or restricted-public data. The strongest iPhone-first-stage and app-category tests require commercial or platform data, but the birth, STI, contraception, demographic, economic, housing, and broadband robustness families are auditable and feasible. The table below separates what can be done immediately from what requires restricted or commercial access.

Table 7. Public and restricted-public data sources for implementing falsification tests

14.5. What evidence would make the iPhone interpretation persuasive?

The iPhone-specific interpretation would become substantially more credible if four conditions held jointly. First, AT&T 3G coverage must strongly predict actual iPhone exposure among teens, young women, partners, or peer groups, not merely average county technology availability. Second, estimated birth effects must begin only after relevant product events plus gestation: the July 2008 App Store for app mechanisms, 2009 push notifications for notification-loop mechanisms, late 2010/2011 Instagram for photo-social mechanisms, and post-2012 for dating-app mechanisms. Third, feature-phone/SMS intensity, PC social networking, BlackBerry/Android adoption, iPod touch/Wi-Fi exposure, broadband, and local economic/reproductive-health trends must not explain the same pattern. Fourth, intermediate outcomes must move in the predicted direction: less in-person interaction and lower sexual frequency for a social-displacement mechanism; lower pregnancies with stable sex for a contraception-information mechanism; STI movements consistent with a less-sex mechanism; and pregnancy/abortion patterns that distinguish fewer conceptions from changed pregnancy resolution.

This paragraph is intentionally symmetric. It identifies what evidence would weaken this comment as well as what evidence would weaken the target paper. A strong county-year first stage for actual iPhone adoption, a clean Verizon handoff after February 2011, null effects for SMS/broadband/iPod touch/Android/BlackBerry placebos, and mechanism-consistent movements in pregnancies, STIs, sexual behavior, and contraception would materially shift the balance toward Myers and Hooper's interpretation.

15. Conclusion

The mechanism audit changes the interpretation of Myers and Hooper (2026). The reduced-form AT&T 3G estimate may be real and important. But the proposed behavioral explanation is not yet identified. The most scalable mechanism during the treatment window - mediated teen interaction - was already available through ordinary cell phones and SMS. PC social networking was already widespread among teens. Mobile social networking existed by 2011, but it was cross-platform and included non-smartphone mobile access. The native app stack was still forming. The modern dating-app and ephemeral image-social channels are largely post-treatment. Pornography is plausible but not measured at the required device/geography level. Contraceptive improvement is directly supported by period-specific data and can explain falling births without a large decline in sexual activity.

The most defensible version of the paper is therefore narrower than its title: early mobile internet and broader digital connectedness may have contributed to fertility postponement and pregnancy prevention in counties at the frontier of digital adoption. That hypothesis is plausible and worth testing. It is not the same as showing that the iPhone itself caused one-third to one-half of the national fertility decline.

A revised paper should explicitly separate four causal objects: AT&T 3G coverage, iPhone ownership, smartphone/mobile internet use, and the mature app-social ecosystem. It should then test mechanisms against their own timing and platform predictions. Without those tests, the iPhone remains a suggestive proxy for a broader bundle rather than a demonstrated birth-control technology.

Data and code availability

This comment does not introduce a new restricted microdata analysis and does not replicate Myers and Hooper's county-year estimates. It adds transparent descriptive calibrations and source-audited mechanism figures based on published aggregate values. The proposed falsification tests require the original county-year fertility panel and, for the strongest mechanism tests, carrier, platform, app, survey, commercial media-market, or restricted geocoded health data. The calibration tables in this draft are algebraic sensitivity checks, not substitute estimates of the actual iPhone first stage.

Competing interests

The author declares no competing interests.

Source audit note

References include auditable URLs or DOI landing pages. Official sources are used where possible: Apple Newsroom for product timing, Pew for survey adoption and texting, Comscore/Nielsen for mobile-platform and app-use context, CDC for births/abortions/STIs/YRBS, publisher or PubMed pages for contraceptive-use papers, the AEA page for Kearney and Levine, and the NBER/DOI page for the target paper. A separate mechanism source-audit log accompanies this manuscript.

References

Apple. 2007a. 'iPhone Premieres This Friday Night at Apple Retail Stores.' Apple Newsroom, June 28. Available at: https://www.apple.com/newsroom/2007/06/28iPhone-Premieres-This-Friday-Night-at-Apple-Retail-Stores/

Apple. 2007b. 'Apple Unveils iPod touch.' Apple Newsroom, September 5. Available at: https://www.apple.com/newsroom/2007/09/05Apple-Unveils-iPod-touch/

Apple. 2008a. 'AT&T and Apple to Launch iPhone 3G in the U.S. on July 11.' Apple Newsroom, June 9. Available at: https://www.apple.com/newsroom/2008/06/09AT-T-and-Apple-to-Launch-iPhone-3G-in-the-US-on-July-11/

Apple. 2008b. 'iPhone App Store Downloads Top 10 Million in First Weekend.' Apple Newsroom, July 14. Available at: https://www.apple.com/newsroom/2008/07/14iPhone-App-Store-Downloads-Top-10-Million-in-First-Weekend/

Apple. 2008c. 'App Store Downloads Top 100 Million Worldwide.' Apple Newsroom, September 9. Available at: https://www.apple.com/newsroom/2008/09/09App-Store-Downloads-Top-100-Million-Worldwide/

Apple. 2009a. 'Apple Previews Developer Beta of iPhone OS 3.0.' Apple Newsroom, March 17. Available at: https://www.apple.com/newsroom/2009/03/17Apple-Previews-Developer-Beta-of-iPhone-OS-3-0/

Apple. 2009b. 'Apple Announces Over 100,000 Apps Now Available on the App Store.' Apple Newsroom, November 4. Available at: https://www.apple.com/newsroom/2009/11/04Apple-Announces-Over-100-000-Apps-Now-Available-on-the-App-Store/

Apple. 2011a. 'New Version of iOS Includes Notification Center, iMessage, Newsstand, Twitter Integration Among 200 New Features.' Apple Newsroom, June 6. Available at: https://www.apple.com/newsroom/2011/06/06New-Version-of-iOS-Includes-Notification-Center-iMessage-Newsstand-Twitter-Integration-Among-200-New-Features/

Apple. 2011b. 'Verizon Wireless & Apple Team Up to Deliver iPhone 4 on Verizon.' Apple Newsroom, January 11. Available at: https://www.apple.com/newsroom/2011/01/11Verizon-Wireless-Apple-Team-Up-to-Deliver-iPhone-4-on-Verizon/

Apple. 2011c. 'Apple Launches iPhone 4S, iOS 5 & iCloud.' Apple Newsroom, October 4. Available at: https://www.apple.com/newsroom/2011/10/04Apple-Launches-iPhone-4S-iOS-5-iCloud/

Apple. 2018. 'The App Store Turns 10.' Apple Newsroom, July 5. Available at: https://www.apple.com/newsroom/2018/07/app-store-turns-10/

Buckles, Kasey, Melanie Guldi, and Lucie Schmidt. 2025. 'The Great Recession's Baby-Less Recovery: The Role of Unintended Births.' Journal of Human Resources 60(1): 224-258. Available at: https://doi.org/10.3368/jhr.1220-11395R3; https://ideas.repec.org/a/uwp/jhriss/v60y2025i1p224-258.html

CDC. 2012a. 'Youth Risk Behavior Surveillance - United States, 2011.' MMWR Surveillance Summaries 61(SS-4): 1-162. Available at: https://www.cdc.gov/mmwr/preview/mmwrhtml/ss6104a1.htm; https://pubmed.ncbi.nlm.nih.gov/22673000/

CDC. 2012b. 'Birth Rates for Females Aged 15-19 Years, by Race/Ethnicity - National Vital Statistics System, United States, 2007 and 2011.' MMWR 61(42): 865. Available at: https://www.cdc.gov/mmwr/preview/mmwrhtml/mm6142a8.htm

CDC. 2013. Sexually Transmitted Disease Surveillance 2011. Atlanta: U.S. Department of Health and Human Services. Available at: https://www.cdc.gov/sti-statistics/media/pdfs/2024/07/Surv2011.pdf

CDC. 2014. 'Abortion Surveillance - United States, 2011.' MMWR Surveillance Summaries 63(SS-11): 1-41. Available at: https://www.cdc.gov/mmwr/preview/mmwrhtml/ss6311a1.htm; https://pubmed.ncbi.nlm.nih.gov/25426741/

CDC. 2008. 'Youth Risk Behavior Surveillance - United States, 2007.' MMWR Surveillance Summaries 57(SS-4): 1-131. Available at: https://www.cdc.gov/mmwr/preview/mmwrhtml/ss5704a1.htm; https://pubmed.ncbi.nlm.nih.gov/18528314/

Comscore. 2011a. 'Comscore Reports December 2010 U.S. Mobile Subscriber Market Share.' February 7. Available at: https://www.comscore.com/Insights/Press-Releases/2011/2/comScore-Reports-December-2010-US-Mobile-Subscriber-Market-Share

Comscore. 2011b. 'Comscore Reports June 2011 U.S. Mobile Subscriber Market Share.' August 4. Available at: https://www.comscore.com/Insights/Press-Releases/2011/8/comScore-Reports-June-2011-US-Mobile-Subscriber-Market-Share

Comscore. 2011c. 'Social Networking On-The-Go: U.S. Mobile Social Media Audience Grows 37 Percent in the Past Year.' October 20. Available at: https://ir.comscore.com/news-releases/news-release-details/social-networking-go-us-mobile-social-media-audience-grows-37

Comscore. 2011d. 'Digital Omnivores: How Tablets, Smartphones and Connected Devices are Changing U.S. Digital Media Consumption Habits.' October 10. Available at: https://www.comscore.com/Insights/Presentations-and-Whitepapers/2011/Digital-Omnivores

Crook, Jordan. 2013. 'Instagram Introduces Instagram Direct.' TechCrunch, December 12. Available at: https://techcrunch.com/2013/12/12/instagram-messaging/

Empson, Rip. 2013. 'Tinder: Finding Traction On Campuses, IAC's New Dating App Makes It Easy To Break The Ice.' TechCrunch, January 3. Available at: https://techcrunch.com/2013/01/03/tinder-finding-traction-on-campuses-hatch-labs-new-dating-app-makes-it-easy-to-break-the-ice/

Guttmacher Institute. 2016. 'Declines in Teen Pregnancy Risk Entirely Driven by Improved Contraceptive Use.' News release, August 30. Available at: https://www.guttmacher.org/news-release/2016/declines-teen-pregnancy-risk-entirely-driven-improved-contraceptive-use

Guttmacher Institute. 2018. 'Improvements in Contraceptive Use Continue to Drive Declines in Pregnancy Among U.S. Adolescents.' News release, August 30. Available at: https://www.guttmacher.org/news-release/2018/improvements-contraceptive-use-continue-drive-declines-pregnancy-among-us

Hinge. n.d. 'Our Story.' Hinge official website. Accessed June 9, 2026. Available at: https://hinge.co/our-story

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. Available at: https://doi.org/10.1257/aer.20140012; https://www.aeaweb.org/articles?id=10.1257/aer.20140012

Lenhart, Amanda. 2010. 'Teens and Mobile Phones.' Pew Research Center, April 20. Available at: https://www.pewresearch.org/internet/2010/04/20/teens-and-mobile-phones/

Lenhart, Amanda. 2012. 'Teens, Smartphones & Texting.' Pew Research Center, March 19. Available at: https://www.pewresearch.org/internet/2012/03/19/teens-smartphones-texting/

Lindberg, Laura D., John S. Santelli, and Sheila Desai. 2016. 'Understanding the Decline in Adolescent Fertility in the United States, 2007-2012.' Journal of Adolescent Health 59(5): 577-583. Available at: https://doi.org/10.1016/j.jadohealth.2016.06.024; https://www.jahonline.org/article/S1054-139X(16)30172-0/fulltext

Lindberg, Laura D., John S. Santelli, and Sheila Desai. 2018. 'Changing Patterns of Contraceptive Use and the Decline in Rates of Pregnancy and Birth Among U.S. Adolescents, 2007-2014.' Journal of Adolescent Health 63(2): 253-256. Available at: https://doi.org/10.1016/j.jadohealth.2018.05.017; https://pubmed.ncbi.nlm.nih.gov/30149926/

Myers, Caitlin K., and Ezekiel Hooper. 2026. 'Is the iPhone Birth Control? Causal Evidence from AT&T's 2007-2011 Carrier Monopoly.' NBER Working Paper No. 35310. Available at: https://doi.org/10.3386/w35310; https://www.nber.org/papers/w35310; https://www.nber.org/system/files/working_papers/w35310/w35310.pdf

Nielsen. 2010. 'The State of Mobile Apps.' Nielsen, June 2010. Available at: https://www.nielsen.com/insights/2010/the-state-of-mobile-apps-2/

Pew Research Center. 2010a. 'Social Media and Young Adults.' February 3. Available at: https://www.pewresearch.org/internet/2010/02/03/social-media-and-young-adults/; https://www.pewresearch.org/internet/2010/02/03/part-3-social-media/

Pew Research Center. 2010b. 'Part 2: Gadget Ownership and Wireless Connectivity.' February 3. Available at: https://www.pewresearch.org/internet/2010/02/03/part-2-gadget-ownership-and-wireless-connectivity/

Rodriguez, Salvador. 2013. 'Snapchat Launches Stories Feature.' Los Angeles Times, October 3. Available at: https://www.latimes.com/business/technology/la-fi-tn-snapchat-stories-feature-20131003-story.html

Siegler, MG. 2010. 'Instagram Launches with the Hope of Igniting Communication Through Images.' TechCrunch, October 6. Available at: https://techcrunch.com/2010/10/06/instagram-launch/

Smith, Aaron. 2011. 'Smartphone Adoption and Usage.' Pew Research Center, July 11. Available at: https://www.pewresearch.org/internet/2011/07/11/smartphone-adoption-and-usage/

Smith, Aaron, and Maeve Duggan. 2013. 'Online Dating & Relationships.' Pew Research Center, October 21. Available at: https://www.pewresearch.org/internet/2013/10/21/online-dating-relationships-3/; https://www.pewresearch.org/internet/2013/10/21/part-2-dating-apps-and-online-dating-sites/; https://www.pewresearch.org/wp-content/uploads/sites/9/media/Files/Reports/2013/PIP_Online-Dating-2013.pdf

Snap Inc. 2017. Registration Statement on Form S-1. U.S. Securities and Exchange Commission, February 2. Available at: https://www.sec.gov/Archives/edgar/data/1564408/000119312517029199/d270216ds1.htm

Tinder. n.d. 'About Tinder.' Tinder Press Room. Accessed June 9, 2026. Available at: https://www.tinderpressroom.com/about

Additional public-data documentation for robustness modules:

BLS. n.d. 'American Time Use Survey.' U.S. Bureau of Labor Statistics. Accessed June 9, 2026. Available at: https://www.bls.gov/tus/

BLS. n.d. 'Local Area Unemployment Statistics: Tables and Maps.' U.S. Bureau of Labor Statistics. Accessed June 9, 2026. Available at: https://www.bls.gov/lau/tables.htm

CDC. n.d. 'Natality Information.' CDC WONDER. Accessed June 9, 2026. Available at: https://wonder.cdc.gov/natality.html

CDC. n.d. 'About AtlasPlus.' National Center for HIV, Viral Hepatitis, STD, and TB Prevention. Accessed June 9, 2026. Available at: https://www.cdc.gov/nchhstp/about/atlasplus.html

CDC. n.d. 'National Survey of Family Growth.' National Center for Health Statistics. Accessed June 9, 2026. Available at: https://www.cdc.gov/nchs/nsfg/index.htm

CDC. n.d. 'Restricted Data: National Survey of Family Growth.' Research Data Center. Accessed June 9, 2026. Available at: https://www.cdc.gov/rdc/b1datatype/Dt1226.htm

CDC. n.d. 'Youth Risk Behavior Surveillance System.' Accessed June 9, 2026. Available at: https://www.cdc.gov/yrbs/index.html

Federal Housing Finance Agency. n.d. 'FHFA House Price Index.' Accessed June 9, 2026. Available at: https://www.fhfa.gov/data/hpi

NTIA. n.d. 'June 30, 2010 National Broadband Map Datasets.' State Broadband Initiative archive. Accessed June 9, 2026. Available at: https://www2.ntia.gov/June-2010-datasets

NTIA. n.d. 'June 30, 2011 National Broadband Map Datasets.' State Broadband Initiative archive. Accessed June 9, 2026. Available at: https://www2.ntia.gov/Jun-2011-datasets

U.S. Bureau of Economic Analysis. n.d. 'Personal Income by County.' Accessed June 9, 2026. Available at: https://www.bea.gov/data/income-saving/personal-income-by-county

U.S. Census Bureau. n.d. 'American Community Survey 5-Year Data (2009-2024).' Accessed June 9, 2026. Available at: https://www.census.gov/data/developers/data-sets/acs-5year.html