Popular in Professional

Human Rights Law

Human Rights and Corruption: Problems and Potential of Individualising a Systemic Problem

Max Planck Institute Anne Peters

Law

A statistical approach to law school citation rankings

University of Virginia Joshua Fischman

Education

High school change

Boston College Andrew Hargreaves

Constitutional Law

What is constitutional ideology?

University of Oxford Nick Barber

Civil Procedure Law

Politics, Identity, and Class Certification on the U.S. Courts of Appeals

University of Pennsylvania Sean Farhang

Culture and Society

The City as a Growth Machine: Toward a Political Economy of Place

New York University Harvey Moltoch

Education

The Role of Education in Reducing Racial Inequality

University of Southern California Pedro A. Noguera

Chinese Exchange Rate Policy: Lessons for Global Investors

View more insights in Professional - Finance

View more figures from University of California San Diego

View the page of Michael Melvin

Read the Study

Next in Finance

Firm Performance Pay as Insurance against Promotion Risk

Alvin Chen | Stockholm School of Economics

Investor decision making within retirement savings schemes

Robert Durand | Curtin University

How can a dynamic currency overlay strategy, using a forecasting model based on the RMB central parity rate policy introduced by the People’s Bank of China in 2015, help global investors manage currency risk in Chinese equity positions through long and short CNH positions? Michael Melvin discusses.

Image courtesy of interviewee. November 7, 2024

Log-in or Sign-up to Faculti
Currently viewing this subject insight as a guest. You have insight(s) remaining for this month. Login to view 8000+ figures on the platform.

Generating Workspace...

Export as PDF

This video discusses the currency forecasting mechanisms of the People's Bank of China (PBOC) and how they set the central parity rate for the Chinese Yuan (CNY) against the US dollar. Previously, the rate was set at the discretion of the PBOC without market signals, but the current system incorporates market forces based on the previous day’s closing rate and supply-demand conditions. This video change provides more predictability for investors.

The speaker, drawing from experience in both academia and industry, highlights that forecasting exchange rates is challenging. Historical models have often failed to outperform a simple random walk, which predicts that today's rate is the best estimator for the future rate. In their work, they introduce a forecasting model that uses the PBOC's morning rate to predict daily changes in the Yuan's value.

They report a success rate of about 67% in predicting the direction of change, which is significantly better than the random walk's 50% chance. This video model is particularly beneficial for investors holding Chinese equities, as it allows them to hedge currency risks dynamically. Unlike static hedging strategies, their model adapts based on directional forecasts, improving performance during market fluctuations.

Overall, the text reflects on the evolution of currency forecasting methods, including the impact of current technologies such as AI on model development and the shift toward data-driven approaches. The speaker emphasizes the importance of practical application of forecasting rather than just academic benchmarks, suggesting that understanding currency dynamics is vital for investment strategy.

Language:

Export as PDF

Export as PDF
Here are five key points extracted from the video:

1. **People's Bank of China (PBOC) and Central Parity Rate**:
The People's Bank of China sets a daily central parity rate for the Chinese Yuan (CNY) against the US dollar based on the previous day's closing rate and current supply and demand conditions. This approach introduces a level of predictability to the market that was previously absent.

2. **Random Walk Theory in Currency Forecasting**:
The random walk theory suggests that predicting exchange rates is effectively as reliable as using today's rate as a forecast. Academic researchers have struggled to develop models that can outperform this baseline, leading to a general belief that future rates cannot be reliably predicted.

3. **Predictability of Currency Changes**:
The author’s forecasting model utilizes the PBOC's central parity rate to predict changes in the CNH (offshore Chinese Yuan) exchange rate. The model claims to outperform the random walk, achieving about a 67% success rate in predicting the direction of currency changes, making it useful for practitioners managing investments.

4. **Dynamic Currency Hedging Strategies**:
The video discusses the advantages of implementing dynamic hedging strategies compared to static hedges in currency risk management. The proposed dynamic hedge adjusts positions based on predicted currency movements, thereby improving performance compared to a static hedge.

5. **Challenges in Currency Forecasting and AI**:
Currency forecasting is described as a difficult field, with new variables and models constantly being proposed by academics. The advent of machine learning and AI tools has changed the landscape of data analysis in forecasting, enabling a more exploratory approach but raising concerns about the rigor of the methodologies behind such models.

Export as PDF
Here are some key quotes and points extracted from the video :

1. **Market Signals and Central Parity Rate:**
- "There were no market based signals. It was discretion on the part of the People's Bank of China as to where they would set a parity rate each day."
- "The people's bank would set a central parity rate before business hours begin... based on what they call the closing rate yesterday."

2. **Forecasting Challenges:**
- "All the forecasting models that people were using in the early seventies, none of them could outperform a random walk."
- "Today's exchange rate is the best you can do... that’s a very disappointing view, I think, of forecasting."

3. **Random Walk Hypothesis:**
- "The random walk has emerged as the benchmark for currency forecasting. It tells you just use today's forecast."
- "We have about a 67% success rate in predicting the direction of change [unlike the random walk which is 50/50]."

4. **Practitioner Perspective:**
- "If you're managing money you care about direction of change and suggested magnitude of that direction of change."
- "The forecasting model... takes the early morning People's Bank of China central parity rate as an indicator of where the exchange rate should move."

5. **Currency Hedging Strategies:**
- "We built an implementable model... combine your long position in Chinese equities with a currency hedge portfolio."
- "We can design a dynamic hedging strategy where you're short when your model forecasts currency depreciation, but you're long when the model forecasts currency appreciation."

6. **IMF and Currency Politics:**
- "I believe this was done in order for the IMF to curry favor with the Chinese authorities... it was politics."
- "The Chinese currency is a managed currency... it doesn't fit with the dollar, the Euro, the other currencies in the SDR."

7. **Dynamic vs Static Hedging:**
- "Dynamic hedge... gives much better performance than the static... hedge."
- "A currency overlay strategy... will add real value relative to a static hedge on your global equity positions."

8. **Future of Currency Forecasting:**
- "Currency forecasting is a very challenging business and the game for academics... is always this mean square error of the forecast error course race against the random walk."
- "Now, with machine learning models and AI tools, it’s just outright high powered data mining."

These quotes encapsulate key themes of market dynamics, the effectiveness of currency forecasting, and the challenges faced by both academics and practitioners in the field of currency predictions and risk management.

Export as PDF
Here are some articles that discuss similar themes related to currency forecasting, central bank interventions, and the challenges of predicting exchange rates:

1. **"The Impact of Central Bank Policies on Currency Fluctuations"**
This article explores how central banks, like the People's Bank of China, influence their currencies through policy tools, discussing the implications for forex traders and institutional investors. It analyzes the shift from fixed to more market-driven exchange rates, like the one China adopted.

2. **"Modeling Exchange Rates: A Random Walk or Something More?"**
A deep dive into forecasting methods for exchange rates, this article critiques the random walk hypothesis and discusses various models that have emerged over the years, examining their successes and failures in predicting currency movements.

3. **"Assessing Currency Forecasting Models: The Role of Market Signals"**
This piece reviews different currency forecasting models, including those that use market signals to predict currency valuation. It emphasizes the importance of understanding market dynamics, such as supply and demand, when making forecasts.

4. **"Dynamic Hedging Strategies for Forex Investors"**
This article discusses various hedging techniques for managing currency risk in international investments. It highlights the benefits of dynamic versus static hedging strategies, especially in the convideo of changing market conditions and central bank policies.

5. **"AI and Machine Learning in Currency Prediction: Revolution or Just Hype?"**
This article investigates the increasing use of AI and machine learning techniques in forecasting currency movements. It discusses the potential advantages and pitfalls of relying solely on data-driven models without understanding the underlying economic principles.

6. **"The Role of Currency Blocks in Global Finance"**
An analysis of how different currencies, including the Chinese Yuan, fit into global financial frameworks and the implications for investors. It examines the restrictions placed on foreign investments and trading, particularly in relation to emerging market currencies.

7. **"Evaluating Risk Management Strategies in International Equity Investments"**
This article reviews various strategies that investors use to manage risk associated with currency fluctuations when holding foreign equities, including the impact of currency depreciation on overall portfolio performance.

8. **"Central Bank Interventions and Their Effect on Currency Markets"**
A thorough exploration of how central banks, including the People's Bank of China, intervene in currency markets to stabilize their currencies, highlighting significant case studies and their outcomes on forex trading.

These articles would provide valuable insights into forecasting currency movements, the influence of central bank policies, and risk management strategies that align closely with the themes discussed in your video.

X-TIMESTAMP-MAP=LOCAL:00:00:00.000,MPEGTS:90000

1
00:00:00.115 --> 00:00:01.725
Were no market based signals.

2
00:00:02.225 --> 00:00:06.565
It was, discretion on the part of the People's Bank

3
00:00:06.565 --> 00:00:10.525
of China as to where they would set a parody rate each day.

4
00:00:10.825 --> 00:00:13.765
And when they moved to this new system,

5
00:00:14.115 --> 00:00:15.685
they made an announcement

6
00:00:15.685 --> 00:00:17.725
that it would be based more on market forces.

7
00:00:18.185 --> 00:00:21.805
So, and that turned out as, as we showed it in our article,

8
00:00:21.825 --> 00:00:23.205
turned out to be something

9
00:00:23.235 --> 00:00:25.765
that we think was very useful for investors.

10
00:00:26.225 --> 00:00:28.445
The people's bank would set a,

11
00:00:28.665 --> 00:00:31.325
or does fill today, sets a central par

12
00:00:31.465 --> 00:00:34.605
so called central parity rate in the morning, really

13
00:00:34.605 --> 00:00:36.245
before business hours begin.

14
00:00:36.905 --> 00:00:39.285
And that rate is set based on

15
00:00:39.285 --> 00:00:42.005
what they call the closing rate yesterday.

16
00:00:42.555 --> 00:00:47.245
What was the value in the market of the exchange rate?

17
00:00:47.905 --> 00:00:51.885
The CNY Chinese Yuan against the US dollar

18
00:00:52.425 --> 00:00:56.245
at late afternoon the day before, and supply

19
00:00:56.405 --> 00:00:58.645
and demand conditions since then.

20
00:00:58.665 --> 00:01:00.725
And that's the nebulous part right there.

21
00:01:00.945 --> 00:01:04.645
Supply and demand conditions are undefined,

22
00:01:05.185 --> 00:01:08.165
but it does, it's suggestive that

23
00:01:08.715 --> 00:01:10.805
when banks are giving input

24
00:01:11.025 --> 00:01:13.805
to the people's bank in the early morning as to

25
00:01:13.805 --> 00:01:17.245
where they think the currency should be trading early in the

26
00:01:17.245 --> 00:01:21.645
morning, they're asked to use yesterday's co quote,

27
00:01:21.675 --> 00:01:25.445
closing rate and supply and demand conditions.

28
00:01:25.705 --> 00:01:28.685
So the commercial banks are supposed to look at

29
00:01:28.685 --> 00:01:30.405
where were we late yesterday afternoon?

30
00:01:30.945 --> 00:01:34.805
What's happened since then in order to come up with a rate

31
00:01:34.805 --> 00:01:37.525
that they submit to the central bank

32
00:01:38.035 --> 00:01:40.485
that the Central bank can then use to come up

33
00:01:40.485 --> 00:01:42.085
with their daily parity rate.

34
00:01:42.345 --> 00:01:46.165
So it adds a degree of predictability to the market

35
00:01:46.235 --> 00:01:47.765
that did not exist before.

36
00:01:47.985 --> 00:01:49.245
You mentioned Macy and Rogoff.

37
00:01:49.515 --> 00:01:51.965
This, Dick Macy was one

38
00:01:52.085 --> 00:01:54.685
of my former colleagues when I was at BlackRock years ago.

39
00:01:54.985 --> 00:01:57.125
So it's a friend of mine

40
00:01:57.225 --> 00:01:59.405
and he wrote an article in the early seventies

41
00:01:59.405 --> 00:02:02.525
with Ken Roff at Harvard that showed that

42
00:02:03.145 --> 00:02:04.565
all the forecasting models

43
00:02:04.915 --> 00:02:07.845
that people were using in the early seventies, none

44
00:02:07.845 --> 00:02:10.725
of them could outperform a random walk.

45
00:02:11.115 --> 00:02:15.085
What does that mean? It means a random walk means your best

46
00:02:15.445 --> 00:02:19.525
forecast of a future exchange rate is today's exchange rate.

47
00:02:20.035 --> 00:02:22.965
It's as likely to go up as it is to go down.

48
00:02:23.745 --> 00:02:26.925
So today's exchange rate is the best you can do.

49
00:02:27.385 --> 00:02:32.005
That's a very disappointing view, I think, of forecasting.

50
00:02:32.705 --> 00:02:36.525
And since that article, that has been the game

51
00:02:36.745 --> 00:02:41.205
for economists who forecast exchange rates to try

52
00:02:41.205 --> 00:02:43.445
and outperform the random walk.

53
00:02:43.505 --> 00:02:46.685
So the random walk has emerged as the benchmark

54
00:02:47.025 --> 00:02:48.405
for currency forecasting.

55
00:02:49.225 --> 00:02:53.685
So the random walk tells you just use today's forecast.

56
00:02:53.825 --> 00:02:56.165
And if you're forecasting, say 30 days ahead

57
00:02:56.265 --> 00:02:58.765
or 90 days ahead, you use today's

58
00:02:59.125 --> 00:03:00.965
Forecast as the best forecast of

59
00:03:00.965 --> 00:03:02.325
that future exchange rate.

60
00:03:02.785 --> 00:03:05.965
And then researchers would build all these forecasting

61
00:03:05.965 --> 00:03:09.205
models to try and outperform the random walk

62
00:03:09.265 --> 00:03:11.765
or outperform to research.

63
00:03:11.775 --> 00:03:14.445
Economists has traditionally been in terms

64
00:03:14.465 --> 00:03:17.605
of means square error of the forecast.

65
00:03:17.905 --> 00:03:21.685
So forecast being square error has been the horse race

66
00:03:22.235 --> 00:03:24.925
that people have run against the random walk.

67
00:03:25.385 --> 00:03:27.845
And that's fine for academic work,

68
00:03:27.985 --> 00:03:31.365
but that's not the way practitioners think in terms of

69
00:03:31.965 --> 00:03:33.965
forecasting for investing purposes.

70
00:03:34.725 --> 00:03:36.965
A practitioner wants to know which way

71
00:03:37.665 --> 00:03:39.005
the exchange rate's going.

72
00:03:39.345 --> 00:03:41.245
Is the currency going to go up in value?

73
00:03:41.385 --> 00:03:42.805
Is it gonna go down in value?

74
00:03:43.305 --> 00:03:45.965
So the practitioner wants direction of change

75
00:03:46.825 --> 00:03:51.325
and the mean forecast error, it can be small

76
00:03:51.545 --> 00:03:54.165
and it can be wrong every time on the direction,

77
00:03:54.305 --> 00:03:56.645
but it could be smaller than a random walk.

78
00:03:56.705 --> 00:03:58.685
And the, the researcher would say,

79
00:03:59.125 --> 00:04:00.165
I beat the random walk.

80
00:04:00.325 --> 00:04:01.845
I have a great forecasting model.

81
00:04:02.545 --> 00:04:06.325
So for us, what we wanted to do, you know,

82
00:04:06.505 --> 00:04:08.245
bef I was a professor for many years,

83
00:04:08.245 --> 00:04:10.165
then I went into industry for 11 years

84
00:04:10.845 --> 00:04:12.005
managing currency hedge funds,

85
00:04:12.105 --> 00:04:15.325
and now I'm back In academia, if you're managing money

86
00:04:15.425 --> 00:04:16.965
and you're investing in this market,

87
00:04:17.425 --> 00:04:19.645
you care about direction of change

88
00:04:20.265 --> 00:04:24.325
and suggested magnitude of that direction of change,

89
00:04:24.985 --> 00:04:27.925
you don't care about mean square forecast error.

90
00:04:28.265 --> 00:04:29.845
You want to be on the right side

91
00:04:30.225 --> 00:04:33.325
of the change in the exchange rate that is realized.

92
00:04:34.105 --> 00:04:37.925
So the forecasting model that we have in this paper takes

93
00:04:37.995 --> 00:04:40.085
that early morning, people's Bank

94
00:04:40.085 --> 00:04:44.605
of China central parity rate as an indicator of

95
00:04:44.605 --> 00:04:46.205
where the exchange rate should move

96
00:04:46.335 --> 00:04:48.365
today relative to yesterday.

97
00:04:49.225 --> 00:04:54.045
So we're forecasting the change in the CNH exchange rate.

98
00:04:54.305 --> 00:04:58.285
That's the onshore trade that you can do in Hong Kong.

99
00:04:58.545 --> 00:05:00.685
That's a market where people

100
00:05:00.685 --> 00:05:04.045
around the world can trade in the, so-called onshore market,

101
00:05:04.465 --> 00:05:08.565
you can't trade the, so-called CNY,

102
00:05:09.305 --> 00:05:12.685
the mainland Chinese currency onshore, you trade

103
00:05:12.685 --> 00:05:14.285
that currency offshore

104
00:05:14.585 --> 00:05:17.605
and hedge funds like I used to manage, they trade.

105
00:05:17.605 --> 00:05:21.285
So-called non-deliverable forward contracts on the CNH

106
00:05:21.285 --> 00:05:24.725
where you don't settle in any, rebi,

107
00:05:24.725 --> 00:05:26.045
you settle in US dollars.

108
00:05:26.545 --> 00:05:29.325
That's the, the market for offshore.

109
00:05:29.465 --> 00:05:32.165
So we're using the CNH, which is a market

110
00:05:32.195 --> 00:05:35.125
that everyone can trade in forecasting

111
00:05:35.185 --> 00:05:37.525
how it changes from yesterday to today,

112
00:05:37.945 --> 00:05:41.525
taking the morning's central parody rate from the central

113
00:05:41.595 --> 00:05:44.365
bank as indicative of where that should change.

114
00:05:44.665 --> 00:05:47.805
And we find it has significant predictability.

115
00:05:48.425 --> 00:05:50.565
We outperform the random, so the game

116
00:05:50.585 --> 00:05:52.245
of the academic economist,

117
00:05:52.425 --> 00:05:56.445
we outperform the random walk in terms of mean square error,

118
00:05:56.865 --> 00:06:01.325
But we also generate about a two thirds probability

119
00:06:01.345 --> 00:06:03.205
of getting the direction correct.

120
00:06:03.435 --> 00:06:06.565
Whereas the random walk, it's 50 50, it's likely

121
00:06:06.585 --> 00:06:07.725
to go up, it's down.

122
00:06:08.225 --> 00:06:12.325
We have about a 67% success rate in predicting the

123
00:06:12.325 --> 00:06:13.485
direction of change.

124
00:06:14.105 --> 00:06:16.605
So that makes it of use to practitioners.

125
00:06:17.025 --> 00:06:18.285
We take it from a perspective

126
00:06:18.745 --> 00:06:20.885
of an investor in Chinese equities.

127
00:06:21.745 --> 00:06:24.485
If I am based in the us which I am in California

128
00:06:24.665 --> 00:06:28.685
and I hold us, I have a portfolio of Chinese equities.

129
00:06:29.185 --> 00:06:30.725
I'm long the Chinese equities,

130
00:06:30.965 --> 00:06:32.405
I have bought Chinese equities,

131
00:06:32.755 --> 00:06:37.165
that means I'm implicitly long, rebi as well

132
00:06:37.165 --> 00:06:40.405
because Chinese equities are priced in the local currency.

133
00:06:41.275 --> 00:06:43.605
Well, when Chinese equities sell off,

134
00:06:44.145 --> 00:06:47.205
say there's a big drop in the market, we see

135
00:06:47.205 --> 00:06:51.085
that generally the currency also sells off at the same time

136
00:06:51.305 --> 00:06:53.485
and you get a big drop in the currency market.

137
00:06:54.385 --> 00:06:57.165
So exploiting that fact for risk,

138
00:06:57.185 --> 00:06:59.045
you mentioned risk management purposes.

139
00:06:59.465 --> 00:07:02.605
If you combine your long position in Chinese equities

140
00:07:02.875 --> 00:07:04.685
with a currency hedge portfolio,

141
00:07:05.675 --> 00:07:08.805
then when the market sells off, you are

142
00:07:09.515 --> 00:07:11.525
winning from your hedge portfolio.

143
00:07:12.185 --> 00:07:15.525
So generally international investors hedge typically

144
00:07:15.525 --> 00:07:17.685
around 50% of the exposure

145
00:07:18.115 --> 00:07:20.885
that they have in a currency from their international

146
00:07:20.885 --> 00:07:21.925
equity investments.

147
00:07:22.385 --> 00:07:25.405
So if I've got say one US dollar worth of Chinese equities

148
00:07:25.865 --> 00:07:30.325
and I short 50 cents US worth of Chinese currency,

149
00:07:31.155 --> 00:07:34.765
then that hedge will give me a good performance when the

150
00:07:35.045 --> 00:07:37.845
currency market's selling off in China when the equity

151
00:07:37.885 --> 00:07:38.925
market's selling off.

152
00:07:38.925 --> 00:07:40.565
So that helps protect me.

153
00:07:40.865 --> 00:07:44.085
On the downside, we built an implementable model, something

154
00:07:44.085 --> 00:07:45.405
that you could actually trade.

155
00:07:46.515 --> 00:07:50.045
Typical currency hedging is a so-called static hedge

156
00:07:50.735 --> 00:07:52.045
where you just always,

157
00:07:52.045 --> 00:07:55.365
you're always short Chinese currency if you're long Chinese

158
00:07:55.805 --> 00:07:57.725
equities, but you, as we did,

159
00:07:57.825 --> 00:08:00.285
we can design a dynamic hedging strategy

160
00:08:00.895 --> 00:08:03.405
where you're short when your model forecast

161
00:08:04.165 --> 00:08:07.205
currency depreciation, but you're long when the model

162
00:08:07.605 --> 00:08:09.325
forecast currency appreciation.

163
00:08:10.105 --> 00:08:11.125
And we found that

164
00:08:11.125 --> 00:08:14.485
that gives much better performance than the static.

165
00:08:14.595 --> 00:08:17.045
So-called static hedge, I believe,

166
00:08:17.045 --> 00:08:20.045
and I have no visibility into the internal politics

167
00:08:20.045 --> 00:08:21.045
of the IMF,

168
00:08:21.385 --> 00:08:25.365
but I believe this was done in order for the IMF

169
00:08:25.665 --> 00:08:28.925
to curry favor with the Chinese authorities

170
00:08:29.235 --> 00:08:33.925
because the currencies that made up the SDR,

171
00:08:34.385 --> 00:08:37.525
if you look at the rules, the IMF rules, they have

172
00:08:37.525 --> 00:08:42.205
to be fully convertible currencies, fully trade, easily,

173
00:08:42.675 --> 00:08:44.645
implementable and strategies

174
00:08:44.645 --> 00:08:46.085
where you can freely trade them.

175
00:08:46.385 --> 00:08:49.045
You can't do that with the Chinese currency.

176
00:08:49.305 --> 00:08:51.285
It is not a freely tradable currency.

177
00:08:51.595 --> 00:08:52.885
It's a managed currency.

178
00:08:53.065 --> 00:08:55.485
And there are restrictions on foreigners holding

179
00:08:55.585 --> 00:08:56.645
and trading the currency.

180
00:08:57.225 --> 00:08:59.485
So it doesn't fit with the dollar, the Euro,

181
00:08:59.625 --> 00:09:01.885
the British pound and the Japanese Yen,

182
00:09:01.905 --> 00:09:03.885
the other currencies in the SDR,

183
00:09:04.465 --> 00:09:08.045
that's why I believe it was a political move had no

184
00:09:08.375 --> 00:09:12.525
importance in terms of how you would trade, the r and b.

185
00:09:12.945 --> 00:09:14.805
It was, politics.

186
00:09:14.985 --> 00:09:16.965
We actually had several different strategies

187
00:09:16.965 --> 00:09:18.045
that we investigated.

188
00:09:18.305 --> 00:09:23.255
We found when international investors can hedge their

189
00:09:23.495 --> 00:09:26.135
currency portfolio specifically by shorting the currency,

190
00:09:26.275 --> 00:09:28.335
as I said, and we built that strategy,

191
00:09:29.115 --> 00:09:33.415
or they could have a dynamic hedge where they vary

192
00:09:34.035 --> 00:09:36.695
the size and sign of their hedge.

193
00:09:36.695 --> 00:09:39.015
Sometimes they're long, sometimes they're short,

194
00:09:39.115 --> 00:09:42.295
but they time it with the signals, trading signals

195
00:09:42.295 --> 00:09:43.495
that come from our model.

196
00:09:44.155 --> 00:09:46.415
Or they can use what's known as a,

197
00:09:46.775 --> 00:09:48.255
currency overlay strategy.

198
00:09:49.035 --> 00:09:50.135
And this is very popular

199
00:09:50.595 --> 00:09:52.615
for currency hedge fund managers like,

200
00:09:52.765 --> 00:09:55.335
like I was when I was at, BGI,

201
00:09:55.335 --> 00:09:57.255
Barclays Global Investors in BlackRock.

202
00:09:57.355 --> 00:10:00.615
So there, a big asset owner

203
00:10:00.915 --> 00:10:05.415
who has global equity positions will buy a currency

204
00:10:06.245 --> 00:10:08.815
overlay strategy, which is an active

205
00:10:09.455 --> 00:10:13.255
currency investing strategy that runs in parallel

206
00:10:13.445 --> 00:10:15.175
with the global equity portfolio.

207
00:10:15.995 --> 00:10:20.215
So the thinking is the correlations are very low generally,

208
00:10:20.715 --> 00:10:23.055
but the value added from going long

209
00:10:23.055 --> 00:10:27.215
or short currencies will add real value relative

210
00:10:27.275 --> 00:10:31.175
to a static hedge on your global equity positions.

211
00:10:31.715 --> 00:10:35.375
And we show in our paper that that does add the most value

212
00:10:35.375 --> 00:10:38.735
because it's the less the least constrained model

213
00:10:39.315 --> 00:10:41.855
and it lets the investor take long

214
00:10:41.855 --> 00:10:43.775
and short positions vary the size

215
00:10:43.775 --> 00:10:45.335
of those positions over time.

216
00:10:45.995 --> 00:10:50.975
And that, increased flexibility in the investment

217
00:10:51.535 --> 00:10:54.215
strategy pays off in the sample period

218
00:10:54.405 --> 00:10:57.695
that we used when we wrote the paper Future research

219
00:10:57.755 --> 00:10:59.175
on currency forecasting.

220
00:11:00.355 --> 00:11:03.005
It's, it's generally the way this has gone over time

221
00:11:03.705 --> 00:11:06.765
is a researcher will come up with a new variable.

222
00:11:06.945 --> 00:11:09.485
For instance, we should look at this variable.

223
00:11:09.545 --> 00:11:12.045
It will be useful in forecasting currencies.

224
00:11:12.305 --> 00:11:14.445
And if you look at the literature over time,

225
00:11:14.825 --> 00:11:18.125
you'll see this sequence of papers that are written

226
00:11:18.275 --> 00:11:19.805
that say, we beat the random wall.

227
00:11:20.065 --> 00:11:23.525
But when you then look at those signals over a longer period

228
00:11:23.645 --> 00:11:25.525
of time, they don't hold up.

229
00:11:26.085 --> 00:11:29.165
Currency forecasting is a very challenging business

230
00:11:29.865 --> 00:11:33.605
and the game for academics again,

231
00:11:33.625 --> 00:11:35.565
is always this mean square error

232
00:11:36.145 --> 00:11:40.165
of the forecast error course race against the random walk.

233
00:11:40.345 --> 00:11:44.645
And so I suspect we'll see researchers in the future doing

234
00:11:44.645 --> 00:11:48.085
the same thing, coming up with new ways of forecasting.

235
00:11:48.085 --> 00:11:50.325
And of course in the modern era right now,

236
00:11:50.865 --> 00:11:52.525
people are using all sorts of

237
00:11:52.865 --> 00:11:56.565
AI tools to generate new models for forecasting.

238
00:11:56.775 --> 00:11:57.805
Where, it's funny

239
00:11:57.805 --> 00:12:01.765
because when I was in the industry quantitative investing as

240
00:12:01.905 --> 00:12:06.085
as I did, I led quant investment teams, at BlackRock

241
00:12:06.085 --> 00:12:09.085
and BGI, there was a rule, no data mining.

242
00:12:09.545 --> 00:12:11.485
You know, you've gotta have a story.

243
00:12:11.985 --> 00:12:13.365
Here's why I think this works.

244
00:12:13.625 --> 00:12:15.165
And then you test it with the data,

245
00:12:15.545 --> 00:12:17.405
but you can't just get a bunch of variables

246
00:12:17.405 --> 00:12:19.485
and see if they work and not understand why.

247
00:12:19.795 --> 00:12:22.565
Well, where we've gone now with machine learning models

248
00:12:22.825 --> 00:12:26.685
and AI tools, it's just outright high powered data mining.

249
00:12:27.345 --> 00:12:29.805
And so we've seen this 300,

250
00:12:30.145 --> 00:12:32.565
I'd say 180 degree swing in terms of

251
00:12:32.585 --> 00:12:34.325
how people think about data mining.

252
00:12:34.785 --> 00:12:36.965
If basically it's, if it works, it works

253
00:12:37.385 --> 00:12:40.845
and we'll figure out maybe why it works if we see

254
00:12:40.845 --> 00:12:43.165
that we've got something that can actually predict.

255
00:12:43.625 --> 00:12:44.805
So there are all sorts

256
00:12:44.805 --> 00:12:48.765
of things being done these days in terms of forecasting

257
00:12:49.335 --> 00:12:51.725
would just have a huge menu of variables

258
00:12:51.945 --> 00:12:54.525
and let the model sort it out, what seems to work

259
00:12:54.905 --> 00:12:57.925
and then we'll tell our stories about why it works.

260
00:12:57.925 --> 00:12:59.525
Perhaps if we can figure that out.

Market Forces and Currency Parity: Analyzing the People's Bank of China's Approach

00:00:00 - Were no market based signals. It was, discretion on the part of the People's Bank of China as to where they would set a parody rate each day. And when they moved to this new system, they made an announcement that it would be based more on market forces. So, and that turned out as, as we showed it in our article, turned out to be something that we think was very useful for investors. The people's bank would set a, or does fill today, sets a central par

00:00:34 - before business hours begin. And that rate is set based on what they call the closing rate yesterday. What was the value in the market of the exchange rate? The CNY Chinese Yuan against the US dollar at late afternoon the day before, and supply and demand conditions since then. And that's the nebulous part right there. Supply and demand conditions are undefined,

00:01:08 - when banks are giving input to the people's bank in the early morning as to where they think the currency should be trading early in the morning, they're asked to use yesterday's co quote, closing rate and supply and demand conditions. So the commercial banks are supposed to look at where were we late yesterday afternoon? What's happened since then in order to come up with a rate that they submit to the central bank

Currency Forecasting: The Random Walk Challenge

00:01:40 - with their daily parity rate. So it adds a degree of predictability to the market that did not exist before. You mentioned Macy and Rogoff. This, Dick Macy was one of my former colleagues when I was at BlackRock years ago. So it's a friend of mine and he wrote an article in the early seventies with Ken Roff at Harvard that showed that all the forecasting models that people were using in the early seventies, none of them could outperform a random walk.

00:02:15 - forecast of a future exchange rate is today's exchange rate. It's as likely to go up as it is to go down. So today's exchange rate is the best you can do. That's a very disappointing view, I think, of forecasting. And since that article, that has been the game for economists who forecast exchange rates to try

00:02:43 - So the random walk has emerged as the benchmark for currency forecasting. So the random walk tells you just use today's forecast. And if you're forecasting, say 30 days ahead or 90 days ahead, you use today's Forecast as the best forecast of that future exchange rate. And then researchers would build all these forecasting models to try and outperform the random walk or outperform to research.

Forecasting Currency Movements: Practitioners vs. Academics

00:03:14 - of means square error of the forecast. So forecast being square error has been the horse race that people have run against the random walk. And that's fine for academic work, but that's not the way practitioners think in terms of forecasting for investing purposes. A practitioner wants to know which way the exchange rate's going. Is the currency going to go up in value? Is it gonna go down in value?

00:03:46 - and the mean forecast error, it can be small and it can be wrong every time on the direction, but it could be smaller than a random walk. And the, the researcher would say, I beat the random walk. I have a great forecasting model. So for us, what we wanted to do, you know, bef I was a professor for many years, then I went into industry for 11 years managing currency hedge funds, and now I'm back In academia, if you're managing money

00:04:17 - you care about direction of change and suggested magnitude of that direction of change, you don't care about mean square forecast error. You want to be on the right side of the change in the exchange rate that is realized. So the forecasting model that we have in this paper takes that early morning, people's Bank of China central parity rate as an indicator of where the exchange rate should move

Forecasting CNH: Predictability and Market Implications

00:04:49 - So we're forecasting the change in the CNH exchange rate. That's the onshore trade that you can do in Hong Kong. That's a market where people around the world can trade in the, so-called onshore market, you can't trade the, so-called CNY, the mainland Chinese currency onshore, you trade that currency offshore and hedge funds like I used to manage, they trade.

00:05:21 - where you don't settle in any, rebi, you settle in US dollars. That's the, the market for offshore. So we're using the CNH, which is a market that everyone can trade in forecasting how it changes from yesterday to today, taking the morning's central parody rate from the central bank as indicative of where that should change. And we find it has significant predictability.

00:05:50 - of the academic economist, we outperform the random walk in terms of mean square error, But we also generate about a two thirds probability of getting the direction correct. Whereas the random walk, it's 50 50, it's likely to go up, it's down. We have about a 67% success rate in predicting the direction of change. So that makes it of use to practitioners. We take it from a perspective

Hedging Strategies for Chinese Equity Investments

00:06:21 - If I am based in the us which I am in California and I hold us, I have a portfolio of Chinese equities. I'm long the Chinese equities, I have bought Chinese equities, that means I'm implicitly long, rebi as well because Chinese equities are priced in the local currency. Well, when Chinese equities sell off, say there's a big drop in the market, we see that generally the currency also sells off at the same time

00:06:54 - So exploiting that fact for risk, you mentioned risk management purposes. If you combine your long position in Chinese equities with a currency hedge portfolio, then when the market sells off, you are winning from your hedge portfolio. So generally international investors hedge typically around 50% of the exposure that they have in a currency from their international equity investments.

00:07:25 - and I short 50 cents US worth of Chinese currency, then that hedge will give me a good performance when the currency market's selling off in China when the equity market's selling off. So that helps protect me. On the downside, we built an implementable model, something that you could actually trade. Typical currency hedging is a so-called static hedge where you just always, you're always short Chinese currency if you're long Chinese

Dynamic Hedging Strategies and Currency Politics

00:07:57 - we can design a dynamic hedging strategy where you're short when your model forecast currency depreciation, but you're long when the model forecast currency appreciation. And we found that that gives much better performance than the static. So-called static hedge, I believe, and I have no visibility into the internal politics of the IMF, but I believe this was done in order for the IMF to curry favor with the Chinese authorities

00:08:34 - if you look at the rules, the IMF rules, they have to be fully convertible currencies, fully trade, easily, implementable and strategies where you can freely trade them. You can't do that with the Chinese currency. It is not a freely tradable currency. It's a managed currency. And there are restrictions on foreigners holding and trading the currency. So it doesn't fit with the dollar, the Euro,

00:09:01 - the other currencies in the SDR, that's why I believe it was a political move had no importance in terms of how you would trade, the r and b. It was, politics. We actually had several different strategies that we investigated. We found when international investors can hedge their currency portfolio specifically by shorting the currency, as I said, and we built that strategy, or they could have a dynamic hedge where they vary

Dynamic Currency Strategies for Enhanced Portfolio Value

00:09:36 - Sometimes they're long, sometimes they're short, but they time it with the signals, trading signals that come from our model. Or they can use what's known as a, currency overlay strategy. And this is very popular for currency hedge fund managers like, like I was when I was at, BGI, Barclays Global Investors in BlackRock. So there, a big asset owner who has global equity positions will buy a currency

00:10:09 - currency investing strategy that runs in parallel with the global equity portfolio. So the thinking is the correlations are very low generally, but the value added from going long or short currencies will add real value relative to a static hedge on your global equity positions. And we show in our paper that that does add the most value because it's the less the least constrained model

00:10:41 - and short positions vary the size of those positions over time. And that, increased flexibility in the investment strategy pays off in the sample period that we used when we wrote the paper Future research on currency forecasting. It's, it's generally the way this has gone over time is a researcher will come up with a new variable. For instance, we should look at this variable.

Challenges and Evolution in Currency Forecasting

00:11:12 - And if you look at the literature over time, you'll see this sequence of papers that are written that say, we beat the random wall. But when you then look at those signals over a longer period of time, they don't hold up. Currency forecasting is a very challenging business and the game for academics again, is always this mean square error of the forecast error course race against the random walk.

00:11:44 - the same thing, coming up with new ways of forecasting. And of course in the modern era right now, people are using all sorts of AI tools to generate new models for forecasting. Where, it's funny because when I was in the industry quantitative investing as as I did, I led quant investment teams, at BlackRock and BGI, there was a rule, no data mining. You know, you've gotta have a story.

00:12:13 - And then you test it with the data, but you can't just get a bunch of variables and see if they work and not understand why. Well, where we've gone now with machine learning models and AI tools, it's just outright high powered data mining. And so we've seen this 300, I'd say 180 degree swing in terms of how people think about data mining. If basically it's, if it works, it works and we'll figure out maybe why it works if we see that we've got something that can actually predict.

Evolution of Forecasting: From Rigid Models to Data-Driven Insights

00:12:57 - of things being done these days in terms of forecasting would just have a huge menu of variables and let the model sort it out, what seems to work and then we'll tell our stories about why it works. Perhaps if we can figure that out.

Faculti AI can make mistakes

Faculti

Popular in Professional

Human Rights and Corruption: Problems and Potential of Individualising a Systemic Problem

A statistical approach to law school citation rankings

High school change

What is constitutional ideology?

Politics, Identity, and Class Certification on the U.S. Courts of Appeals

The City as a Growth Machine: Toward a Political Economy of Place

The Role of Education in Reducing Racial Inequality

Chinese Exchange Rate Policy: Lessons for Global Investors

Next in Finance

Guide

Platform and Category Pages

Contact Us

FAQs

What is Faculti?

How do you select interviews?

Interview Process

Page Structure

Popular in Professional

Guide

Platform and Category Pages

Contact Us

FAQs

What is Faculti?

How do you select interviews?

Interview Process

Accessibility Options

Contrast Adjustment

Text Size Adjustment

Text Spacing Adjustment

Line Height Adjustment

Cursor Size Adjustment

Page Structure Options

Page Structure