"""
Chart 2: Finance & insurance value added as % of US GDP, 1997-2025

Source: BEA Industry Economic Accounts, Table 1.10 (TVA110-A annual)
URL: https://apps.bea.gov/industry/Release/XLS/GDPxInd/ValueAdded.xlsx
Published April 9, 2026. Annual frequency. NAICS-based.

Key story: NO COMPRESSION. Finance and insurance share of GDP went from
6.7% (1997) to 7.9% (2025), oscillating between a 5.9% GFC trough (2008)
and an 8.0% peak (2020). Every prior tech wave was supposed to compress this.
It didn't.

Data gap: pre-1997 data uses SIC industry codes (not NAICS) and lives in
BEA historical archives. The Greenwood/Ialenti/Scharfstein 2025 Annual
Review paper and Philippon's 2015 AER paper provide the longer-run series
back to 1880. We don't reproduce their data here — we cite it in caption.
"""

from pathlib import Path
import csv
import matplotlib.pyplot as plt
import matplotlib.ticker as mtick

HERE = Path(__file__).parent
OUT = HERE / "chart_02_finance_share_gdp.png"

years, vals = [], []
with open(HERE / "data" / "bea_finance_share_gdp_1997_2025.csv") as f:
    r = csv.DictReader(f)
    for row in r:
        years.append(int(row["year"]))
        vals.append(float(row["finance_insurance_pct_gdp"]))

fig, ax = plt.subplots(figsize=(10, 5.5), dpi=200)
ax.plot(years, vals, color="#1f4e79", linewidth=2.0, marker="o", markersize=4)
ax.fill_between(years, vals, alpha=0.12, color="#1f4e79")

# Reference: Philippon "constant unit cost of intermediation" implies stable ~2% of GDP for the
# *unit cost* metric — different from value-added, but worth annotating as the related claim.
# The relevant horizontal here is the 1997 baseline of 6.7%.
ax.axhline(6.7, color="#888", linestyle="--", linewidth=0.8, alpha=0.6)
ax.text(1997.5, 6.55, "1997 baseline: 6.7%", fontsize=8, color="#666", va="top")

# Annotate key points
ax.annotate(
    f"GFC trough\n2008: {vals[years.index(2008)]:.1f}%",
    xy=(2008, vals[years.index(2008)]),
    xytext=(2003.5, 5.4),
    fontsize=8,
    arrowprops=dict(arrowstyle="->", color="#666", lw=0.6),
)

peak_idx = vals.index(max(vals))
ax.annotate(
    f"Peak\n{years[peak_idx]}: {vals[peak_idx]:.1f}%",
    xy=(years[peak_idx], vals[peak_idx]),
    xytext=(years[peak_idx] - 4, vals[peak_idx] + 0.5),
    fontsize=8,
    arrowprops=dict(arrowstyle="->", color="#666", lw=0.6),
)

ax.annotate(
    f"2025: {vals[-1]:.1f}%",
    xy=(years[-1], vals[-1]),
    xytext=(years[-1] - 5, vals[-1] - 0.5),
    fontsize=8,
    arrowprops=dict(arrowstyle="->", color="#666", lw=0.6),
)

ax.set_title(
    "Finance & insurance value added as % of US GDP, 1997–2025",
    fontsize=13,
    fontweight="bold",
    pad=14,
)
ax.set_xlabel("Year", fontsize=10)
ax.set_ylabel("Finance & insurance value added / GDP", fontsize=10)
ax.yaxis.set_major_formatter(mtick.PercentFormatter(decimals=1))
ax.set_ylim(5.0, 9.0)
ax.set_xlim(1996, 2026)
ax.grid(alpha=0.25)
ax.spines["top"].set_visible(False)
ax.spines["right"].set_visible(False)

caption = (
    "Source: BEA Industry Economic Accounts, Table 1.10 (Value Added as % of GDP), April 2026. NAICS-based.\n"
    "Pre-1997 series exists in Philippon (AER 2015) and Greenwood/Ialenti/Scharfstein (Annual Review of\n"
    "Financial Economics 2025); these show finance share rose from ~2.8% in 1950 to current levels."
)
fig.text(0.02, 0.005, caption, fontsize=7, color="#555")

plt.tight_layout(rect=[0, 0.045, 1, 1])
plt.savefig(OUT, dpi=200, bbox_inches="tight")
print(f"wrote {OUT}")
print(f"data span: {years[0]}–{years[-1]} ({len(years)} years)")
print(f"min: {min(vals)}% in {years[vals.index(min(vals))]}")
print(f"max: {max(vals)}% in {years[vals.index(max(vals))]}")
print(f"latest: {vals[-1]}% ({years[-1]})")