The question beneath the behavior
The work began with a hard question: if young people know that certain choices can harm them, why can those choices still feel reasonable in the moment?
Abraham Humphrey proposed a multicausal explanation. He asked whether shifts in family expectations, social pressure, social media and influencer culture, weakened friendship and romantic connection, normalized cheating, gaslighting, and disrespect, and changing substance-use patterns might help explain why certain risks can feel reasonable to young people. He also asked whether vaping might help explain lower reported drinking among Gen Z. Life that Counts treated those observations as questions that generated tests—not as conclusions the data were made to confirm.
There is no defensible date when a generation “went bad.” Reported sexual activity and alcohol use have fallen substantially over the long term. Persistent sadness and hopelessness rose. Some forms of risk became more private, digital, immediate, discreet, and easily concealed. A highly visible minority can make a behavior feel universal even when most young people are not doing it.
Adolescent risk becomes more likely when ordinary desires for belonging, intimacy, status, autonomy, and relief encounter exaggerated peer norms, inconsistent relationships, distress or trauma, digital pressure, limited confidential help, and an uncertain or unconvincing path into adulthood.
Evidence posture: Understanding what a behavior does for a young person in the moment is not the same as approving it. Coercion, exploitation, assault, and abuse are never the victim’s fault. This article denormalizes risk without demoralizing the young person.
What the evidence supported, challenged, and left unresolved
This matrix records the proposition tested, the evidence checked, and the bounded conclusion. It distinguishes evidence that challenged an initial explanation from mechanisms that remain plausible and questions that remain unresolved.
Proposition | Evidence checked | Conclusion |
|---|---|---|
Traditional risk behavior increased across the board. | National and Alabama YRBS trends. [1][2][3] | Challenged. Several measured sexual- and substance-risk indicators declined while emotional distress increased. |
Peer visibility and perceived approval can make risk feel normal. | Descriptive- and injunctive-norm research and the rural peer-influence experiment. [4] | Plausible and supported as a mechanism, but the experiment measured hypothetical responses rather than later behavior. |
Family monitoring and school connection may be protective. | National 2021 YRBS analyses. [5][6] | Supported as cross-sectional associations; causation was not established. |
Social media and influencer culture caused the change. | Trend timing, YRBS context, and available local-data limitations. [1][2] | Unresolved. Digital environments may increase visibility and rapid feedback, but this review cannot isolate them as causes. |
Vaping explains lower adolescent drinking. | Timing of alcohol and vaping trends and prospective cross-substance research. [12][13][14] | Not supported as the primary historical explanation. Substitution may occur in particular settings while clustering and shared vulnerability remain plausible. |
Accountable work and mentoring relationships cause healthier behavior. | Direct observation, positive-youth-development mechanisms, and historical program evidence. [15][16][17] | Untested practice hypothesis—not demonstrated causal impact. |
Narrative description: the trend claim was challenged; peer norms and connected relationships remain plausible mechanisms with important design limits; social media causation remains unresolved; vaping is not supported as the principal historical explanation for lower drinking; and accountable mentoring remains an untested practice hypothesis.
The trends challenged the easy story
Nationally, the share of high-school students reporting that they had ever had sexual intercourse fell from 47% in 2013 to 32% in 2023. Current sexual activity fell from 34% to 21%. Current alcohol use fell from 35% to 22%, and current marijuana use fell from 23% to 17%. Over the same period, persistent sadness or hopelessness rose from 30% to 40%. [1]
Reported national indicator | 2013 | 2023 | Direction |
|---|---|---|---|
Ever had sexual intercourse | 47% | 32% | Down 15 points |
Currently sexually active | 34% | 21% | Down 13 points |
Current alcohol use | 35% | 22% | Down 13 points |
Current marijuana use | 23% | 17% | Down 6 points |
Persistent sadness or hopelessness | 30% | 40% | Up 10 points |
Source: CDC, 2013–2023 YRBS Trends Report. National high-school surveys show lower reported sexual activity, alcohol use, and marijuana use in 2023 than in 2013, while persistent sadness or hopelessness was higher. These measures do not describe every sexual, relational, digital, or substance-use experience.
Three different meanings of “normal”
Actual behavior is what peers report doing. A descriptive norm is what a young person believes most peers do. An injunctive norm is what a young person believes respected peers approve, reward, or expect. Those three do not have to match.
In an experiment involving 300 early adolescents from three rural, low-income Southeastern middle schools, 236 of 300—about 79%—gave a riskier hypothetical answer after seeing supposedly peer-endorsed responses in a simulated chat. The study had no control condition, could not rule out demand effects, and did not measure later sexual behavior. It does not prove that a particular peer message causes real-world behavior. [4]
Why a risky choice can feel useful
A useful prevention model asks what immediate function a behavior may serve: belonging and status, reassurance and intimacy, coping and relief, digital visibility, or a shortcut around an uncertain future. Prevention has to replace the function as well as prohibit the behavior.
National 2021 YRBS analyses found that high parental monitoring and school connectedness were associated with lower prevalence across many risks. These cross-sectional analyses do not show that monitoring or connectedness caused the differences. They do suggest that standards are more likely to hold when healthy behavior is connected to warmth, membership, respect, and a safe path for disclosure. [5][6]
Alabama trend points
Weighted secondary analysis of CDC public-use Alabama high-school YRBS files shows a mixed long-term pattern. These are original weighted Alabama point estimates, not a CDC-published Alabama trend table. The 2019 public-use file contains 2,040 Alabama records; the 2021 file contains 419. Item-specific denominators are smaller and vary. Verified confidence intervals are not displayed in this article. The Alabama values are therefore presented as historical descriptive estimates only. No statistical-significance claim or inferential comparison is made beyond the reported estimates. Alabama did not produce a representative high-school YRBS estimate in 2023. These are not Winston County estimates. [2][3]
Indicator | Earlier point | 2019 | 2021 |
|---|---|---|---|
Ever had sexual intercourse | 56.9% (2003) | 40.7% | 28.8% |
Sexually active in previous 3 months | 41.9% (2003) | 30.6% | 18.8% |
Current alcohol use | 46.8% (1991) | 22.5% | 18.8% |
Current marijuana use | 19.2% (2013) | 15.3% | 11.3% |
Current electronic-vapor-product use | 24.5% (2015) | 19.4% | 17.5% |
Persistent sadness or hopelessness | 27.0% (1999) | 37.5% | 36.9% |
Seriously considered suicide | 16.9% (1999) | 20.7% | 21.6% |
The defensible conclusion is not “young people are doing more of everything.” Several traditional risk behaviors are less prevalent than in earlier decades. Emotional distress is more prevalent. Risk has partly migrated toward private, digital, immediate, and concealable forms. Visibility and perceived approval can rise even when reported behavior declines.
What is specifically rural—and what is not
Winston County is part of Alabama’s Appalachian region. Rural Alabama should not be treated as a smaller or deficient version of an urban community. It has distinctive constraints and distinctive protective assets. [7]
Current local indicators describe serious access and opportunity conditions. Winston County’s teen-birth rate for 2018–2024 is 35.0 births per 1,000 females ages 15–19, compared with 22.5 for Alabama and 14.6 nationally. In 2025, the county had approximately one mental-health provider per 4,774 residents, compared with 1:639 statewide and 1:287 nationally. About 65.7% of children were eligible for free or reduced-price lunch, and estimated food insecurity was 20.2%. These indicators do not establish why any individual behaves as they do. Teen birth rates are not the same as current high-school sexual activity. [8][9]
Context indicator | Winston County | Alabama | United States |
|---|---|---|---|
Teen births per 1,000, ages 15–19 | 35.0 | 22.5 | 14.6 |
Residents per mental-health provider | 4,774:1 | 639:1 | 287:1 |
Children eligible for free/reduced-price lunch | 65.7% | 60.8% | 56.1% |
Food insecurity | 20.2% | 17.5% | 14.3% |
Did vaping replace drinking?
Substitution may occur in particular situations, but temporal order does not support vaping as the principal historical explanation for the decades-long decline in adolescent drinking. Prospective research links e-cigarette use with later alcohol, marijuana, cigarette, and nonmedical prescription-drug use; that pattern is more consistent with clustering or shared vulnerability than a simple replacement account. These associations do not prove that vaping caused later substance use. [12][13][14]
Three complementary evidence lenses: YRBS, RELENTLESS, and RAMP
These datasets answer different questions and must remain analytically separate. They must not be merged into one trend, pooled sample, or direct effectiveness comparison. Sample size answers how much information was collected. Study design answers which inference the information can support.
Evidence lens | Strongest use | Cannot establish |
|---|---|---|
YRBS | Weighted population prevalence and trends for the geography and year reported | Winston prevalence, program experience, causal program effects, or the full digital and relational environment |
RELENTLESS | Historical participant baseline/post patterns, implementation history, and experience | Representative prevalence, within-person change from unmatched surveys, or causal impact without an equivalent comparison group |
RAMP | Historical middle-school implementation, survey flow, participant experience, retrospective self-assessment, and evaluator recommendations. | Unique participant counts, retention, individual change, long-term behavior change, or causal effectiveness. |
More records. Different purpose.
Life that Counts collected substantially more RAMP and RELENTLESS survey records than the 419 records in Alabama’s 2021 public-use YRBS file. That does not make the program records stronger prevalence evidence. Sample size answers how much information was collected. Study design answers which inference the information can support.
Dataset | Records | Permitted interpretation |
|---|---|---|
Alabama 2021 YRBS | 419 unweighted public-use high-school records | Weighted statewide prevalence; not Winston County prevalence. |
RELENTLESS 2019 Alabama | 434 baseline; 343 post-program | Unmatched historical program records. |
RELENTLESS 2020–2021 Alabama | 183 baseline; 104 post-program | Unmatched historical program records. |
RELENTLESS 2020–2021 both states | 389 baseline; 247 post-program | Unmatched historical program records; item denominators may vary. |
RAMP/WPRE | 2,541 entry; 2,204 exit survey administrations | Historical middle-school survey flow across Alabama, Montana, and three curriculum levels; not verified unique participants or retention. |
RELENTLESS: bounded historical evidence
In 2019, RELENTLESS collected 434 baseline and 343 post-program questionnaire records in Alabama. Baseline and post surveys were unmatched. Adjusted comparisons on 0–100 scales showed positive changes in abstinence/risk-avoidance attitude (+6.6, p=.0003), risk-reduction knowledge (+10.9, p=.0001), and risk-reduction skill (+4.8, p=.0002). Behavior intention changed +3.2 and was not statistically significant. These are scale-point estimates, not percentages of participants who improved; no individual change or causal impact can be inferred. [15]
The pandemic-affected 2020–2021 cycle included 183 baseline and 104 post-program records in Alabama, and 389 baseline and 247 post-program records across both states. It did not reproduce significant positive changes in the principal Alabama outcomes. Reported unfavorable changes included an Alabama positive-future-attitude estimate of −4.6 scale points (p=.033). Participant-experience findings were 79.8% reporting clear material, 73.2% reporting that discussions or activities helped them learn, and 84.5% reporting feeling respected, all or most of the time. The combined file reported 247 post surveys but did not provide an item-specific denominator for every percentage; 247 is not the denominator for each item. Historical RELENTLESS findings are not current BELIEVE effectiveness findings. [15]
RAMP: descriptive historical evidence
Across three historical middle-school curriculum levels delivered in Alabama and Montana from 2020 through 2022, Washington Park Research and Evaluation compiled 2,541 entry and 2,204 exit survey administrations. Dividing those totals produces an aggregate exit-to-entry survey-record ratio of 86.7%. Because entry and exit records were not linked to the same individuals, that ratio is not a participant completion rate, retention rate, or measure of individual change. [16]
Historical level | Entry survey administrations | Exit survey administrations |
|---|---|---|
Level 1 | 1,910 | 1,681 |
Level 2 | 526 | 414 |
Level 3 | 105 | 109 |
The historical report does not identify the precise reason Level 3 had more exit records than entry records. These counts must not be interpreted as unique participants, retention, completion, or within-person change.
Post-program measure | Level 1 | Level 2 | Level 3 |
|---|---|---|---|
Material was clear all/most of the time | 85.3% (1,341/1,572) | 88.0% (346/393) | 86.9% (93/107) |
Felt respected all/most of the time | 83.5% (1,305/1,562) | 88.0% (346/393) | 85.0% (91/107) |
More likely to plan toward goals | 79.8% (1,259/1,578) | 86.8% (264/304) | 73.6% (78/106) |
Greater understanding of healthy relationships | 81.7% (1,326/1,623) | 84.5% (257/304) | 79.2% (84/106) |
These are level-specific post-program participant reports. Item denominators vary. The retrospective items were not measured at baseline, entry and exit records were not student-linked, and no comparison group, confidence intervals, effect sizes, or significance tests were reported for these items. Ninety-six early Level 2 exit respondents were not asked the retrospective items. The evidence supports statements about historical implementation, participant experience, and organizational learning—not causal effectiveness or the effectiveness of current BELIEVE content. [16]
What accountable relationships may change
The federal Sexual Risk Avoidance Education literature describes youth behavior as multilevel: individual development, parents and caregivers, peers and partners, school and community connection, and media all matter. Current federal program language also emphasizes positive youth development and trauma-informed practice. Those requirements do not prove that a local practice causes outcomes. [16][17]
Mentoring in ordinary places: the practice hypothesis
In the gym, on the farm, and alongside an earthwork crew, John Williams has watched young men respond to being known, expected to contribute, corrected without rejection, trusted with responsibility, and invited to imagine a future larger than immediate impulse. Life that Counts’ working practice hypothesis is that these ordinary, repeated relationships may contain several protective ingredients emphasized in positive youth development: structured settings, close relationships, positive adult role models, meaningful skills, contribution, self-worth, and belonging.
That observation helps define a research question; it does not answer it. Life that Counts has not isolated which element produced which outcome, compared participants with an equivalent group, or established that mentoring caused changes in sexual or substance-use behavior. Practice hypothesis—not a demonstrated causal pathway: trusted-adult connection, program belonging, self-worth, emotion regulation, goal clarity, accurate perceived peer norms, relationship knowledge, refusal confidence, future orientation, and help-seeking confidence may be measurable mechanisms worth testing.
Life that Counts’ organizational position
Life that Counts treats sexual intimacy as meaningful and properly situated within marriage and stable family life. Its prevention posture includes medically accurate, age-appropriate education; the voluntary avoidance of nonmarital sexual activity and substance use; healthy relationships and future planning; and explicit protection against pressure, manipulation, threats, guilt, force, coercion, exploitation, dating violence, assault, and abuse. A past choice does not erase a young person’s dignity or determine their future.
A six-part social-norming framework
Measure locally. Find the truthful protective norm. State it precisely with population, geography, survey year, question, and timeframe. Use credible messengers. Pair the norm with practiced efficacy. Repeat and test whether perceived norms become more accurate and whether belonging, skills, intentions, help-seeking, and later behavior change.
Use language such as: “In the 2023 national survey, 21% of high-school students reported current sexual activity. Nearly four in five did not.” “Choosing not to have sex—or choosing to stop—is a real and respected option.” “A past choice does not erase your dignity or determine your future.” Avoid shame, disgust, humiliation, rural stereotypes, fear-only messaging, or high-risk prevalence presented so prominently that it becomes an advertisement.
What this means for program design
A defensible rural Alabama theory of change is that youth are more likely to avoid nonmarital sexual activity and substance use when they experience accurate healthy-peer norms, warm and accountable adult relationships, meaningful belonging, practiced self-regulation and refusal skills, confidential pathways to help, and attainable educational, occupational, marital, and family futures. This is a theory to test, not a demonstrated BELIEVE effect.
Practical design choices include locally measured norm messages, credible peer messengers with clear boundaries, repeated small-group contact, warm caregiver monitoring, realistic skills rehearsal, meaningful alternatives and contribution, concrete local pathways into adulthood, honest referral limits, and measurement of mechanisms as well as behavior.
How we examined—and tried to disprove—the theory
Converted Abraham’s observations into separate testable propositions.
Checked whether measured behaviors had actually increased.
Compared like YRBS questions across reported years.
Preserved exact geography and survey year.
Distinguished behavior from visibility, perceived prevalence, and perceived approval.
Displayed observed years without inventing a smooth timeline.
Treated Alabama 2021 as historical statewide context, not a 2026 Winston County estimate.
Refused to use teen births as a proxy for current high-school sexual activity.
Used temporal order to test the vaping-substitution proposition.
Checked polysubstance clustering and shared vulnerability as rival explanations.
Separated association, participant experience, pre/post estimates, and causal impact.
Preserved favorable, null, and unfavorable historical findings.
Did not convert participant satisfaction into behavior change.
Treated direct observation as hypothesis-generating.
Treated federal language as program context rather than endorsement.
Kept organizational values separate from empirical findings.
Technical note: Alabama YRBS secondary analysis
The displayed estimates identify the survey year, Alabama geography, and unweighted public-use file counts where verified. Item-specific denominators vary and missing responses were excluded from each reported item. The current public record does not identify verified weight, stratum, or PSU variable names; software and survey-design procedure; item-specific 95% confidence intervals; or a publicly archived code file. Those details are therefore not invented here, and the article avoids inferential claims beyond the displayed weighted point estimates. [2][3]
The question we should carry forward
The deepest issue is not simply that young people have stopped hearing rules. It is that many do not consistently experience a community in which truth, belonging, adult integrity, self-control, forgiveness, and an attainable future reinforce one another.
The work ahead is larger than delivering a warning. It is to make the healthy choice credible, socially supported, practiced, and possible—to denormalize the risky behavior without demoralizing the young person. That is the testable promise behind Life that Counts’ next work, not a claim that one relationship has already proved the answer.