Capital Markets Memorandum Capital Markets and Asset Allocation insights from Frontier Advisors Frontier s Medium Term Fundamental Currency Model November 014 Alvin Tan is a member of the Capital Markets Team and Quantitative Solutions Group and is responsible for asset allocation and quantitative research. Alvin also has client responsibilities for one of Australia s largest industry superannuation funds. Prior to joining Frontier in May 013, Alvin worked at Goldman Sachs Asset Management as a Lead Portfolio Manager and Executive Director within their Quantitative Investment Strategies group. Alvin has also held roles at the Reserve Bank of Australia, Macquarie Funds Management and AllianceBernstein. Alvin holds a Bachelor of Economics (Honours) and a PhD in Econometrics. He is also a CFA charterholder.
Medium term exchange rate model The inability to anticipate changes in supply and demand for a currency is at the root of the statistically robust finding that forecasting exchange rates has a success rate no better than that of forecasting the outcome of a coin toss. Mr Alan Greenspan, FRB speech, 004 Currency forecasting is a notoriously difficult task to get right consistently. However, we have found from our research that over the medium to longer term, a range of fundamentals are useful in determining the fair value of the Australian dollar/us dollar (AUD) exchange rate. A common belief is that there is one true model to predict movements in the exchange rate. However, if one were to ask market participants about what drives a currency, you would be likely to get a wide range of market or economic drivers suggested and, consequently, a broad range of outcomes. Frontier s Capital Market and Asset Allocation Team (CMAAT) uses a variety of models in framing our views on the value of the AUD a very long term model (based on Purchasing Power Parity (PPP)) which we use as the basis for our foreign currency Dynamic Asset Allocation (DAA) hedging advice for clients; and additional models which we have formulated to enable us to better evaluate medium and short term influences on the Australian dollar. These additional models consist of a medium term fundamental model (which is the subject of this memo and is based on economic fundamentals); and a shorter term model (based on economic flows) to reflect the different market participants views on the market. These models, which are useful in assisting us in forecasting the direction of the exchange rate, are most useful when their signals are at extremes and are used to augment our long term PPP valuation. In this memo, we discuss our medium term Australian dollar (AUD)/US dollar (USD) model which has a current AUD fair value estimate of 87 cents to the US dollar. This estimate (which is elevated compared to our long term PPP valuation) is higher mainly due to the lingering effects of the recent commodities boom (likely in its early stages of decline) and is based on current market pricing of several key variables. In the absence of such a boom, our medium term fundamental fair value model suggests a level closer to its long run average of 75 cents (which is in line with our PPP valuation). We would be delighted to discuss our work with you. Page 1
Modelling framework The theory underlying most medium term exchange rate models focuses on fundamentals such as current accounts, productivity and interest rates 1. For small open economies such as Australia, the terms of trade is often an important variable to consider as exports account for approximately 0% of Australia s GDP. At Frontier, in the process of building a medium term model to augment our long term PPP model, we examined a whole range of economic variables (using data from January 1988) and found the following variables to be most useful for forecasting the AUD over the medium term: Interest rate differentials; Inflation rate differentials; and Terms of trade (the relative price of exports to imports). Interest rate differentials Interest rate differentials have been an important driver for the high Australian dollar over the past few years, particularly in an environment of globally low interest rates and investors search for yield (in currency markets this is also known as the carry trade ). For interest rate differentials as inputs into our model, we considered the short rate (i.e. 90-day interest rate) and the year government bond rate. We chose the year rate to reflect monetary policy expectations given we are trying to construct a forwardlooking model. (In fact the impact on results is statistically insignificant if one series is substituted for the other). All things being equal, we would expect a currency with a higher interest rate to appreciate relative to a country with a lower interest rate. As an example, the year interest rate differential between Australia and the US was over 4% in 010/011, which partially explained the high level of the Australian dollar against the US dollar during that time. However, what we are particularly interested in, in modelling this relationship, is whether the AUD overshot its fundamentals during that time. Inflation As well as nominal interest rate differentials, relative inflation is also a driver of currency performance. High inflation is typically negative for a currency as it impacts the relative competitiveness of domestically manufactured goods relative to foreign goods. For an adjustment back to equilibrium, the currency with a higher inflation rate typically depreciates. For inflation rate differentials, we chose the headline Consumer Price Index (CPI) inflation rate. Some other studies have used the real interest rate rather than decomposing the variables separately into nominal interest rate differentials and inflation differentials. We chose the latter two-step method as we wanted to make sure that each of the individual components had the coefficient sign that we would expect, and were both statistically significant. Chart 1: AUD/USD vs interest rate differentials Chart : AUD/USD vs inflation differentials AUDUSD % 10 8 6 4 0-83 86 89 9 AUD/USD (LHS) 95 98 01 04 07 10 13 -yr Interest Rate Differential Source: Frontier/Bloomberg AUDUSD 83 86 89 9 95 AUD/USD (LHS) 98 01 04 07 10 13 Inflation Differential Source: Frontier/Bloomberg % -4-0 4 6 Page
Modelling framework Terms of trade It has been well documented in exchange research that the terms of trade is a significant driver of the AUD/USD exchange rate because it affects our dollar via 3 : The Income effect : Higher incomes via higher terms of trade result in a greater demand for Australian goods, making Australian produced goods more expensive relative to imported goods. To achieve equilibrium, the exchange rate will need to appreciate to make imported goods cheaper relative to domestic produced goods. For the purpose of forecasting however, we find the terms of trade comes out with a -month lag from the Australian Bureau of Statistics. As we are interested in timeliness for the purpose of forecasting, we use the tradable component of the RBA Commodity Price Index as a proxy for Australia s terms of trade (see Table 1 for components and weights). The Investment effect : Increased investment to expand capacity in the resources sector. This will result in capital inflows from foreign investors and thereby boosting the currency. Table 1: Tradable components of Australia s Commodity Price Index Commodity Weight (%) Copper and copper ore 1.5 Aluminium and alumina 1.5 Nickel 1.6 Zinc 1.6 Lead 1.6 Natural gas 15.1 Brent crude 0.1 Wheat 10.1 Sugar.4 Cotton.4 Gold 0.1 Chart 3: AUD/USD vs RBA Tradable Commodity Price Index AUDUSD Index 1.1 4 4 0.9 3 3 0.7 1 0.5 1 0 83 86 89 9 95 98 01 04 07 10 13 AUD/USD (LHS) RBA Commodity Prices (Tradable) Source: Nomura/RBA Source: Frontier/RBA Page 3
AUD/USD Index Per cent of GDP Results Combining all the different factors in our Medium Term AUD model, we determine a fair value estimate of 87 cents as of 30 September which is in line to the actual currency level of 87 cents as of 4 November 014 (see Chart 4). Despite some market participants viewing the reason for the high Australian dollar as being due to Australia s relatively high interest rates (they often focus on the ratio of Australian to US interest rates rather than differentials), our model suggests the interest rate differential is not the key driver for the Australian dollar at present as the interest rate spread is currently around its long term average. In fact, according to our model, the main reason for the high Australian dollar is due to commodity prices remaining at a high level relative to their own history and, in the absence of high commodity prices, we would expect the Australian dollar to be closer to 75 cents. This result is also consistent with our fundamental view that the very significant increase in mining investment over the past decade has resulted in significant capital inflows into Australia from offshore investors, which partially explains the Australian dollar moving to levels significantly higher than its fair value level (particularly in 011 when the Australian dollar hit 1.10). Conversely, once the investment boom in the mining sector comes to an end, we would expect capital inflows to decline significantly. This process has already begun and is likely to continue over the next few years (see Chart 5) and lead consequently to a lower Australian dollar. Chart 4: Fundamental Fair Value Model 1. 1.1 0.9 0.7 0.5 87 89 91 93 95 97 99 01 03 05 07 09 11 13 AUD/USD Rate Fair Value +/- Std. Dev. Chart 5: Terms of Trade vs Work Yet to be Done 15 115 105 95 85 75 65 55 (% GDP) 45 87 90 93 96 99 0 05 08 11 Terms of Trade (lhs) Work Yet to Be Done to GDP 50% 40% 30% 0% 10% 0% Source: Frontier/Bloomberg/Datastream Source: Frontier/Bloomberg Page 4
AUD/USD Results How useful would this model have been in the past for forecasting future performance We find that models are useful as a standalone signal at extremes? At Frontier, we use various quantitative models to assess the historical drivers of the Australian dollar and to give us a better understanding on what drivers we should be focussing on when combining our qualitative/fundamental analysis. We are also interested in understanding whether models such as our medium term fundamental exchange rate model are useful as standalone signals in predicting future movements of the Australian dollar. In order to assess the usefulness of the model, we examined what the model would have predicted if we only used the information available at the point in time when a forecast was made i.e. without the benefit of hindsight (this is a process known as testing out of sample forecasting accuracy, which measures whether the model would have been useful if it had been applied through time). To illustrate this, we investigate the out of sample forecasts from January 1999 onwards (we used the first eleven years of estimation from January 1988 to get our initial forecast). We find that when the Australian dollar is 1 standard deviation away from its fair value, the model provides a reasonable indication of future movements in the Australian dollar (see Chart 6). As an example, since January 1998 the currency has appreciated by an annualised average of 3.3% per year. However, during periods when the currency was 1 standard deviation cheap or expensive, it returned an average annualised return of 6.6% and - 3.0% respectively over the subsequent 1 months (Table ). Chart 6: Fundamental Fair Value Model in real time 1. 1.1 0.9 0.7 0.5 98 00 0 04 06 08 10 1 14 AUD/USD Rate Fair Value +/- 1 Std. Dev. Source: Frontier/Bloomberg/Datastream Table : Average returns on currency when 1 standard deviations expensive/cheap Average 1 Month Forward Percentage Change Expensive (%) Cheap (%) Average (%) -3.0 6.6 3.3 Number of Months 31 60 189 Source: Bloomberg, DataStream, Frontier Page 5
Conclusion Our analysis draws us to a couple of conclusions. First, it suggests that, contrary to the notion that currencies are impossible to forecast, when currency models suggest the AUD/USD is at an extreme valuation, they are useful in forecasting future movements in the Australian dollar. Second, our AUD/USD model helps us quantify the significance of the different variables that are used to explain the currency, and how their relative significance changes over time. This analysis helps us understand the impact expected further falls commodity prices will have on the Australian dollar. These findings are particularly useful for Frontier s dynamic asset allocation process, as they assist Frontier s CMAAT to calibrate both qualitative and quantitative views using all the available information in the most efficient manner. 1 See IMF, 006, Methodology for CGER Exchange Rate Assessments, Research Department, International Monetary Fund, November 006 for further details. We also considered other variables such as relative debt (both short term and long term) and productivity differentials but found the key drivers to be the three drivers mentioned above and hence is the focus of this memo. 3 See Treasury, 01, Understanding the appreciation of the Australian dollar and its policy implications, Macroeconomic Policy Division, the Australian Treasury for further details. Page 6
Level 16, Exhibition Street Melbourne, Victoria 3000 Tel: +61 3 8648 4300 www.frontieradvisors.com.au @frontier_adv About Frontier Advisors: Frontier Advisors is one of Australia s leading asset consultants. We offer a wide range of services and solutions to some of the nation s largest institutional investors including superannuation funds, government/sovereign wealth funds and universities. Our services range from asset allocation and portfolio configuration advice, through to fund manager research and rating, investment auditing and assurance, quantitative modelling and analysis, and general investment consulting advice. With around $0 billion in funds under advice we have been providing investment advice to clients since 1994. Our advice is fully independent of product, manager, or broker conflicts which means our focus is firmly on tailoring optimal solutions and opportunities for our clients. At Frontier, we re on your side. Frontier does not warrant the accuracy of any information or projections in this paper and does not undertake to publish any new information that may become available. Investors should seek individual advice prior to taking action on any of the issues raised in this paper. While this information is believed to be reliable, no responsibility for errors or omissions is accepted by Frontier or any director or employee of the company. Frontier Advisors Pty Ltd ABN 1 074 87 406. AFS Licence No. 4166.