Original paper
Abstract
This paper has a twofold objective. First, we contribute to the stream of literature that investigates whether traditional asset pricing factors show any predictive power for the cross-section of Real Estate Investment Trust (REIT) returns. In particular, we investigate the existence of a premium associated to the Value, Size, Momentum, Investment, and Profitability factors over the period 1993-2018. We find support for all the pricing factors but for the Profitability one. Second, we investigate whether a set of smart beta strategies, based on the combination of the identified factors, may outperform similar allocation techniques that do not exploit factors. We find that all the proposed factor-based strategies display a higher risk-adjusted out-of-sample performance than a simple buy-and-hold investment in the real estate market (proxied by the FTSE NAREIT All REITs Index). In addition, we find that when factor-based strategies are implemented, REIT-only portfolios display risk-adjusted performances comparable to those of diversified portfolios that include equity, bond, and commodities.
Keywords:Â REITs, real estate factors, factor investing, smart beta strategies
Trading rules
- The selection scope includes all REITs listed on NYSE, AMEX, and NASDAQ.
- Remove REITs that have negative B/M values as well as those in the extreme 0.5% at both ends (highest and lowest).
- Annually, rank REITs based on their book-to-market ratios.
- The cheap (value) portfolio consists of REITs in the top 20% quintile of the B/M ratio (BM1).
- The expensive (growth) portfolio consists of REITs in the lowest 20% quintile of the B/M ratio (BM5).
- The strategy involves taking a long position on cheap REITs.
- Hold the selected REITs for a duration of one year.
- Each REIT in the collection is given equal weight.
Python code
Backtrader
import backtrader as bt
import pandas as pd
class REITsStrategy(bt.Strategy):
def __init__(self):
# Define universe
self.universe = pd.read_csv('nyse_amex_nasdaq.csv')['Symbol'].tolist()
# Quarter percentiles to exclude
self.bm_exclusion_percentiles = (0.005, 0.995)
# Time parameters
self.year_start = self.datas[0].datetime.date(0).year
self.rebalance_year = self.year
# Set length of holding period
self.holding_period = 252 # trading days in a year
# Keep track of cheap and expensive REITs
self.expensive_reits = []
self.cheap_reits = []
def start_year(self):
# Update current year
self.year = self.datas[0].datetime.date(0).year
# Update list of investment universe
self.universe = pd.read_csv('nyse_amex_nasdaq.csv')['Symbol'].tolist()
# Get fundamental data from Yahoo Finance
self.bm_ratio = self.get_bm_ratio()
# Exclude extreme B/M values
self.exclude_extreme_bms()
# Rank the remaining REITs based on their B/M ratio
self.rank_reits()
# Assign REITs to expensive and cheap portfolios
self.assign_portfolios()
# Calculate weights for each REIT in the portfolio
self.calculate_weights()
# Go long on cheap REITs
for data in self.cheap_reits:
self.order_target_percent(data, target=self.weights[data])
def next(self):
# Skip if not at the end of the year
if self.datas[0].datetime.date(0).year == self.rebalance_year:
return
# Set the new rebalancing year
self.rebalance_year = self.datas[0].datetime.date(0).year
# Close all existing positions
self.close()
# Assign REITs to expensive and cheap portfolios
self.assign_portfolios()
# Calculate weights for each REIT in the portfolio
self.calculate_weights()
# Go long on cheap REITs
for data in self.cheap_reits:
self.order_target_percent(data, target=self.weights[data])
def get_bm_ratio(self):
bm_ratio = {}
for sym in self.universe:
data = bt.feeds.YahooFinanceData(dataname=sym, fromdate=datetime(self.year, 1, 1),
todate=datetime(self.year, 12, 31),
buffered= True,
reverse= False)
df = data.to_pandas()
bm_ratio[sym] = df['Book Value'][0]/df['Market Capitalization'][0]
return bm_ratio
def exclude_extreme_bms(self):
sorted_bms = sorted(self.bm_ratio.items(), key=lambda x:x[1])
num_excluded = int(len(sorted_bms) * 0.005)
self.universe = [x[0] for x in sorted_bms[num_excluded:len(sorted_bms)-num_excluded]]
def rank_reits(self):
sorted_bms = sorted(self.bm_ratio.items(), key=lambda x:x[1], reverse=True)
self.sorted_reits = [x[0] for x in sorted_bms]
return self.sorted_reits
def assign_portfolios(self):
bm1_num = int(len(self.sorted_reits) * 0.2)
bm5_num = int(len(self.sorted_reits) * 0.8)
self.expensive_reits = self.sorted_reits[len(self.sorted_reits)-bm1_num:]
self.cheap_reits = self.sorted_reits[:bm5_num]
def calculate_weights(self):
num_expensive = len(self.expensive_reits)
num_cheap = len(self.cheap_reits)
# Calculate equal weights
weight_expensive = 1 / num_expensive
weight_cheap = 1 / num_cheap
# Assign weights to each REIT in the portfolio
self.weights = {}
for data in self.expensive_reits:
self.weights[data] = weight_expensive
for data in self.cheap_reits:
self.weights[data] = weight_cheap