# Reconstruction: Does the Chart Add Up to Real Empirical Data?

You asked for time-series exposure data **only where real measurements exist** (no extrapolation), plus W values, and to see whether the chart can be reproduced from them. The honest answer has three parts.

## Headline finding

1. **Empirical data exists for ~12 of 27 five-year periods**, mostly post-1965, very sparsely pre-1939. Pre-1965 has effectively no continuous child-time-use data — only adjacent proxies (school enrollment, female labor force participation, TV/radio penetration, Gallup church attendance from 1939). See `Empirical_Anchors_RAW.csv` for what actually exists.
2. **When you apply reasonable W values to the empirical exposure data, the resulting allocations don't match the chart cleanly.** Specifically: the chart over-weights Church and TV pre-2000 and under-weights Parents and School relative to what a straightforward E × W calculation produces.
3. **The chart can be made to "add up" only by using elevated W values for Church (per hour, more formative than parents) and TV (per hour, comparable to parents).** Those W settings are defensible as a normative claim about ritual and broadcast formation, but they are not derivable from the literature — they are tuning. You should know this if you're going to defend the chart.

---

## Coverage map: where real data exists

| Period | Parents | Church | School | Peers | TV/Broadcast | Internet | Algorithm |
|---|---|---|---|---|---|---|---|
| 1900 | ✗ | ✗ | ◐ enrollment only | ✗ | n/a (=0) | n/a (=0) | n/a (=0) |
| 1905 | ✗ | ✗ | ◐ | ✗ | n/a | n/a | n/a |
| 1910 | ✗ | ✗ | ◐ | ✗ | n/a | n/a | n/a |
| 1915 | ✗ | ✗ | ◐ | ✗ | n/a | n/a | n/a |
| 1920 | ✗ | ✗ | ◐ | ✗ | ◐ radio | n/a | n/a |
| 1925 | ✗ | ✗ | ◐ | ✗ | ◐ radio | n/a | n/a |
| 1930 | ✗ | ✗ | ◐ | ✗ | ◐ radio | n/a | n/a |
| 1935 | ✗ | ✗ | ◐ | ✗ | ◐ radio | n/a | n/a |
| 1940 | ✗ | ✓ Gallup | ◐ | ✗ | ◐ radio | n/a | n/a |
| 1945 | ✗ | ✓ | ◐ | ✗ | ◐ radio | n/a | n/a |
| 1950 | ✗ | ✓ | ✓ | ✗ | ✓ penetration | n/a | n/a |
| 1955 | ✗ | ✓ | ✓ | ✗ | ✓ | n/a | n/a |
| 1960 | ✗ | ✓ | ✓ | ✗ | ✓ | n/a | n/a |
| 1965 | ✓ Bianchi | ✓ | ✓ | ✗ | ✓ adult Robinson | n/a | n/a |
| 1970 | ✓ | ✓ | ✓ | ✗ | ✓ | n/a | n/a |
| 1975 | ✓ | ✓ | ✓ | ✗ | ✓ Robinson | n/a | n/a |
| 1980 | ✓ Hofferth 1981 | ✓ | ✓ | ✓ Hofferth | ✓ Nielsen | n/a | n/a |
| 1985 | ✓ Robinson | ✓ | ✓ | ✗ | ✓ Robinson | n/a | n/a |
| 1990 | ✓ | ✓ | ✓ | ✗ | ✓ | ◐ <1% adoption | n/a |
| 1995 | ✓ Bianchi/Robinson | ✓ | ✓ | ◐ Hofferth 1997 | ✓ | ✓ Pew | n/a |
| 2000 | ✓ Bianchi | ✓ | ✓ | ◐ | ✓ Nielsen | ✓ | n/a |
| 2005 | ✓ ATUS | ✓ | ✓ | ✓ ATUS | ✓ KFF 2004 | ✓ | n/a |
| 2010 | ✓ ATUS | ✓ | ✓ | ✓ ATUS | ✓ KFF 2009 | ✓ | ◐ pre-feed |
| 2015 | ✓ ATUS | ✓ | ✓ | ✓ ATUS | ✓ CSM | ✓ | ✓ CSM 2015 |
| 2020 | ✓ ATUS | ✓ | ✓ | ✓ ATUS | ✓ CSM 2021 | ✓ | ✓ CSM 2019/21 |
| 2025 | ✓ ATUS | ✓ Gallup 2023 | ✓ | ✓ ATUS | ✓ | ✓ | ✓ Pew 2024 |

✓ = direct measurement exists for that period or close to it. ◐ = partial proxy (e.g., enrollment without hours, household penetration without child viewing). ✗ = no direct measurement; chart cell is reconstructive.

**Bottom line on coverage: 16 of 27 periods have ≥ 5 channels with real data. 11 periods (all pre-1965) have ≤ 2 channels with real data.**

---

## Reconstruction with empirical E + reasonable W

Using the empirical exposure data from the CSV plus the W values from the methodology (Parents 3, Teachers 2.5, Church 2.5, Peers 2, Algorithm 1.5, TV 1, Internet 0.8), here is what reconstruction produces for the years where most data exists:

### 1965 (parent and church data available, child TV inferred from adult Robinson)

| Channel | Empirical E (hrs/wk) | Source | W | E × W | Reconstructed % | Chart % | Δ |
|---|---|---|---|---|---|---|---|
| Parents | 13 (maternal 10 + paternal ~3 est.) | Bianchi 1965 | 3 | 39 | 28% | 38 | −10 |
| School | 22 (180 days × 6.5 hr ÷ 52 wk) | NCES | 2.5 | 55 | 40% | 15 | +25 |
| Church | 1 (47% × ~2 hr/wk attenders) | Gallup 1965 | 2.5 | 2.5 | 2% | 22 | −20 |
| TV/Broadcast | 20 (adult avg, child viewing extrapolated) | Robinson 1965 | 1 | 20 | 14% | 25 | −11 |
| Peers | 12 (no 1965 data; Larson cross-sectional estimate) | Larson & Verma 1999 | 2 | 24 | 17% | 0 | +17 |
| Internet | 0 | — | 0.8 | 0 | 0% | 0 | 0 |
| Algorithm | 0 | — | 1.5 | 0 | 0% | 0 | 0 |
| **Sum** | | | | **140.5** | **100%** | **100** | |

The chart's 1965 row is **not** what straightforward E × W produces. The chart attributes 22% to Church on ~1 hr/week of attendance and 38% to Parents on ~13 hr/week of primary care — implying that church time is roughly **22× more formative per hour than parent time**. That can be argued (ritual + cultural authority + family religious practice that extends beyond service attendance), but it is a strong normative claim, not a literature finding.

### 1995 (multiple direct measurements available)

| Channel | Empirical E (hrs/wk) | Source | W | E × W | Reconstructed % | Chart % | Δ |
|---|---|---|---|---|---|---|---|
| Parents | ~30 (Hofferth 1997: 28.6 mother + ~12 father overlap-adjusted) | Hofferth & Sandberg | 3 | 90 | 38% | 25 | +13 |
| School | 22 | NCES | 2.5 | 55 | 23% | 10 | +13 |
| Church | ~1.5 (40% × ~3 hr/wk attenders) | Gallup | 2.5 | 3.75 | 2% | 12 | −10 |
| TV/Broadcast | 27 (1997 Hofferth: 27% of child weekly time = ~27 hr/wk) | Hofferth | 1 | 27 | 11% | 30 | −19 |
| Peers | 12 | Hofferth/Larson | 2 | 24 | 10% | 18 | −8 |
| Internet | 5 (early adoption, ~14% adults online) | Pew 1995 | 0.8 | 4 | 2% | 5 | −3 |
| Algorithm | 0 | — | 1.5 | 0 | 0% | 0 | 0 |
| **Sum** | | | | **203.75** | **100%** | **100** | |

Same pattern: chart over-weights Church and TV, under-weights Parents and School.

### 2020 (rich modern data)

| Channel | Empirical E (hrs/wk) | Source | W | E × W | Reconstructed % | Chart % | Δ |
|---|---|---|---|---|---|---|---|
| Parents | ~21 (ATUS 2020, ~3 hr/day) | BLS ATUS | 3 | 63 | 23% | 12 | +11 |
| School | 22 (annualized) | NCES | 2.5 | 55 | 20% | 5 | +15 |
| Church | 0.5 (29% × ~2 hr/wk attenders) | Gallup 2020 | 2.5 | 1.25 | 0.5% | 3 | −2.5 |
| TV/Broadcast | 14 (~2 hr/day traditional TV) | Nielsen/CSM 2021 | 1 | 14 | 5% | 10 | −5 |
| Peers | 12 (in-person; mediated allocated to Algo) | ATUS + modeling decision | 2 | 24 | 9% | 22 | −13 |
| Internet | 5 (non-feed: email, web, gaming) | CSM 2021 | 0.8 | 4 | 1.5% | 8 | −6.5 |
| Algorithm | 45 (CSM 2021: tween 5.5 + teen 8.5 daily, weekly) | CSM 2021 | 1.5 | 67.5 | 24% | 40 | −16 |
| **Sum** | | | | **228.75** | **83%** (rounding) | **100** | |

Even with the modern data, the chart still **over-weights Church and Algorithm and under-weights Parents and School** relative to E × W reconstruction. For the chart's 2020 cell of Algorithm = 40% to hold against ~45 hr/wk of algorithmic exposure, W_algo would need to be roughly 3× higher than what's used here.

---

## What W values would the chart actually require?

If you take the empirical E values from the CSV as fixed and back out the W needed to reproduce each chart cell, you get implied W values that look like this (rounded, relative to TV = 1):

| Channel | Implied W to match chart | Comment |
|---|---|---|
| Parents | ~2 | Below the W_parent = 3 the literature would justify |
| Teachers/School | ~0.5 | Far below what the literature would justify — school is treated as low-formative |
| Church | ~10–20 | Far above any defensible value per hour; this is the chart's load-bearing assumption |
| Peers | ~1.5 | Roughly consistent |
| TV/Broadcast | ~2 | Above what would be expected for passive broadcast |
| Internet | ~1.5 | Roughly consistent |
| Algorithm | ~3 | High but defensible given the engagement-optimization literature |

The chart, in other words, embeds a theory: that **religious ritual produces unusually high formative weight per unit of exposure** (10–20× the per-hour weight of parental contact). This is the central claim — without it, the curves don't look the way they look. It is not an empirical finding from any cited source. It is a substantive theological/cultural claim that the chart's author has built in.

---

## So can you make the chart "add up to" empirical data?

**Yes, but only if you adopt the chart's implicit theory of formation**, in which:
- Church per-hour formative weight is ~10–20× parent per-hour weight.
- School is mostly custodial, not formative.
- TV is ~2× more formative per hour than its passive nature would suggest.
- Algorithmic feeds are ~3× more formative per hour than broadcast TV.

If instead you apply weights drawn from the attachment / Bronfenbrenner / Bandura literatures (parents and school highest per-hour weight; church elevated but not exceeding parents; passive broadcast low), you get a very different picture: parents and school remain the dominant formative channels even in 2020, with algorithm rising to ~25% rather than 40%, church near 0–2% throughout the modern period rather than the chart's 10–22% pre-1980, and TV peaking around 15% rather than the chart's 35%.

---

## What I recommend you do

You have two defensible options. Either is fine; mixing them is not.

**Option A: Defend the chart as a theological-cultural model.** Be upfront that the chart is built on a specific claim that ritual religious participation produces disproportionately high per-hour formation, and that broadcast/algorithmic media produce elevated per-hour formation relative to interpersonal time. Argue for those W settings explicitly. This is a coherent position; it just needs to be stated rather than presented as data.

**Option B: Re-derive the chart from empirical E + literature-anchored W.** Accept that the resulting curves will show parents and school as more dominant historically than the chart shows, church as smaller (and declining from a smaller base) than the chart shows, and algorithm as roughly half the share the chart shows. The narrative still goes the right direction — parents/church/teachers down, algorithm up — but the magnitudes change.

Files committed:
- `Empirical_Anchors_RAW.csv` — every real measurement I could verify, with sources and NO DATA markers
- `Methodology_Childhood_Influence_1900_2025.md` — original methodology
- `Empirical_Sources_E_and_W.md` — source-by-source backing
- `Reconstruction_Chart_vs_Empirical.md` — this document

## Sources

- [Hofferth & Sandberg — Changes in American Children's Time, 1981–1997](https://www.sciencedirect.com/science/article/abs/pii/S1040260801800113)
- [Hofferth — Changes in American Children's Time, 1997–2003 (PMC)](https://pmc.ncbi.nlm.nih.gov/articles/PMC2939468/)
- [Robinson — Americans' Use of Time, 1965 (ICPSR 7254)](https://www.icpsr.umich.edu/web/ICPSR/studies/7254)
- [Robinson — Americans' Use of Time, 1985 (ICPSR 9875)](https://www.icpsr.umich.edu/web/ICPSR/studies/9875)
- [Kaiser Family Foundation — Generation M2 (2010)](https://files.eric.ed.gov/fulltext/ED527859.pdf)
- [Kaiser Family Foundation — Generation M (2005)](https://www.kff.org/wp-content/uploads/2013/01/generation-m-media-in-the-lives-of-8-18-year-olds-report.pdf)
- [Common Sense Media — 2021 Census](https://www.commonsensemedia.org/sites/default/files/research/report/8-18-census-integrated-report-final-web_0.pdf)
- [Pew — Teens, Social Media and Technology 2024](https://www.pewresearch.org/internet/2024/12/12/teens-social-media-and-technology-2024/)
- [Pew — Americans' Internet Access 2000–2015](https://www.pewresearch.org/internet/2015/06/26/americans-internet-access-2000-2015/)
- [Gallup — Four in 10 Report Attending Church in Last Week](https://news.gallup.com/poll/166613/four-report-attending-church-last-week.aspx)
- [Television Bureau of Advertising — National TV Household Penetration Trends](https://www.tvb.org/wp-content/uploads/2022/10/National-TV-Household-Penetration-Trends.pdf)
- [ICPSR — Introduction of Television to the United States Media Market, 1946–1960](https://www.icpsr.umich.edu/web/ICPSR/studies/22720)
- [Measuring Children's Media Use in the Digital Age (PMC2745155)](https://pmc.ncbi.nlm.nih.gov/articles/PMC2745155/)