Original paper

SSRN-id2559375.pdf1134.2KB

Abstract

This paper examines the profitability of value strategies based on four different measures of currency valuation. Real exchange rate level strategies generate the largest excess returns on intervals up to 1 month, while real exchange rate changes produce the largest excess returns for intervals of 1-12 months. Purchasing power parity and Big-Mac index based strategies underperform. The returns are not explained by economic state variables or currency risk factors and are larger than estimated transaction costs. A composite strategy based on all four valuation approaches produces superior mean returns but volatility is also higher due to portfolio concentration.

Keywords: Currency, value strategies, real exchange rate, purchasing power parity, Big-Mac Index

Trading rules

Investable assets: 39 currency pairs relative to USD.
Indicator: Real exchange rate change (5-year cumulative change)

Formula: Nominal exchange rate * (Local consumer goods price level / US consumer goods price level)

Monthly actions:

Rank currencies by value (lowest to highest 5-year return)
Open a long position in the top quintile (most undervalued).
Open a short position in the bottom quintile (most overpriced).

Portfolio management:

Equally weighted
Rebalanced monthly

Python code

Backtrader

import backtrader as bt
import pandas as pd

class RealExchangeRateStrategy(bt.Strategy):
    params = (
        ('top_pct', 0.2),
        ('bottom_pct', 0.2),
        ('lookback_years', 5),
    )

    def __init__(self):
        self.inds = dict()
        for d in self.datas:
            self.inds[d] = dict()
            self.inds[d]['nominal_rate'] = d.close
            self.inds[d]['local_cpi'] = bt.feeds.PandasData(dataname=pd.DataFrame(d.params.local_cpi))
            self.inds[d]['us_cpi'] = bt.feeds.PandasData(dataname=pd.DataFrame(d.params.us_cpi))
            self.inds[d]['real_rate'] = bt.Div(bt.Mult(self.inds[d]['nominal_rate'], self.inds[d]['local_cpi']), self.inds[d]['us_cpi'])
            self.inds[d]['real_rate_change'] = bt.ind.PctChange(self.inds[d]['real_rate'], period=self.p.lookback_years * 12)

    def next(self):
        real_rate_changes = {d: self.inds[d]['real_rate_change'][0] for d in self.datas}
        ranked_data = sorted(self.datas, key=lambda d: real_rate_changes[d])

        long_size = int(self.p.top_pct * len(ranked_data))
        short_size = int(self.p.bottom_pct * len(ranked_data))

        for i, d in enumerate(ranked_data):
            if i < long_size:
                self.order_target_percent(d, target=1.0 / long_size)
            elif i >= len(ranked_data) - short_size:
                self.order_target_percent(d, target=-1.0 / short_size)
            else:
                self.order_target_percent(d, target=0)

    def rebalance(self):
        self.next()

cerebro = bt.Cerebro()

# Add currency pairs data feeds with local_cpi and us_cpi parameters
# It's assumed that this part is completed separately

cerebro.addstrategy(RealExchangeRateStrategy)
cerebro.addanalyzer(bt.analyzers.SharpeRatio, _name='sharpe_ratio')
cerebro.addanalyzer(bt.analyzers.DrawDown, _name='drawdown')
cerebro.addanalyzer(bt.analyzers.AnnualReturn, _name='annual_return')

results = cerebro.run()

sharpe_ratio = results[0].analyzers.sharpe_ratio.get_analysis()
drawdown = results[0].analyzers.drawdown.get_analysis()
annual_return = results[0].analyzers.annual_return.get_analysis()

print(f'Sharpe Ratio: {sharpe_ratio}\nMaximum Drawdown: {drawdown.max.drawdown}\nAnnual Return: {annual_return}')

Please note that this code is a starting point, and you will need to adjust it according to your specific data feeds and requirements. The local_cpi and us_cpi parameters should be provided as Pandas DataFrames with appropriate date indices.

Are Value Strategies Profitable in the Foreign Exchange Market?

Original paper

Abstract

Trading rules

Python code

Backtrader