def strategy_capm(returns, start_date):
"""
CAPM using lagged optimal market portfolio tracker to predict
company stock return. Positions are assigned as follows:
- Buy when predicted return > expected return
- Sell when predicted return < expected return
- No transaction costs. For now.
"""
mkt_return = pdr.get_data_famafrench(
'F-F_Research_Data_Factors_daily',
start='1-1-2010')[0]['Mkt-RF'] / 100
rf = pdr.get_data_famafrench('F-F_Research_Data_Factors_daily',
start='1-1-2010',
end=start_date)[0]['RF']
ols_capm = {
'slope': None,
'intercept': None,
'r_value': None,
'p_value': None,
'std_err': None
}
returns = returns.shift(1)
for ret in returns.columns:
slope, intercept, r_value, p_value, std_err = linregress(
mkt_return, y=returns[ret])
ols_capm['slope'].add(slope)
ols_capm['intercept'].add(intercept)
ols_capm['r_value'].add(r_value)
ols_capm['p_value'].add(p_value)
ols_capm['std_err'].add(std_err)
return ols_capm
def load_industries():
"""Load industry portfolio returns from Ken French's website.
Returns
-------
industries : dictionary of Pandas DataFrames
Each key is a portfolio group.
Example
-------
>>> from pyfinance import datasets
>>> ind = datasets.load_industries()
# Monthly returns to 5 industry portfolios
>>> ind[5].head()
Cnsmr Manuf HiTec Hlth Other
Date
1950-01-31 1.26 1.47 3.21 1.06 3.19
1950-02-28 1.91 1.29 2.06 1.92 1.02
1950-03-31 0.28 1.93 3.46 -2.90 -0.68
1950-04-30 3.22 5.21 3.58 5.52 1.50
1950-05-31 3.81 6.18 1.07 3.96 1.36
"""
n = [5, 10, 12, 17, 30, 38, 48]
port = ("%s_Industry_Portfolios" % i for i in n)
rets = []
for p in port:
ret = pdr.get_data_famafrench(p, start=DSTART)[0]
rets.append(ret.to_timestamp(how="end", copy=False))
industries = dict(zip(n, rets))
return industries
import statsmodels.api as sm
from statsmodels.regression.rolling import RollingOLS
seaborn.set_style("darkgrid")
pd.plotting.register_matplotlib_converters()
# `pandas-datareader` is used to download data from
# [Ken French's website](https://mba.tuck.dartmouth.edu/pages/faculty/ken.
# french/data_library.html).
# The two data sets downloaded are the 3 Fama-French factors and the 10
# industry portfolios.
# Data is available from 1926.
#
# The data are monthly returns for the factors or industry portfolios.
factors = pdr.get_data_famafrench("F-F_Research_Data_Factors",
start="1-1-1926")[0]
factors.head()
industries = pdr.get_data_famafrench("10_Industry_Portfolios",
start="1-1-1926")[0]
industries.head()
# The first model estimated is a rolling version of the CAPM that
# regresses
# the excess return of Technology sector firms on the excess return of the
# market.
#
# The window is 60 months, and so results are available after the first 60
# (`window`)
# months. The first 59 (`window - 1`) estimates are all `nan` filled.
linestyle='--',
linewidth=1.0)
# compare the return distribution of 3 firms visually...
grid = sns.pairplot(daily_pct_change, diag_kind='kde', kind="reg")
grid.map_offdiag(plot_unity) # how cool is that!
###############################################################################
# get the factor loadings
###############################################################################
# get FF factors merged with the stock returns
ff = pdr.get_data_famafrench(
'F-F_Research_Data_5_Factors_2x3_daily', start=2006
)[0] # the [0] is because the imported obect is a dictionary, and key=0 is the dataframe
ff.rename(columns={"Mkt-RF": "mkt_excess"}, inplace=True) # cleaner name
ff = ff.join(daily_pct_change, how='inner') # merge with stock returns
for stock in ['MSFT', 'AAPL', 'VZ']:
ff[stock] = ff[
stock] * 100 # FF store variables as percents, so convert to that
ff[stock + '_excess'] = ff[stock] - ff[
'RF'] # convert to excess returns in prep for regressions
print(ff.describe()) # ugly...
pd.set_option('display.float_format',
lambda x: '%.2f' % x) # show fewer digits
pd.options.display.max_columns = ff.shape[1] # show more columns
print(ff.describe(include='all')) # better!
# run the models-
import pandas_datareader as pdr
import pandas as pd
import statsmodels.api as sm
from statsmodels.regression.rolling import RollingOLS
from statsmodels.api import OLS, add_constant
import matplotlib.pyplot as plt
import seaborn
seaborn.set_style('darkgrid')
pd.plotting.register_matplotlib_converters()
#%%
factors = pdr.get_data_famafrench('F-F_Research_Data_Factors', start='1-1-2020')[0]
# print(factors.head())
industries = pdr.get_data_famafrench('17_Industry_Portfolios', start='1-1-2020')[0]
# print(industries.head())
#%%
exog_vars = ['Mkt-RF', 'SMB', 'HML','RF']
exog = sm.add_constant(factors[exog_vars])
returns = industries.sub(factors.RF, axis=0)
#%%
betas = []
rsquared = []
#%%
for industry in returns.columns:
# print(returns.loc[returns.index, industry])
endog = returns.loc[returns.index, industry]
rols = RollingOLS(endog, exog, window=12)
rres = rols.fit()