Evaluate index performance

Manipulating Time Series Data in Python

Stefan Jansen

Founder & Lead Data Scientist at Applied Artificial Intelligence

Evaluate your value-weighted index

Index return:
- Total index return
- Contribution by component
Performance vs Benchmark
- Total period return
- Rolling returns for sub periods

Value-based index - recap

agg_market_cap = market_cap_series.sum(axis=1)

index = agg_market_cap.div(agg_market_cap.iloc[0]).mul(100)

index.plot(title='Market-Cap Weighted Index')

Value contribution by stock

agg_market_cap.iloc[-1] - agg_market_cap.iloc[0]

315,037.71

Value contribution by stock

change = market_cap_series.first('D').append(market_cap_series.last('D'))

change.diff().iloc[-1].sort_values() # or: .loc['2016-12-30']

TM     -6,365.07
KO     -4,034.49
ABB     7,592.41
ORCL   11,109.65
PG     14,597.48
UPS    17,212.08
WMT    23,232.85
BABA   27,800.00
JNJ    39,931.44
T      50,229.33
XOM    53,075.38
JPM    80,656.65
Name: 2016-12-30 00:00:00, dtype: float64

Market-cap based weights

market_cap = components['Market Capitalization']

weights = market_cap.div(market_cap.sum())

weights.sort_values().mul(100)

Stock Symbol
ABB     1.85
UPS     3.45
TM      5.96
ORCL    6.93
KO      7.03
WMT     8.50
PG      8.81
T       9.47
BABA   10.55
JPM    11.50
XOM    12.97
JNJ    12.97
Name: Market Capitalization, dtype: float64

Value-weighted component returns

index_return = (index.iloc[-1] / index.iloc[0] - 1) * 100

14.06

weighted_returns = weights.mul(index_return)

weighted_returns.sort_values().plot(kind='barh')

Performance vs benchmark

data = index.to_frame('Index') # Convert pd.Series to pd.DataFrame

data['SP500'] = pd.read_csv('sp500.csv', parse_dates=['Date'],
                            index_col='Date')

data.SP500 = data.SP500.div(data.SP500.iloc[0], axis=0).mul(100)

Performance vs benchmark: 30D rolling return

def multi_period_return(r):
    return (np.prod(r + 1) - 1) * 100

data.pct_change().rolling('30D').apply(multi_period_return).plot()

Let's practice!

Manipulating Time Series Data in Python