Technology Adoption

Adoption.py

#!/usr/bin/env python3

import requests
import pandas as pd
import numpy as np
from io import StringIO

URL = "https://ourworldindata.org/grapher/technology-adoption-by-households-in-the-united-states.csv"

print("Downloading dataset...")
resp = requests.get(URL)
resp.raise_for_status()

df = pd.read_csv(StringIO(resp.text))

# convert to long if needed
if "Entity" not in df.columns:
	df = df.melt(id_vars=["Year"], var_name="Entity", value_name="Value")

df = df.rename(columns={
	"Technology Diffusion (Comin and Hobijn (2004) and others)": "Value"
})

results = []

for entity, sub in df.groupby("Entity"):
	sub = sub.dropna().sort_values("Year")

	if len(sub) < 5:
		continue

	t = sub["Year"].values
	y = sub["Value"].values

	# normalize (optional)
	if np.max(y) < 20:
		continue

	# compute slope (dy/dt)
	dy = np.gradient(y, t)

	# smooth a bit to avoid noise spikes
	dy_smooth = pd.Series(dy).rolling(3, center=True).mean().to_numpy()

	if np.all(np.isnan(dy_smooth)):
		continue

	idx = np.nanargmax(dy_smooth)
	midpoint_year = t[idx]
	slope = dy_smooth[idx]

	results.append((entity, midpoint_year, slope))

results.sort(key=lambda x: x[1])

print("\n=== MIDPOINT YEARS (max slope) ===\n")

for entity, year, slope in results:
	print(f"{entity:35s}  midpoint ≈ {year:.0f}   (slope={slope:.2f})")

print("\nDone.")