High-Frequency Capital Flow Proxies Are Good At Predicting BoP Flows

Introduction

Policymakers need timely information on capital flows to inform decisions on monetary policy and foreign exchange intervention. Official balance of payments (BoP) data are of little help, given their long release lag. Instead, high-frequency capital flow proxies can serve as early warning indicators. The availability of such data has improved dramatically over the past two decades, but a lack of transparency from private sector data providers combined with conceptual and measurement issues means there are no comparable datasets.

A new IMF working paper is set out as a guide for those who need timely information on capital flow developments. The paper:

• Provides an overview of high-frequency datasets currently used as proxies for portfolio/capital flows.
• Conducts a meta-analysis on the use of the data sources in the empirical literature.
• Addresses a gap in the availability of free, timely and reliable portfolio flow data.
• Measures how well widely used portfolio flow proxies track portfolio flow data in real-time.

The key conclusions are as follows:

• There are four key capital flow data sources: the IMF BoP Statistics, the BIS Locational and Consolidated Banking Statistics, the IIF Capital Flow Data and Portfolio Flow Trackers, and the EPFR Fund Flow Data.
• A review of the literature shows that academic researchers prefer the IMF balance of payments database, while policymakers favour the high-frequency data sources, such as the Emerging Portfolio Fund Research (EPFR) database.
• In a nowcasting ‘horse race’, the paper shows that all high-frequency portfolio flow proxies have significant predictive content for BoP portfolio flows. Among the various predictors, IIF portfolio flow trackers generally outperform EPFR fund flow data.

On this last result, Figure 1 shows the results from the nowcasting horse race. The capital flow proxies are weekly/monthly EPFR data, daily/weekly IIF data and monthly KP data (explained later). The performance of these proxies at predicting actual equity flows (left) and debt flows (right) is measured by the root mean squared forecasting error (RMSFE) in percentage of GDP – the lower the percentage, the better the estimate.

On both counts, daily/monthly IIF proxies are better high-frequency predictors of actual BoP flows, and the predicting power improves over the quarter. The RHS axis expresses the forecast performance as a share of absolute average quarterly flows. The RMSFEs of debt and equity flows based on daily and monthly IIF data outperform weekly and monthly EPFR data, respectively, by around 20% early in the quarter and 50% at the end of the quarter.

Source: Page 27 of ‘Koepke and Paetzold (2020)’

Existing Capital Flow Datasets

The IMF paper looks at 88 academic studies and 111 policy publications to decipher and provide an overview of the most commonly used data sources. I provide links to a number of the sources at the end of this Deep Dive.

Source: Page 7 of ‘Koepke and Paetzold (2020)’

Four key data sources emerge:

1. IMF Balance of Payments Statistics: Provides the most comprehensive coverage of capital flow data. It covers countries, various capital flow components, as well as major international transactions. For the majority of countries, data are available on a quarterly and annual basis and are typically released with a lag of two to four months.
2. BIS Locational Banking Statistics and Consolidated Banking Statistics: The cross-border banking data provides a comprehensive coverage of international banking flows and positions, including currency composition, instrument type, and sector and residency of counterparty. The LBS data are available on a quarterly basis, while CBS data are available at both a semi-annual and quarterly frequency.
3. IIF Capital Flow and Portfolio Flow Trackers: These provide independent, private sector estimates of capital flows to and from emerging market economies. Up until 2013, IIF data was exclusively at the annual frequency. Since 2014, the data is monthly and daily. The dataset serves as timely proxies for BoP-based portfolio flow data, constructed using country-level data from individual national sources.
4. Emerging Portfolio Fund Research (EPFR): The most widely used data provider of subscription-based data on flows to investment funds. It is available at monthly, weekly and daily intervals for EMs as a group or at a country level. The data covers more than 18,000 reporting equity funds and more than 9,000 reporting debt funds, amounting to 96% of assets under management (AUM) of the global investment fund industry.

High-Frequency Portfolio Flow Proxies

Portfolio flows (see ‘A Primer on Balance of Payment’ section below for more details) have received substantial attention from academics, policymakers and market participants. Since the GFC they have grown rapidly and are the most volatile component of capital flows. Moreover, portfolio debt and equity flows are most closely tied to asset price and exchange rate fluctuations, making them relevant for central bank policy decisions. In addition, data availability on portfolio flows is far better than for any other component of capital flows, since there are no comparable monthly, weekly or daily data on banking flows, FDI, or ‘other investment’ flows.

The growing use of high-frequency portfolio flow proxies is evident in both academic and policy papers. As it is alluded to above, there are two main datasets on high-frequency portfolio flows, and they differ widely in scope. The IMF paper compares and contrasts them.

IIF Capital Flow:

The IIF’s monthly tracker for overall emerging market flows uses an econometric model in which financial variables and bond issuance data supplement underlying portfolio flow data to minimize the statistical deviation from quarterly BoP portfolio flow data. Daily data are simply the sum of all reported country data. In terms of the underlying data, it is worth noting that for some countries, daily flows are estimated based on stock data, which are sometimes published in local currency terms. Also, purchases of newly issued bonds and redemptions of maturing bonds are not included.

The IIF’s monthly and daily data aim to be consistent with BoP accounting principles, in which an inflow is recorded if there is a transaction between a non-resident and a resident. The dataset matches closely with BoP portfolio equity flows, but not so well for bond flows.

In terms of coverage, the IIF monthly and especially daily flows, do not include specific types of portfolio transactions if these are not reported by the original data provider (such as a national central bank). For example, some countries’ data do not include hard currency debt or corporate debt. Finally, on the country sample, the emerging market countries are limited to those who make available timely portfolio flow data at monthly (35 countries) and daily (21 countries) frequencies.

Fund Flow data from EPFR:

A major benefit of EPFR data is that fund flow data are available by fund types (i.e. ETFs vs. mutual funds etc.). Moreover, the data can be disaggregated by the fund domicile, investment benchmark, thematic fund category, and by local vs hard currency denomination. The data can also be disaggregated by specific categories for both equity and bond funds.

On coverage, an important caveat is that EPFR data do not cover all types of emerging market investors, only those investing via mutual funds and exchange traded funds. For example, large institutional investors like sovereign wealth funds, pension funds, hedge funds, and banks’ proprietary trading desks typically purchase EM securities directly and are generally not reflected in EPFR data.

Furthermore, while IIF capital flow data is consistent with BoP principles, flows into investment funds do not necessarily result in cross-border transactions. This issue also applies to residents of emerging markets who purchase shares of funds that invest domestically and is particularly relevant for Thailand and India. Finally, in terms of the country sample, the emerging market universe is guided by which countries are included in key benchmark indices and covers data for a total of 98 countries for equity and 113 countries for debt flows.

Which High-Frequency Dataset Should You Use?

Both datasets are useful for monitoring directional shifts in investor interest in emerging market assets. Overall, EPFR data seem best suited for analysing questions relating to (fund) investor behaviour. In academic research, such questions are more likely to arise in the finance literature than in the international economics literature. Additionally, fund flow data may be useful to investment professionals for informing asset allocation decisions.

By contrast, IIF data seem most appropriate for analysing portfolio flows in a macroeconomic and external financing context, which are more common in the international economics literature than in the finance literature. IIF data may also be preferable for answering most policy-related questions and/or for undertaking country-level analysis.

In terms of real-time tracking, the paper conducts a quantitative assessment of how well EPFR and IIF data predict official BoP data. The sample period for the nowcasting exercise runs from 2010:Q1 until 2019:Q2. The results are shown in Figure 1.

In greater detail:

• Daily IIF data and weekly EPFR data are available as early as two and seven days into the quarter, much earlier than either of the monthly proxies. Forecast performance was therefore relatively poor to start with, but the daily IIF data improve notably once 20-30 days’ worth of data are included.
• The monthly proxies first become available about 33 and 45 days into the quarter for IIF and EPFR data, respectively. The IIF’s monthly tracker performs better than either of the EPFR datasets (weekly and monthly), but only outperforms the IIF daily flows data after the second data release (third month).

The Alternative IMF Monthly Portfolio Flow Dataset

There is one key benefit of high-frequency data for empirical research, and that is to facilitate event studies (i.e. what happened to financial flows at the onset of the Fed’s pandemic response). The move entails some downside risk: research has shown that fund flow data are more affected by push factors. This biases empirical findings towards greater importance of push factors. Private data providers are subscription-based and thus costly.

The IMF addresses this gap by providing a monthly portfolio flow dataset (KP dataset). The KP dataset is constructed using data from national sources in a set of 19 EMs, with data for aggregate portfolio flows for 19 countries and debt and equity flows for 18 countries. Data availability begins in 2010 for most countries. Of the 19 countries, 13 report monthly data that correspond to the quarterly BoP published by the central bank or statistical authority. For six countries, they collect monthly proxies for one or more portfolio flow type.

On the datasets performance, the paper compares total EM equity and debt flows for the KP dataset to three main comparators (IIF monthly tracker, IIF daily, EPFR monthly). Equity data from the KP dataset and the IIF’s monthly portfolio flow tracker are highly correlated with quarterly BoP-based equity flows. For the debt component, the KP data has a high correlation with BoP flows at the country level, but somewhat lower at an EM aggregate level.

Source: Page 24 of ‘Koepke and Paetzold (2020)’

The results from the horse race also show that the KP dataset has a similar performance to the IIF’s monthly tracker (although the monthly data from countries contained in the KP dataset is subject to a greater release lag within the current quarter). Its accuracy may also be linked to regular data revisions, which the IIF and EPFR do not benefit from.

The Bottom Line

The move towards high-frequency capital flow proxies is evident in academia as well as in the policy world. The data is shown to be good, in the sense that it has predictive power. It is, however, costly and contains biases. So which dataset should you choose? If monthly capital flow data suffice for your research, use the free IMF KP dataset. If you are a financial professional, consider using fund flow data (from either EPFR, Lipper Fund Flows and Trounceflow, State Street and BNY Mellon). If you are policymaker looking for reliable high-frequency data that predicts actual capital flows well, IIF data seems to be the best option.

Capital Flow Dataset Links:

• IMF Balance of Payments Statistics + Guide
• IMF Coordinated Direct Investment Survey
• BIS Locational Banking Statistics + Consolidated Banking Statistics + Debt Securities
• IIF Capital Flow + EM Capital Flows
• Emerging Portfolio Fund Research (EPFR)
• ECB Security Holdings Statistics
• Morningstar Investment Fund
• UNCTAD FDI
• SWIFT

Capital Flow Datasets By Country:

• Canada (Portfolio Investment – monthly) (Direct Investment – quarterly flow, annual stock)
• Germany (Portfolio Investment – quarterly) (Direct Investment – quarterly)
• Japan (Portfolio Investment – quarterly) (Direct Investment – quarterly)
• United States (Portfolio Investment – monthly stock + flows) (Direct investment – quarterly flow, annual stock)
• Brazil (Direct Investment)
• Indonesia(Direct Investment)
• Russia (Portfolio Investment) (Direct Investment)
• Korea (Portfolio Investment) (Direct Investment)
• Turkey (Portfolio Investment) (Direct Investment)
• Austria (Portfolio Investment) (Direct Investment)
• Denmark (Direct Investment)
• Netherlands (Direct Investment)
• Spain (Portfolio Investment) (Direct Investment)

A Primer on Balance of Payments:

A country’s balance of payments accounts keep track of both its payments to and its receipts from foreigners. Any transaction resulting in a receipt from foreigners is entered in the BoP accounts as a credit. Any transaction resulting in a payment to foreigners is entered as a debt. Three types of international transactions are recorded in the BoP:

1. Transactions that arise from the export or import of goods and services and therefore enter directly into the current account.
2. Transactions that arise from the purchase or sale of financial assets. These transactions are recorded in the financial account of the BoP.
3. Certain other activities resulting in transfers of wealth between countries are recorded in the capital account. These are typically very small.

The fundamental balance of payments identify is as follows:

Current Account (1) + Capital Account (3) = Financial Account (2)

For the purpose of this Deep Dive, we focus on the financial account. Many common misconceptions lie in the two-way nature of international financial flows that make up this account. The financial account measures the difference between acquisitions of assets from foreigners (inward investment) and the build-up of liabilities to them (outward investment). Both inward and outward investment flows can be positive or negative, and the two can be netted against each other to obtain a ‘net’ measure of capital flows. We can zoom in on the financial account balance in the following figure.

Source: Page 15 of ‘Koepke and Paetzold (2020)’

The IMF paper outlines three sources of confusion:

1. Net vs Gross: The net change in liabilities is a ‘net’ measure of capital flows (netted against FDI, portfolio flows etc.) but is a ‘gross’ measure of capital flows in the sense that changes in liabilities are not netted against changes in assets.
2. Sign convention: Change in the sixth edition of the IMF’s BPM in 2009 – an increase in foreign assets (including reserves) is shown with a positive sign.
3. Components of capital flows: There are many, but the main components are FDI, portfolio flows and other investments. The term ‘capital flow’ is also confused with ‘fund flow’ (net purchases of investment fund shares).

To view the full paper, please click here.