Global Tactical Asset Allocation and Stock Selection
International Fund Flows and the Predictability of Equity Returns
For Global Tactical Asset Allocation and Stock Selection
Team: Plutonium 233
February 28, 2002
[pic]
Harley Adams
[pic]
Brenna Copeland
[pic]
Van Menard
[pic]
Mary Rachide
[pic]
Erik Schneider
Top
Click to jump to:
Project Hypothesis
Data
Background Discussion
Methodology
Model
Trading Strategy
Regression Output
Conclusions
Areas for Further Exploration
Project Hypothesis
We are hypothesizing that data about the flow of investment between Latin American countries and the US (in a limited set of asset classes) can be used to predict country equity performance. The rationale for this hypothesis is that marginal changes in the flow of capital between countries with significant ties could result in liquidity (or lack thereof) that impacts the prices of equities.
(return to top)
Data
The key independent variable tested in this analysis measures the flow of capital between foreign countries and the US. We pulled international funds flow data from the Treasury International Capital (TIC) reporting system (see ). TIC collects monthly mandatory data from banks, securities dealers, and investors on the volume of assets purchased from US residents by foreign residents (capital inflow to the US) and the volume of assets sold to US residents by foreign residents (capital outflow from the US).
The data has several limitations. First, the reports have a two-month time lag to the public release of the data. Second, the data series are revised for up to 24 months after the reporting date, and the data we used starts in January of 1988. The data reflects only transactions between US residents and counter parties located outside the US. Published portfolio flows data that is collected by TIC identify only the country in which the account is located and through which a transaction is made, not the country where the foreign security is issued nor the country where the account benefactor resides.
The TIC data reports the gross flows (from both the US buyer’s and the US seller’s perspective) of six classifications of securities:
Domestic
US Treasury bonds and notes
Bonds of US government corporations and federally-sponsored agencies
US corporate and other bonds
US corporate and other stocks
Foreign
Foreign bonds
Foreign stocks
Other economic variables were considered to create a benchmark predictive model, by which we could measure the value provided by TIC data. We pulled the following additional data from Datastream:
• MSCI country monthly index returns in USD
• Local foreign currency exchange rates (vs. USD)
• Salomon Brothers Brady Bond local country monthly index return
• MSCI US total return index and MSCI World monthly total return index
• Goldman Sachs World Energy Returns Index
• 30-day Eurocurrency rate
(return to top)
Background Discussion
A simple economic interpretation of funds flows to and from a country is not possible. There are a number of contradictory explanations of why funds would move out of local investments and into US or foreign investments: 1) the local market is under-performing 2) the local economy is booming and the market is considered to be over-valued 3) political instability is generating sovereign risk in the local markets and 4) no economic impacts. By regressing information about funds flows against market indices, we are hoping to determine what correlation there might be between funds flows and market performance, if any.
Our primary source of data for funds flow information is the Treasury International Capital reporting system. However, there is some question whether this data accurately proxies for true funds flow in and out of a country, since it merely captures where the transaction takes place (not where the actual transactor lives or the country of the securities are purchased). This question of geographical accuracy of data was extensively explored by Warnock and Mason in their paper entitled, “The Geography of Capital Flows: What We Can Learn From Benchmark Surveys of Foreign Equity Holdings.” Warnock and Mason conclude that while capital flows attributed to certain countries may represent true holdings in other countries, “for some countries, such as Canada and much of Latin America, holdings estimates are quite accurate.” Furthermore, they show that TIC transaction data showed estimated holdings of Mexican and Argentinean stocks were within 5% of actual, and Chilean stocks were within 15%, but Brazilian stocks were underestimated by 32%.
Therefore, we narrowed our focus to relevant companies as measured by most likely to have accurate TIC data, markets related to the US, markets with significant depth of data, and markets of significant size. The countries we considered are:
• Argentina
• Brazil
• Mexico
• Venezuela
(return to top)
Methodology
We looked at our data from three perspectives: benchmark, net, and gross. First, after obtaining logical economic variables, we set out to create benchmark regressions without the TIC data. We wanted to know whether what we found was any different than what the analysis of the TIC data would show.
After finding a benchmark on economic variables, we looked at the net TIC data. Net figures are derived by taking the difference between “Gross Sales of Foreign Equities by U.S. Residents” and “Gross Purchases of Foreign Equities by U.S. Residents.” A positive difference indicates that the local country is purchasing more foreign equities from U.S. Residents than it is selling to U.S. Residents. A negative difference indicates that capital is flowing out of the US and into the local country equities.
TIC reports six classes of securities. After looking at preliminary results of the predictability of all six asset classes, we narrowed the analysis to the net purchases of foreign stocks by US residents. This makes economic sense, since we were testing the predictability of foreign stock returns, and helped prevent data mining.
We grouped the data a number of different ways in order to predict country returns:
1) Net flows (as defined above) both using the actual US$ of the flows as well as the ratio of sales of foreign equities by US investors over the purchases of foreign equities by US investors
2) Gross flows where sales of foreign equities by US investors and purchases of foreign equities by US investors are regressed against country returns separately
3) Binned Net Flows (using the raw data); the data is divided into two bins, one where the lagged net flow of funds is positive and one where the lagged net flow of funds is negative
4) Binned Net Flows (using the ratio data); the data is divided into two bins, one where the lagged ratio of sales to purchases is greater than one and one where the ratio is less than one
(return to top)
Model
We designed a series of models to predict country returns in Argentina, Brazil, Venezuela and Mexico. Based on our methodology, we created base case regressions, net and gross flows predictions and binned net and gross flow predictions. We then evaluated a basic trading strategy for effectiveness and compiled metrics for each subsidiary model. The aggregate statistics about directional counts (in terms of the number of time our model predicted the return in the same direction as the actual return) and the average return and standard deviation of the trading strategy return are all reported in our results.
(return to top)
Trading Strategy
The trading strategy depends on the way the data is segmented. For the net flows and gross flows analyses, the trading strategy compares the predicted return (generated using a 3-month lagged TIC net or gross flow) to the 30-day Eurodollar deposit. When the predicted equity return exceeds the risk-free return, we invest in the local market index. When the predicted equity return is less than the risk-free return, we invest in the risk-free security. We then measure the average returns of our trading strategy versus the average return to the market investment over a period of time. We also look at the volatility of this trading strategy.
(return to top)
Regression Output
We have summarized regression outputs for each of our four countries:
(return to top)
Conclusions
It appears that our models work best in the Brazilian and Venezuelan markets and generally, the binned net flows methodology appears to provide the best trading returns. The trading strategy is not effective in Mexico. Furthermore, the trading strategy delivers superior returns during the time periods where the lagged net funds flow is positive. To implement this strategy, one would need to find an alternate investment strategy when the lagged net funds flows are negative (only Brazil and Venezuela outperform in the negative periods as well as the positive periods). The trading strategy delivers average returns in excess of a buy-and-hold strategy as well as delivering significantly lower standard deviation in most circumstances. This result suggests that the net flows seem to be a leading indicator of market movements. Furthermore, when the fund flows are positive (indicating that gross sales of foreign equities by US residents exceed gross purchases of foreign equities by US residents) the return to the trading strategy tends to be higher. This could reflect that generally speaking, the local equity market is expected to go up three months after a positive net fund flows.
(return to top)
Areas for Further Exploration
Another interesting area to explore in the TIC data would be to look at whether or not funds flowing out of localities and into US treasuries actually precede negative returns in the local equity market.
One might also want to consider whether currency returns could be separated out from equity returns (all our analysis was conducted in US denominated returns).
As these markets develop, bond information would also be worth examining.
It would be interesting to examine how large a portion of the total country market the US investor (net US investment) represents.
One would want to use this data as an input to an optimizer before executing the trading strategy.
It would be interesting to look at an analysis of Canada’s data.
(return to top)
................
................
In order to avoid copyright disputes, this page is only a partial summary.
To fulfill the demand for quickly locating and searching documents.
It is intelligent file search solution for home and business.
Related download
- 1 american businesses get their external funds primarily from
- 3q09 qdia notices no responses american funds
- global tactical asset allocation and stock selection
- chapter eleven new york university
- evaluation of returns for an orp provider
- chapter 4 net present value finance department
- american stock market development and performance
Related searches
- vanguard asset allocation model
- asset allocation in retirement portfolio
- asset allocation after retirement
- asset allocation models by age
- asset allocation strategies by age
- blackrock global asset allocation fund
- best asset allocation during retirement
- morningstar asset allocation model portfolios
- kiplinger asset allocation by age
- bogleheads asset allocation model
- best asset allocation models
- asset allocation calculator