def big_drop(after=10):
# 20 years of ZIRP to exhaust bonds under Prime Harvesting
for i in range(after):
yield AnnualChange(year=0, stocks=Decimal('.04'), bonds=Decimal('0'), inflation=Decimal('.02'))
# big drop of stocks to allow rebalacing to "buy stocks cheap"
yield AnnualChange(year=0, stocks=Decimal('-.25'), bonds=Decimal('0'), inflation=Decimal('.02'))
# now we just have normal (but not amazing) returns.....
while True:
yield AnnualChange(year=0, stocks=Decimal('.08'), bonds=Decimal('.04'), inflation=Decimal('.03'))
def fmt(self, row):
(stocks, bonds, inflation) = (Decimal(row[x]) / 100
for x in ("VFINX", "IT Bonds", "CPI-U"))
return AnnualChange(year=row['Year'],
stocks=stocks,
bonds=bonds,
inflation=inflation)
def fmt(self, row):
return AnnualChange(
year=row.name,
stocks=row['NIKKEI225'],
bonds=row['Spliced Bond'],
inflation=row['CPI Japan']
)
def fmt(self, row):
return AnnualChange(
year=row.name,
stocks=row['Equities'],
bonds=row['IT Bond'],
inflation=row['CPI']
)
def fmt(self, row):
# the database only provides the consumer price index level, not the
# annual change. I don't feel like going back and calculating it.
return AnnualChange(year=row.year,
stocks=Decimal(row.eq_tr),
bonds=Decimal(row.bond_tr),
inflation=Decimal(row.cpi))
def fmt(self, row):
return AnnualChange(
year=row.name.year,
stocks=Decimal(row['S&P %']),
bonds=Decimal(row['Bond %']),
inflation=0 # it is already reported in real terms
)
def random_year(self):
s = random.lognormvariate(self.stocks_mean, self.stocks_stddev) - 1
s = Decimal(s)
b = random.lognormvariate(self.bonds_mean, self.bonds_stddev) - 1
b = Decimal(b)
return AnnualChange(year=self.year, stocks=s, bonds=b, inflation=Decimal(0))
def fmt(self, row):
stock_performance = [row[x] * self.weights._asdict()[x] for x in self.asset_classes]
stock_performance = sum(stock_performance)
return AnnualChange(
year=row['Year'],
stocks=Decimal(stock_performance) / 100,
bonds=Decimal(row['IT Bonds']) / 100,
inflation=Decimal(row['CPI-U']) / 100
)
def random_year(self):
if random.randint(1, 100) < 75:
stddev = .08
else:
stddev = .30
r = random.lognormvariate(self.mean, stddev) - 1
r = Decimal(r)
return AnnualChange(year=0, stocks=r, bonds=r, inflation=Decimal(0))
def fmt(self, row):
(stocks, bonds, inflation) = (Decimal(row[x])
for x in ("Real Equity", "Real Gilt",
"Inflation"))
return AnnualChange(year=row.name.year,
stocks=stocks + inflation,
bonds=bonds + inflation,
inflation=inflation)
def fmt(self, row):
(stocks, bonds, inflation) = (Decimal(row[x]) for x in ("Real Equity", "Real Gilt", "Inflation"))
if stocks < -1:
print(row)
import pdb;pdb.set_trace()
return AnnualChange(
year=row.name.year,
stocks=stocks,
bonds=bonds,
inflation=0 # always 0 since the others are 'real'
)
def random_year(self):
y_prev = self.y_prev
# first determine bond yields based on previous year
ey = random.normalvariate(0, .009)
y_new = self.ay + (self.by * y_prev)
y_new += ey
y_new = clip(y_new, .01, .10)
delta_y = (y_new - y_prev)
# now determine total returns for cash
ec = random.normalvariate(0, .01)
rc = self.ac + (self.bc * y_new) + (self.byc * delta_y)
rc += ec
rc = clip(rc, 0, .10)
# now determine return for bonds, stocks, and inflation
def calc_returns(ai, byi, bc, by_delta_i, e_stddev, min, max):
ei = random.normalvariate(0, e_stddev)
n = (
ai +
(bc * rc) +
(byi * y_new) +
(by_delta_i * delta_y)
)
n += ei
return clip(n, min, max)
rs = {}
for k in self.coeffs:
# Does this generate real or nominal returns for the stocks & bonds?
rs[k] = calc_returns(*self.coeffs[k])
if self.logging:
self.log.loc[self.year] = (y_prev, y_new, rc, rs['stocks'], rs['bonds'], rs['inflation'])
self.y_prev = y_new
self.year += 1
return AnnualChange(
year=self.year,
stocks=Decimal(rs['stocks']),
bonds=Decimal(rs['bonds']),
inflation=Decimal(rs['inflation'])
)
def random_year(self):
y_prev = self.y_prev
ey = random.normalvariate(0, .0125)
y_new = 0.225/100 + (0.95 * y_prev) + ey
y_new = clip(y_new, .01, .10)
ebond = random.normalvariate(0, .015)
r_bonds = 0 + (1.0 * y_new) + (-8.0 * (y_new - y_prev)) + ebond
# Blanchett doesn't provide min/max for r_bonds
# I'll reuse the min/max from LowYieldsAutoRegression
r_bonds = clip(r_bonds, -.15, .40)
ebill = random.normalvariate(0, .01)
r_bill = -.02 + y_new + 0.75 * (y_new - y_prev) + ebill
r_bill = clip(r_bill, .01, .10)
# the tables max out at 15 years...
lookup_year = min(15, self.year)
estocks = random.normalvariate(0, self.estocks_stdev[lookup_year])
r_stocks = self.a_cape[lookup_year] + self.b_cape[lookup_year] * self.initial_cape + estocks
# Blanchett doesn't provide min/max for r_stocks
# I'll reuse the min/max from LowYieldsAutoRegression
r_stocks = clip(r_stocks, -1, 2)
einflation = random.normalvariate(0, .01)
r_inflation = 0.0125 + (y_new - y_prev) + 0.5 * r_bill + einflation
r_inflation = clip(r_inflation, -.05, .10)
self.y_prev = y_new
self.year += 1
return AnnualChange(
year=self.year,
stocks=Decimal(r_stocks),
bonds=Decimal(r_bonds),
inflation=Decimal(r_inflation)
)
def constant_returns(stocks=Decimal('.04'),
bonds=Decimal('.02'),
inflation=Decimal('.02')):
return itertools.repeat(
AnnualChange(year=0, stocks=stocks, bonds=bonds, inflation=inflation))
def random_year(self):
r = random.lognormvariate(self.mean, self.stddev) - 1
r = Decimal(r)
return AnnualChange(year=self.year, stocks=r, bonds=r, inflation=Decimal(0))
def random_year(self):
''' This is the same method name as in market.US_1871_Returns...which
allows this to be a drop-in replacement for that when simulating data '''
r = lognormal(self.mean, self.sigma) - 1
r = Decimal(r)
return AnnualChange(year=0, stocks=r, bonds=r, inflation=Decimal(0))