Novo
Nordisk analyse

# ============================================================
# Novo Nordisk-analyse
# Formål:
#   Undersøge om Novo Nordisks kursudvikling kan forklares af
#   generelle markedsbevægelser, eller om den primært drives af
#   selskabsspecifikke events, regimeskift og konkurrence mod Eli Lilly.
# ============================================================

from pathlib import Path

import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
import matplotlib.ticker as mticker
import matplotlib.dates as mdates
import yfinance as yf
from scipy import stats

# ------------------------------------------------------------
# 1) Opsætning
# ------------------------------------------------------------
Path("assets").mkdir(exist_ok=True)

novo = "NOVO-B.CO"       # Novo Nordisk B, København
sp500 = "^GSPC"          # S&P 500
lilly = "LLY"            # Eli Lilly

START = "2021-07-06"
SLUT = "2026-06-09"      # fastlåst, så analysen ikke ændrer sig automatisk

novo_peak_date = pd.to_datetime("2024-06-26")

novo_events = {
    "SELECT trial": "2023-08-08",
    "CagriSema skuffer": "2024-12-20",
    "Nedjustering": "2025-07-29",
    "Varslet salgsfald": "2026-02-03",
    "Novo bund": "2026-03-03",
}

sp500_events = {
    "Liberation Day": "2025-04-02",
}

# ------------------------------------------------------------
# 2) Data
# ------------------------------------------------------------
def hent_close(ticker: str) -> pd.Series:
    """Henter autojusteret slutkurs fra Yahoo Finance."""
    data = yf.download(
        ticker,
        start=START,
        end=SLUT,
        auto_adjust=True,
        progress=False,
    )

    if data.empty:
        raise ValueError(f"Ingen data hentet for {ticker}")

    return data["Close"].squeeze().dropna()


pris_novo = hent_close(novo)
pris_sp500 = hent_close(sp500)
pris_lilly = hent_close(lilly)

# ------------------------------------------------------------
# 3) Indekseret kursudvikling
# ------------------------------------------------------------
common_start = max(
    pris_novo.index[0],
    pris_sp500.index[0],
    pris_lilly.index[0],
)

idx_novo = (pris_novo / pris_novo.loc[common_start]) * 100
idx_sp500 = (pris_sp500 / pris_sp500.loc[common_start]) * 100
idx_lilly = (pris_lilly / pris_lilly.loc[common_start]) * 100

fig, ax = plt.subplots(figsize=(13, 6))

ax.plot(idx_novo, linewidth=2.2, label="Novo Nordisk")
ax.plot(idx_lilly, linewidth=2.2, label="Eli Lilly")
ax.plot(idx_sp500, linewidth=1.8, linestyle="--", label="S&P 500")

ax.axhline(100, linewidth=0.9, alpha=0.45)
ax.axvline(novo_peak_date, linestyle=":", linewidth=1.8, alpha=0.75, label="Novo peak")

for series, name in [(idx_novo, "Novo"), (idx_lilly, "Lilly"), (idx_sp500, "S&P 500")]:
    last_val = series.dropna().iloc[-1]
    last_date = series.dropna().index[-1]
    ax.annotate(
        f"{name}: {last_val:.0f}",
        xy=(last_date, last_val),
        xytext=(8, 0),
        textcoords="offset points",
        fontsize=9,
        va="center",
    )

ax.set_title("Indekseret kursudvikling: Novo Nordisk, Eli Lilly og S&P 500")
ax.set_xlabel("Dato")
ax.set_ylabel("Indekseret kurs, start = 100")
ax.xaxis.set_major_formatter(mdates.DateFormatter("%b %Y"))
ax.xaxis.set_major_locator(mdates.MonthLocator(bymonth=[1, 7]))
ax.grid(True, alpha=0.3)
ax.legend()
fig.autofmt_xdate(rotation=35)
plt.tight_layout()
plt.savefig("assets/01_indekseret_kursudvikling.png", dpi=160, bbox_inches="tight")
plt.show()

# ------------------------------------------------------------
# 4) Afkast
# ------------------------------------------------------------
returns_novo = pris_novo.pct_change()
returns_sp500 = pris_sp500.pct_change()
returns_lilly = pris_lilly.pct_change()

aligned = pd.concat(
    [returns_novo, returns_sp500, returns_lilly],
    axis=1,
    sort=False,
).dropna()

aligned.columns = ["novo", "sp500", "lilly"]

fig, axes = plt.subplots(1, 2, figsize=(16, 6))

axes[0].plot(aligned["novo"], linewidth=1)
axes[0].axvline(novo_peak_date, linestyle="--", linewidth=1.5, alpha=0.8)
axes[0].set_title("Novo Nordisk daglige afkast")
axes[0].set_xlabel("Dato")
axes[0].set_ylabel("Dagligt afkast")
axes[0].yaxis.set_major_formatter(mticker.PercentFormatter(1.0))
axes[0].grid(True, alpha=0.3)

axes[1].plot(aligned["sp500"], linewidth=1)
axes[1].axvline(novo_peak_date, linestyle="--", linewidth=1.5, alpha=0.8)
axes[1].set_title("S&P 500 daglige afkast")
axes[1].set_xlabel("Dato")
axes[1].set_ylabel("Dagligt afkast")
axes[1].yaxis.set_major_formatter(mticker.PercentFormatter(1.0))
axes[1].grid(True, alpha=0.3)

plt.suptitle("Daglige procentvise afkast")
plt.tight_layout()
plt.savefig("assets/02_daglige_afkast.png", dpi=160, bbox_inches="tight")
plt.show()

# ------------------------------------------------------------
# 5) Regression
# ------------------------------------------------------------
def run_regression(y: pd.Series, x: pd.Series) -> dict:
    """Returnerer alpha, beta, R², p-værdi, standardfejl og antal observationer."""
    regression_data = pd.concat([y, x], axis=1).dropna()
    regression_data.columns = ["y", "x"]

    beta, alpha, r_value, p_value, std_err = stats.linregress(
        regression_data["x"],
        regression_data["y"],
    )

    return {
        "alpha": alpha,
        "beta": beta,
        "r_squared": r_value**2,
        "p_value": p_value,
        "std_error": std_err,
        "n_obs": len(regression_data),
    }


reg_full = run_regression(aligned["novo"], aligned["sp500"])

print("── Full-period regression: Novo vs. S&P 500 ───────────────")
print(f"Beta:       {reg_full['beta']:.4f}")
print(f"Alpha:      {reg_full['alpha']:.6f}")
print(f"R²:         {reg_full['r_squared']:.4f}")
print(f"P-værdi:    {reg_full['p_value']:.6f}")
print(f"Std. error: {reg_full['std_error']:.6f}")
print(f"Obs.:       {reg_full['n_obs']}")

# ------------------------------------------------------------
# 6) Split-period regression
# ------------------------------------------------------------
pre_peak = aligned.loc[aligned.index < novo_peak_date]
post_peak = aligned.loc[aligned.index >= novo_peak_date]

reg_pre = run_regression(pre_peak["novo"], pre_peak["sp500"])
reg_post = run_regression(post_peak["novo"], post_peak["sp500"])

split_regression_results = pd.DataFrame(
    [reg_full, reg_pre, reg_post],
    index=["Hele perioden", "Før peak", "Efter peak"],
)

print("\n── Split-period regression: Novo vs. S&P 500 ───────────────")
print(split_regression_results)

fig, ax = plt.subplots(figsize=(9, 5))
split_regression_results["beta"].plot(kind="bar", ax=ax)
ax.axhline(0, linestyle="--", linewidth=1, alpha=0.7)
ax.set_title("Beta før og efter Novo-peak")
ax.set_xlabel("Periode")
ax.set_ylabel("Beta mod S&P 500")
ax.grid(True, axis="y", alpha=0.3)
plt.xticks(rotation=0)
plt.tight_layout()
plt.savefig("assets/03_split_period_beta.png", dpi=160, bbox_inches="tight")
plt.show()

# ------------------------------------------------------------
# 7) Rolling beta
# ------------------------------------------------------------
rolling_window = 252
rolling_cov = aligned["novo"].rolling(rolling_window).cov(aligned["sp500"])
rolling_var = aligned["sp500"].rolling(rolling_window).var()
rolling_beta_12m = rolling_cov / rolling_var

fig, ax = plt.subplots(figsize=(14, 6))
ax.plot(rolling_beta_12m, linewidth=2, label="12-måneders rolling beta")
ax.axhline(0, linestyle="--", linewidth=1, alpha=0.6)
ax.axvline(novo_peak_date, linestyle="--", linewidth=1.5, alpha=0.8, label="Novo peak")
ax.set_title("Novo Nordisk: 12-måneders rolling beta mod S&P 500")
ax.set_xlabel("Dato")
ax.set_ylabel("Beta")
ax.grid(True, alpha=0.3)
ax.legend()
plt.tight_layout()
plt.savefig("assets/04_rolling_beta.png", dpi=160, bbox_inches="tight")
plt.show()

# ------------------------------------------------------------
# 8) Rolling volatilitet
# ------------------------------------------------------------
rolling_vol_novo_annual = aligned["novo"].rolling(window=30).std() * np.sqrt(252)
rolling_vol_sp500_annual = aligned["sp500"].rolling(window=30).std() * np.sqrt(252)

fig, ax = plt.subplots(figsize=(14, 6))
ax.plot(rolling_vol_novo_annual, linewidth=2, label="Novo Nordisk")
ax.plot(rolling_vol_sp500_annual, linewidth=2, linestyle="--", label="S&P 500")

for label, date in novo_events.items():
    event_date = pd.to_datetime(date)
    ax.axvline(event_date, linestyle=":", linewidth=1.1, alpha=0.65)
    ax.annotate(
        label,
        xy=(event_date, 0.05),
        xycoords=("data", "axes fraction"),
        xytext=(5, 0),
        textcoords="offset points",
        fontsize=8,
        va="bottom",
    )

for label, date in sp500_events.items():
    event_date = pd.to_datetime(date)
    ax.axvline(event_date, linestyle="--", linewidth=1.2, alpha=0.65)
    ax.annotate(
        label,
        xy=(event_date, 0.90),
        xycoords=("data", "axes fraction"),
        xytext=(5, 0),
        textcoords="offset points",
        fontsize=8,
        va="bottom",
    )

ax.set_title("30-dages rolling annualiseret volatilitet")
ax.set_xlabel("Dato")
ax.set_ylabel("Annualiseret volatilitet")
ax.yaxis.set_major_formatter(mticker.PercentFormatter(1.0))
ax.grid(True, alpha=0.3)
ax.legend()
plt.tight_layout()
plt.savefig("assets/05_rolling_volatilitet.png", dpi=160, bbox_inches="tight")
plt.show()

# ------------------------------------------------------------
# 9) Novo vs. Eli Lilly
# ------------------------------------------------------------
novo_lilly_corr = aligned["novo"].corr(aligned["lilly"])
reg_novo_lilly = run_regression(y=aligned["novo"], x=aligned["lilly"])

print("\n── Novo vs. Eli Lilly ─────────────────────────────────────")
print(f"Korrelation:    {novo_lilly_corr:.4f}")
print(f"Beta mod Lilly: {reg_novo_lilly['beta']:.4f}")
print(f"Alpha:          {reg_novo_lilly['alpha']:.6f}")
print(f"R²:             {reg_novo_lilly['r_squared']:.4f}")
print(f"P-værdi:        {reg_novo_lilly['p_value']:.6f}")
print(f"Std. error:     {reg_novo_lilly['std_error']:.6f}")
print(f"Obs.:           {reg_novo_lilly['n_obs']}")

fig, ax = plt.subplots(figsize=(8, 6))
ax.scatter(aligned["lilly"], aligned["novo"], alpha=0.35, s=18)
x_vals = np.linspace(aligned["lilly"].min(), aligned["lilly"].max(), 100)
y_vals = reg_novo_lilly["alpha"] + reg_novo_lilly["beta"] * x_vals
ax.plot(x_vals, y_vals, linewidth=2, label="Regressionslinje")
ax.set_title("Daglige afkast: Novo Nordisk vs. Eli Lilly")
ax.set_xlabel("Eli Lilly dagligt afkast")
ax.set_ylabel("Novo Nordisk dagligt afkast")
ax.xaxis.set_major_formatter(mticker.PercentFormatter(1.0))
ax.yaxis.set_major_formatter(mticker.PercentFormatter(1.0))
ax.grid(True, alpha=0.3)
ax.legend()
plt.tight_layout()
plt.savefig("assets/06_novo_vs_lilly_regression.png", dpi=160, bbox_inches="tight")
plt.show()

# ------------------------------------------------------------
# 10) Event study
# ------------------------------------------------------------
def event_study(
    returns_data: pd.DataFrame,
    event_dict: dict,
    estimation_window: int = 252,
    event_window_before: int = 1,
    event_window_after: int = 3,
) -> pd.DataFrame:
    """Beregner kumulativ abnormal return omkring events."""
    results = []

    for event_name, event_date_str in event_dict.items():
        event_date = pd.to_datetime(event_date_str)
        valid_dates = returns_data.index[returns_data.index >= event_date]

        if len(valid_dates) == 0:
            continue

        trading_event_date = valid_dates[0]
        event_pos = returns_data.index.get_loc(trading_event_date)

        estimation_start_pos = event_pos - estimation_window - event_window_before
        estimation_end_pos = event_pos - event_window_before
        event_start_pos = event_pos - event_window_before
        event_end_pos = event_pos + event_window_after

        if estimation_start_pos < 0 or event_end_pos >= len(returns_data):
            continue

        estimation_data = returns_data.iloc[estimation_start_pos:estimation_end_pos]
        event_data = returns_data.iloc[event_start_pos:event_end_pos + 1]

        reg_event = run_regression(
            y=estimation_data["novo"],
            x=estimation_data["sp500"],
        )

        expected_return = reg_event["alpha"] + reg_event["beta"] * event_data["sp500"]
        abnormal_return = event_data["novo"] - expected_return

        results.append({
            "event": event_name,
            "dato": event_date.date(),
            "handelsdato": trading_event_date.date(),
            "beta_estimat": reg_event["beta"],
            "alpha_estimat": reg_event["alpha"],
            "faktisk_afkast_window": (1 + event_data["novo"]).prod() - 1,
            "sp500_afkast_window": (1 + event_data["sp500"]).prod() - 1,
            "abnormal_return_window": abnormal_return.sum(),
            "event_window": f"[-{event_window_before}, +{event_window_after}]",
        })

    return pd.DataFrame(results)


event_results = event_study(
    returns_data=aligned[["novo", "sp500"]],
    event_dict=novo_events,
    estimation_window=252,
    event_window_before=1,
    event_window_after=3,
)

print("\n── Event study: abnormal returns ───────────────────────────")
print(event_results)

fig, ax = plt.subplots(figsize=(12, 6))
ax.bar(event_results["event"], event_results["abnormal_return_window"])
ax.axhline(0, linestyle="--", linewidth=1, alpha=0.7)
ax.set_title("Kumulativ abnormal return omkring Novo-begivenheder")
ax.set_xlabel("Begivenhed")
ax.set_ylabel("Kumulativ abnormal return")
ax.yaxis.set_major_formatter(mticker.PercentFormatter(1.0))
ax.grid(True, axis="y", alpha=0.3)
plt.xticks(rotation=25, ha="right")
plt.tight_layout()
plt.savefig("assets/07_event_study_abnormal_returns.png", dpi=160, bbox_inches="tight")
plt.show()

# ------------------------------------------------------------
# 11) Drawdown
# ------------------------------------------------------------
def calculate_drawdown(return_series: pd.Series) -> tuple[pd.Series, float]:
    cumulative_return = (1 + return_series.dropna()).cumprod()
    rolling_peak = cumulative_return.cummax()
    drawdown = (cumulative_return - rolling_peak) / rolling_peak
    return drawdown, drawdown.min()


drawdown_novo, max_dd_novo = calculate_drawdown(aligned["novo"])
drawdown_sp500, max_dd_sp500 = calculate_drawdown(aligned["sp500"])
drawdown_lilly, max_dd_lilly = calculate_drawdown(aligned["lilly"])

drawdown_summary = pd.DataFrame({
    "Max drawdown": [max_dd_novo, max_dd_lilly, max_dd_sp500],
}, index=["Novo Nordisk", "Eli Lilly", "S&P 500"])

print("\n── Komparativ max drawdown ────────────────────────────────")
print(drawdown_summary)

fig, ax = plt.subplots(figsize=(14, 6))
ax.plot(drawdown_novo, linewidth=2, label="Novo Nordisk")
ax.plot(drawdown_lilly, linewidth=2, label="Eli Lilly")
ax.plot(drawdown_sp500, linewidth=2, linestyle="--", label="S&P 500")
ax.axvline(novo_peak_date, linestyle="--", linewidth=1.5, alpha=0.8, label="Novo peak")
ax.set_title("Komparativ drawdown: Novo Nordisk, Eli Lilly og S&P 500")
ax.set_xlabel("Dato")
ax.set_ylabel("Drawdown")
ax.yaxis.set_major_formatter(mticker.PercentFormatter(1.0))
ax.grid(True, alpha=0.3)
ax.legend()
plt.tight_layout()
plt.savefig("assets/08_komparativ_drawdown.png", dpi=160, bbox_inches="tight")
plt.show()

# ------------------------------------------------------------
# 12) Eksport
# ------------------------------------------------------------
split_regression_results.to_csv("assets/split_regression_results.csv")
event_results.to_csv("assets/event_study_results.csv", index=False)
drawdown_summary.to_csv("assets/drawdown_summary.csv")

print("\nFigurer og tabeller er gemt i assets/-mappen.")