Unconventional Monetary Policies in Emerging Markets and Frontier Countries

Contributor Notes

Author’s E-Mail Address: cfratto@imf.org; bmircheva@imf.org; ddepadua@imf.org; hpoirson@imf.org; brendan.vannier@psemail.eu
Publication Date:
22 Jan 2021
eISBN:
9781513567211
Language:
English
Keywords:
Unconventional monetary policy; emerging markets; COVID-19; local currency bond markets.; WP; market dysfunctionalities; app announcement; CB transparency; market functioning; markets query; tackling market dysfunctionality; transparency Index
Download PDF (3.7 MB)

The COVID-19 crisis induced an unprecedented launch of unconventional monetary policy through asset purchase programs (APPs) by emerging market and developing economies. This paper presents a new dataset of APP announcements and implementation from March until August 2020 for 27 emerging markets and 8 small advanced economies. APPs’ effects on bond yields, exchange rates, equities, and debt spreads are estimated using different methodologies. The results confirm that APPs were successful in significantly reducing bond yields in EMDEs, and these effects were stronger than those of policy rate cuts, suggesting that such UMP could be important tools for EMDEs during financial market stress.

Abstract

The COVID-19 crisis induced an unprecedented launch of unconventional monetary policy through asset purchase programs (APPs) by emerging market and developing economies. This paper presents a new dataset of APP announcements and implementation from March until August 2020 for 27 emerging markets and 8 small advanced economies. APPs’ effects on bond yields, exchange rates, equities, and debt spreads are estimated using different methodologies. The results confirm that APPs were successful in significantly reducing bond yields in EMDEs, and these effects were stronger than those of policy rate cuts, suggesting that such UMP could be important tools for EMDEs during financial market stress.

I. Introduction1

Even before experiencing their own outbreaks of COVID-19, emerging markets and developing economies (EMDEs) were affected adversely by the economic and financial fallout of the pandemic in March 2020. A marked increase in global investors’ risk aversion combined with lower commodity prices translated into a rise in local currency bond yields, capital outflows, and sharp currency depreciations. EMDEs responded to the COVID-19 crisis with a combination of conventional and unconventional measures. In particular, EMDEs introduced their first unconventional monetary policy (UMP) measures in the form of asset purchase programs (APPs), similar to those used by advanced economies (AEs) during the global financial crisis (GFC).2

In contrast to small open AEs (Sweden, Switzerland) during the GFC, which deployed UMP at or near the effective lower bound in the context of capital inflows and exchange rate appreciation, most EMDEs in our sample had policy rates above zero and faced external pressures when they launched APPs. In a majority of cases, the programs aimed at affecting the sovereign yield curve, easing stress, and bolstering liquidity of the targeted financial markets. By stabilizing the sovereign secondary bond market and easing broader financial conditions, the APPs improved funding conditions for governments; however, only about a quarter of the EMDE programs (and less than a fifth of small open AE programs) announced support for pandemic-related fiscal needs as a main objective.

While the empirical literature evaluating the effects of UMP in AEs is extensive, and focuses on both financial factors (e.g., term premia) and macro factors (e.g., output and inflation expectations) few studies exist on the effectiveness of UMP in EMDEs —reflecting in part the novelty of these measures. The existing studies focus on the effects of UMP on select financial market variables (e.g., long-term bond yields and exchange rates), and do not explore effects on other financial market variables or second-round effects on related markets. For example, Arslan et al (2020)—using an event study approach for 13 EMs—find that central bank bond purchases on average reduced benchmark (long-term) bond yields in a significant and persistent manner and interrupted depreciation trends. The results are confirmed by panel regressions controlling for confounding factors. Other studies, using somewhat different samples, suggest greater effects of EM APP policy announcements on sovereign bond yields than in AEs (Hartley and Rebucci (2020)), and that the effect on the exchange rate is relatively small or insignificant (International Monetary Fund (2020)).

Against this background, this paper first develops a novel and extensive database of APP announcements and implementation in EMDEs. The database aims at illustrating the recent experience with such UMP schemes in EMDEs, including their objectives and modalities, and at helping to draw some initial lessons from these policy experiments. Specifically, the database contains recent COVID19-related UMP measures for 27 EMDEs, highlighting different characteristics including: (i) the nature of the program—purpose, size, targeted market, etc.; (ii) whether the announcement were made jointly with other authorities or coincided with other central bank policy announcements; and (iii) transactions data where publicly available. The same information is also collected for 8 small AEs, for comparison.

The information collected in the database on UMP suggests that a number of countries carried out several APPs during the period under consideration (January-August 2020), or tweaked their programs, with programs exhibiting different characteristics depending on the objective. Most of the interventions aimed at boosting confidence and improving market functioning, and targeted the government bond market, although a few countries (BEAC, Brazil, Chile, Ethiopia, Hungary, Israel, Korea, Mauritius, and Norway) aimed at the corporate or bank bond market. Egypt is the only country that purchased equities. The size of the programs in EMDEs was comparable to that of AEs in our sample. About two thirds of these programs were quantity-based (fixed or maximum amount of purchases). Examples include Bolivia, Ethiopia, Ghana, Iceland, India, Mauritius, Mexico, New Zealand, Norway, and Thailand. Other programs were also quantity-based, but more flexible, with purchase amounts calibrated to market conditions and/or the economic and inflation outlook (e.g., Angola, Canada, Colombia, Costa Rica, and Hungary). Except for Chile (price-based APP targeting bank bonds), none of the programs were price-based.

Building and expanding on existing studies, we use the database to empirically analyze the effects of these APPs on financial market variables for 15 emerging markets and 8 small AEs. First, we look at the distribution of sovereign bond yields and exchange rates, following previous studies; and also at the distribution of equity prices, corporate bond yields, and EMBI spreads. For each variable and country, we investigate whether there is a different impact response one, two, and three days following announcements. These effects are also estimated for conventional monetary policy actions for comparison. Second, we use country-specific regressions to investigate whether other policies announced at the same time as the APP, external factors, and the COVID-19 pandemic’s impact on activity alter the impact of the announcements. The results help us identify the main channels of impact as well as whether the effect of UMP is different from that of conventional policy easing. Finally, we test whether the main findings hold when we employ a panel regression, controlling for various factors. The analysis covers the period January through August of 2020.

The empirical results for EMDEs confirm a negative and statistically significant multi-day effect on bond yields. However, the magnitude and persistence depend on the focus of the APP, i.e. whether the program was quantity- or price-based as well as whether it involved purchases of government bonds or private securities, or both. The results still hold when we exclude APP announcements that coincide with policy rate cuts as well as when the various controls are added and in a panel regression setting. Similarly, conventional monetary policy measures (policy rate cut announcements) have a negative and statistically significant multi-day effect on bond yields across the different maturities but of a smaller magnitude than that of APP announcements. For AEs, the results are broadly consistent with the findings in the literature.

This paper is organized as follows. Section II provides a brief overview of the motivation for the use of APPs by EMDEs and discusses their channels of transmission and findings in the literature on their effectiveness. Section III mentions the main elements of the database while section IV provides some empirical findings on the effectiveness of APP announcements. Section V concludes.

II. Motivation for Apps and Previous Findings on Effectiveness

This section discusses conditions facing EMDEs and small AEs at the onset of the COVID outbreak, that prompted the deployment of UMP tools, followed by a discussion of the theoretical channels of transmission, and a summary of the literature findings on effectiveness of such unconventional measures.

A. Motivation for APPs in EMDEs3

As the COVID pandemic intensified with a corresponding rise in investor risk aversion, a number of countries in our sample faced pressures. Specifically, Figure 1 shows, for 14 EMDEs and 7 small AEs, the number of markets facing large fluctuations of the 10-year government bond yield and the exchange rate. The number of countries facing stress in both markets increased sharply at the beginning of March. Shortly after stress peaked, in mid-March, the monetary authorities of many of those economies started announcing APPs.

Figure 1.
Figure 1.

Stress in Financial Markets, January–July 2020

Citation: IMF Working Papers 2021, 014; 10.5089/9781513567211.001.A001

Note: A market is defined as highly stressed when the 1-day price change is in the tail of the distribution (2.58 times the historical standard deviation above the historical average, computed over 2017–2019). Thresholds for mid and low stress are equal to 1.96 and 1.64, respectively.Sources: Thomson Reuters and IMF staff calculations.

Overall, EMDEs in our sample started announcing UMP measures even while most of them still had room for policy rate cuts, and relatively low public debt levels. Figure 2 (left panel) shows policy rates for 24 EMDEs at the beginning of March 2020. The countries that announced APPs are represented in orange. All countries faced inflation close to or above target except for Thailand. The same countries except India, Hungary, and South Africa entered the crisis with relatively low public debt levels—similar to the majority of EMDEs— and announced fiscal easing measures in response to the COVID-19 shock (Figure 2, right panel).

Figure 2.
Figure 2.

Policy Rate and Public Debt Levels Pre-COVID-19 Crisis in EMDEs

Citation: IMF Working Papers 2021, 014; 10.5089/9781513567211.001.A001

Note: Orange dots indicate the countries included in our database, blue dots indicate all other EMs.

B. Transmission Channels from Economic Theory

In normal times, when the central bank (CB) implements conventional monetary policy (CMP), changes in the policy rate affect prices. Consequently, financial market participants rebalance their asset portfolio, thus diffusing the effect of the CMP throughout the economy. However, as the COVID19 pandemic hit the world economy, financial markets participants, fearing uncertainty, demanded higher risk premia or fled some markets altogether. This behavior impaired the normal functioning of markets and hindered the transmission mechanism of monetary policy, preventing CMP’s intended effect on the aggregate economy. Figure 3 offers an illustration of the theoretical transmission of CMP and UMP (in the form of APPs) on the aggregate economy.

Figure 3.
Figure 3.

CMP and UMP Transmission to the Economy

Citation: IMF Working Papers 2021, 014; 10.5089/9781513567211.001.A001

Source: Staff analysis.

EMDEs introduced APPs to offset financial markets’ impairments and/or provide direct support to the economy. Based on the experience of AEs, the literature has identified five main mechanisms through which these measures impact the economic and financial environment. One of these channels is direct. Specifically, as APPs entail a change in the liquidity available on the targeted market, the equilibrium price will need to adjust to the new supply and demand conditions. The other four channels of APP transmission are indirect. The two most prominent ones are the portfolio-rebalancing and signaling channels. Through the portfolio-rebalancing channel, market participants who benefit from the intervention reallocate newly found liquidity across their asset portfolio. If markets are well interconnected, the initial impact can spill over and ease liquidity pressures in other financial markets. Through the signaling channel, as CBs announce APPs, they also signal and clarify their objectives to market participants. Given the novelty of most APPs, these announcements surprise financial markets and the discussion or commitment to future interventions by the CB helps coordinate agents’ expectations about future short-term interest rates in uncertain times. Both channels affect long-term interest rates.

APPs may also lead to increased liquidity and to exchange rate depreciation. Through the liquidity channel, APPs affect the economy indirectly when banks and credit establishments see their liquidity constraints reduced and as a result increase credit supply. To the extent that asset purchases have contributed to a change in interest rate differentials, or that they targeted foreign-currency denominated assets, an exchange rate channel of UMP can also be observed. The exchange rate repercussions of UMP in turn can affect its effectiveness, especially in EMDEs where exchange rate fluctuations are a primary source of volatility and vulnerability. If this exchange rate channel is strong, the cost of funding for domestic participants might not fall as expected, thus countering the initial impact of the program.4

With these transmission channels in mind, the impact and effectiveness of APPs might differ across countries. This may be due to underlying structural features. For example, if the CB is more credible or if there is a larger foreign investor participation in the local currency bond market, the confidence-restoring signaling effect of the announcement will be stronger. If a country’s financial markets are more interconnected, the portfolio-rebalancing channel might be more efficient. In financial markets that are more exposed to foreign participants, the exchange rate channel could be of sizable importance.

C. Effectiveness Results from the Literature

Bhattarai and Neely (2020) provide an extensive overview of the literature on effectiveness of UMP, mostly focusing on AEs. Therefore, this section only mentions a few key findings from previous research. In a first strand of the literature, the majority of the findings—based on the use of event study methodologies—confirm that UMP influenced asset prices and risk premia as would be expected. A number of studies illustrate this for the US (Gagnon et al (2011), Vissing-Jorgensen and Krishnamurthy (2011), and Swanson (2011)), showing a decrease in the 10-year bond yield of about 0.2 percent after the announcement of the measure. Briciu and Lis (2015) look at seven balance sheet policies launched by the ECB between 2008 and 2015 and find a cumulative two-day effect on the 10-year bond yield ranging from 33 to -17 bps. Roache and Rousset (2013) use an event study as well to show that “tail risk” diminishes in the immediate aftermath of UMP events, particularly downside left tail risk.

In a second strand of the literature on effects of UMP in AEs, calibrated modeling and vector auto-regressive (VAR) exercises are used to show that these unconventional macroeconomic policies improved macroeconomic outcomes as well. A number of studies illustrate this relationship for advanced economies (MacDonald and Popiel (2017), Gambacorta et al (2013), Weale and Wieladek (2016)). The effects range from 0.25 to 0.62 percent for estimates of the peak responses of real GDP and inflation. As the macroeconomic effects are difficult to disentangle between conventional and unconventional periods, Wu and Xia (2016) use their shadow rate to show that macroeconomic effects of an unanticipated change in the policy instrument are similar to the pre-ELB period.

Only a few studies focus on the impact of UMP in EMs. Hartley and Rebucci (2020) and Arslan et al. (2020), both based on event study approaches, find that on average, bond purchases reduced benchmark bond yields. Hartley and Rebucci (2020) find an average response of -0.28 to -0.43 percent of the 10-year bond yield in EMs, which is higher than that for developed markets (-0.11 to -0.14 percent). In addition, Arslan et al (2020) find that APPs appear to have shored up the exchange rate on average, although the results vary by countries. Some of the event studies in the literature also use control variables (Rai and Suchanek (2014); Falagiarda, McQuade, and Tirpak (2015); and IMF (2016)) to investigate the effectiveness of UMP. They find that, in general, the portfolio rebalancing and signaling channels play a key role. Last, Sever et al. (2020) builds on the literature by expanding the country sample, accounting for multiple APP announcements for each country, and introducing controls for cost of borrowing and various global factors. The results show that APP announcements had a significant impact on bond yields and helped turn around sentiment but did not lead to depreciation of emerging market currencies.

III. A New Database on UMP in EMDES

This paper presents a new and detailed database on recent UMP announcements in emerging market and developing economies and the realized transactions data (when available).5 Specifically, we collect detailed information on 50 asset purchases programs announced between March and August 2020 in a total of 27 EMDEs and 8 small advanced economies (the latter, included as a benchmark for comparison). Table 4 includes the list of countries in the database and summary statistics.

Table 1.

Availability of Transaction Data

article image
Source: Staff analysis.
Table 3.

Policy Measures Introduced by CBs on the Same Day as APP Announcements

article image

Green denotes introduction of the policy at the time of announcement of an APP. Yellow denotes no contemporaneous introduction of the policy.

Source: Staff analysis.
Table 4.

Summary of Asset Purchase Programs

article image
article image
article image
article image
article image

The initial sample of countries has been selected from the policy announcements recorded as APP in the IMF Central Bank Intervention Database (CBID), the Yale tracker, and the IMF COVID policy tracker.6 We drew on central bank press releases as the primary source of information for the database, including for information on implementation so far.7

The dataset consolidates the existing information dispersed in several other databases and broadens their coverage to a longer time period and more countries. In addition, it seeks to extract and present the real-time flow of information that central banks provide to market participants when they announce, augment, change, or implement APPs. To this effect, the dataset includes: (i) both the initial program announcement and all subsequent announcements; (ii) implementation dates and related published information (to allow separation of announcement effects from implementation effects); (iii) broader time and country coverage; and (iv) the information provided by the central banks and/or other authorities at the time of the announcements and implementation. Specifically, we identify the CB’s communication style and how the information is shared in subsequent announcements, whether the announcement was made jointly with other national authorities and was part of a policy package, etc.8

The database covers all CB purchases (or sales) of private and/or public securities on the primary or secondary markets. Such APPs often result in permanent changes to the CB’s balance sheet. However, the database also includes information on measures such as (i) twist operations to purchase long-term and sell short-term government securities that do not result in expansion of CB balance sheets, but affect the yield curve (Colombia, India, and Mexico), (ii) the establishment of special purpose vehicles (Korea, Mauritius) or investment funds (Norway9, Thailand10) to purchase equities and other private securities, (iii) direct monetary financing of the government (one-off contribution made by the Bank of Mauritius to the government), and (iv) purchase of loans to SMEs by the People’s Bank of China.

Overall, this new dataset allows for much more granular and operational analysis of the types of schemes being applied, including their modalities and communication aspects and whether these make a difference in impact (Box 1). The next sections present the timeline of announcements of APPs, followed by a detailed summary of their key features based on information in the database.

A Tale of Two Central Banks: India and South Africa

The cases of India and South Africa help to highlight the diverse approaches taken by EMDE central banks in announcing and implementing UMP. Figure 3 presents the yields of sovereign bonds across different maturities. The vertical bars represent the dates at which the CBs announced the APP or provided new information. In both cases, purchases took place in secondary markets, were quantity-based, and there were no joint announcements with other national authorities. In the case of India, the second press release coincided with the implementation of a previously announced purchase (second darker green vertical bar).

Both countries faced high sovereign bond yields in mid-March. In India, they had reached a historic high, whereas in South Africa, a 3 percent upward shift of the yield curve coincided with the onset of the pandemic. In response and with the objective of shoring up market confidence and tackling market dysfunctionality, both CBs announced APPs of government securities. However, their approaches were rather different.

The South African Reserve Bank (SARB) issued only one press release on March 25th. The press release gave no details on implementation (size, frequency, preannouncement, target maturities) of the program, but stated clearly that the new measure did not signal changes in the future monetary policy stance. The next day, the SARB published a Q&A to answer markets queries and detailing heavily the APP objectives. The SARB also detailed how this program related to other recently introduced policies and specified that the program was neither direct financing nor debt monetization nor quantitative easing, and again did not represent a shift in the CB’s mandate. Bond yields across all maturities were lower after the announcement.

The Reserve Bank of India (RBI) published numerous daily press releases related to the UMP, including diverse policy announcements, but (unlike the SARB) no detail on how it fit in with other policies. Instead, the press release remained factual and focused on the practicalities of the program. Implementation was on a discretionary basis, preannounced two days ahead. On March 18th, the RBI announced a first open market operation involving government dated securities. The quantity-based program implemented two days later received a “positive response”. The RBI announced two 1.5-bigger tranche purchases on March 24th and 30th, covering longer maturities. On March 23rd, the second purchase (announced on March 30th) was later advanced by four days (to March 26th). On April 23rd, the RBI announced a new type of “twist program” that replaced the existing program. The “twist program” included the simultaneous purchase of long-term securities and sale of short-term maturities, to be implemented two days later. On June 29th, the RBI announced another “twist”-like intervention to be made two days later in a similar amount, targeting slightly longer maturities. Overall, the APP announcements had only a slight effect on the bond yields at medium- and long-term maturities. Regarding the maturity of the assets, the RBI changed its target repeatedly over this period, initially purchasing 2 to 5 years bonds, then 2 to 9 years; and when the twist operation was introduced, sold securities with 2 months to 1 year maturity and purchased those with 6 to 10 years maturity, and subsequently sold securities with 3 to 10 months maturity and purchased those with 7 to 13 years maturity.

These two examples highlight both the diversity of country approaches to UMP and the importance of policy communication and other operational aspects. Some aspects that appear relevant from these two cases were frequency of announcements, whether the UMP was part of a policy package or not, and how much detailed information was included on the program (size, maturity of the securities purchased, pre-announced frequency and dates of implementation, counterparties), type and length of commentary on the rationale of the policy and its implications for the CB and the CB monetary policy stance, as well as differences in innovativeness (swap) and implementation (expansion of existing program or new program).

A. Timeline of APP Announcements

This section highlights the timeline of the asset purchase program announcements recorded in the database.

APPs among EMDEs were announced mostly around the second half of March. Most of the APPs involved purchases of sovereign bonds (Figure 4).

Figure 4.
Figure 4.

APP Announcements by EMDEs by Type of Securities Purchased

Citation: IMF Working Papers 2021, 014; 10.5089/9781513567211.001.A001

Source: Staff analysis.

Indonesia was the first, on March 2nd, 2020, to announce the start of sovereign bond purchases in order to “maintain monetary and financial market stability, including mitigating the risk of COVID-19.”11 The program was announced as part of a broader policy package including a reduction in reserve requirements, FX intervention, and repo operations. Mexico followed on March 12th, Croatia on March 13th, and then many other EMDEs in the second half of March. AEs also made similar announcements around the same time. While the U.S. Federal Reserve was establishing the Secondary Market Credit Facility (SMCCF, March 18th), involving the purchase of private securities, small advanced economies including Israel, Norway, Sweden, New Zealand were doing the same thing.

The last announcement in our sample was made at the end of June by the Brazilian central bank. It required a change in the CB mandate in order to conduct such interventions and clarified that this was “yet another available instrument, with no obligation to use.” Other CBs required a change in the mandate as well in order to be able to conduct APPs.

EMDEs that announced APPs so far are very diverse in terms of their policy frameworks (Figure 5). The figure classifies countries according to three framework dimensions, monetary framework (inflation targeting (IT) vs non-IT), exchange rate framework (floating vs non-floating), and central bank transparency, and shows the share of EMDE countries announcing APPs in each country group. So, for instance, there is only one IT country among all EMDEs, with high level of central bank transparency and with a non-floating exchange rate, Indonesia, which announced an APP, so the share of EMDEs in that country group announcing APP is 100 percent. Most of the countries conducting APPs (58 percent) are classified as having inflation targeting (IT) frameworks and floating exchange rate regimes. By contrast, only 8 percent of the countries that announced an APP did not have an IT framework and floating exchange rate regime (Angola, Bolivia, Capo Verde, China, Croatia, Egypt, Ethiopia, and Papua New Guinea). Indonesia and Costa Rica pursue inflation targeting monetary policies but have respectively a stabilized exchange rate arrangement and a crawl-like arrangement. Mauritius has a form of “hybrid inflation targeting.”

Figure 5.
Figure 5.

APP Announcements by EMDEs and Policy Frameworks

Citation: IMF Working Papers 2021, 014; 10.5089/9781513567211.001.A001

Note: The Central Bank Transparency Index is obtained from the Dincer/Eichengreen/Geraats revised Central Bank Transparency Data Set for 1998–2015. High central bank transparency is defined as those countries with above EMDE-average transparency index. Information on the IT and ER frameworks is obtained from the 2017 IMF Annual Report on Exchange Arrangements and Exchange Restrictions Database (AREAER). The EMDE average is computed over the sample of all EMDEs for which data is available.

The transparency of the central bank is correlated with the choice of starting an APP, as 41 percent of EMDEs with above average transparency of the central bank started APPs. 12 The fraction is even higher among those who have above average transparency of the central bank and high non-residential investment share (50 percent). Countries with low CB transparency were less likely to announce an APP: only Jamaica among the countries with high non-residential investment share and low CB transparency announced an APP; and 17 percent of those with low non-residential investment share and low CB transparency announced an APP.

B. Implementation

Information on implementation of the APPs is shown in Table 1, which lists the APPs for which transaction data (realized asset purchases) are publicly available (nearly all programs in AEs and less than a third of programs in EMDEs in our sample), the level of aggregation, and the number of observations.

Figure 6 shows the size of the programs, in US$ billion and percent of GDP. Some are sizeable relative to GDP (Bolivia, Iceland, and Mauritius) or in US$ terms (Australia, Colombia, Israel, and Poland) or both (New Zealand).

Figure 6.
Figure 6.

Size of Announced and Realized CB Interventions

Citation: IMF Working Papers 2021, 014; 10.5089/9781513567211.001.A001

Among countries for which the database includes both the announced size and the realized purchases as of August 2020, the latter are often smaller than announced, suggesting CBs still have room for more asset purchases in the near future (Figure 6). 13 14 15 For example, the Chilean central bank made purchases in March 2020 amounting to US$ 2.5 billion and it has not recorded any transaction ever since, despite initially announcing a US$ 4 billion program, and then doubling it to US$ 8 billion in April. Some other CBs are still conducting APPs and publishing transactions data as of August (e.g., India, Croatia, and Hungary).

Central banks in EMDEs announced and implemented APPs quickly (Figure 7).16 Croatia implemented its APP on the same day as the announcement, while Poland started the implementation three days after the program was officially mentioned by the NBP management. The speed of implementation is comparable, if not faster, to what we observe among AEs, where implementation ranged from one day since the first announcement (Australia, Korea), to nine days (Sweden).

Figure 7.
Figure 7.

Implementation of APPs Over Time 1/

Citation: IMF Working Papers 2021, 014; 10.5089/9781513567211.001.A001

1/ Light green are days in which the APP has already been announced but implementation has not started yet. Dart green are days in which APP implementation has started. Korea includes only the program involving the purchase of sovereign bonds.Source: Staff analysis.
Figure 8.
Figure 8.

Realized Size of Interventions

Citation: IMF Working Papers 2021, 014; 10.5089/9781513567211.001.A001

Source: Staff analysis.

C. Key Features of the APPs

Stated Objectives

Concerns related to market dysfunctionalities and the need to restore confidence are among the top objectives mentioned by CBs in EMDEs when announcing these new programs (Figure 9).17 In line with their initial conditions, only in 9 percent of the cases the objective of the APP in EMDEs is to provide monetary stimulus (Bolivia, Angola, Costa Rica, Mauritius).This is in line with stated objectives for APPs in AEs.18 17 percent of the programs mention explicitly the need to alleviate the costs of COVID-19 outbreak on the population as their main objective (Angola, Bolivia, Cabo Verde, Indonesia, Mauritius, Poland, Rwanda). Supporting the fiscal needs is stated as an objective in 10 percent of the cases (Ghana, Indonesia, Mauritius). Poland mentions the need to “mitigate the negative impact of Coronavirus spread,” when referring to the whole policy package. 19 Angola mentions the need to mitigate the “possible impacts of the COVID-19 pandemic on the national economy, with particular attention to the external accounts, and its implications in the conduct of the monetary and exchange rate policies” and “to provide direct support to the productive sector and to ease the pressure on the cash flow of these companies, major employers, allowing them to continue operating and, consequently, maintain jobs.” 20 Other countries, instead, mention the budget financing as a goal of the program: Indonesia declared that the APP was “a source of funding for the government in the context of national economic recovery”, and they saw it “as the last resort in the event that the market capacity is unable to absorb and/or causes a yield increase that is too high.” 21 Mauritius stated that the APP had the goal of “assisting the government in its fiscal measures to stabilize the economy.22 “The central bank in Ghana declared in their press release dated May 15, 2020, that “under the circumstances […], the Bank of Ghana has triggered the emergency financing provisions, which permits the Bank to increase the limit of BOG’s purchases of government securities in the event of any emergency to help finance the residual financing gap.” 23 Indonesia is the only country declaring the exchange rate to be one of the objectives of the APP, while also mentioning market dysfunctionalities and support for fiscal needs among the objectives. Specifically, the Indonesian central bank introduced the APP in order “to minimize the risk of increasing volatility in the Rupiah exchange rate” and as part of a broader intervention to ensure “that the Rupiah exchange rate moves according to its fundamentals and follows market mechanisms.”

Figure 9.
Figure 9.

Stated Objective of APP Announcements in EMDEs

Citation: IMF Working Papers 2021, 014; 10.5089/9781513567211.001.A001

Source: Staff analysis.

Except in two cases (Turkey and BEAC), APP announcements in EMDEs coincided with other policy measures by CBs (Table 2 and Figure 10). For AEs, by contrast, there were more cases of stand-alone APP announcements (Iceland, Korea, New Zealand, Norway).24 In particular, APP announcements coincided25 with policy rate cuts in the case of Egypt, Philippines26, Chile, Poland, Hungary, Romania, Capo Verde, Indonesia, Mexico, Uganda, and Papua New Guinea. This table does not include the instances in which the central bank cut its interest rates in days in which there were no APP announcements.

Figure 10.
Figure 10.

APP Stated Objectives and Other Policy Measures 1/

Citation: IMF Working Papers 2021, 014; 10.5089/9781513567211.001.A001

Source: Staff analysis.1/ Change in CAPB is the difference between cyclically adjusted primary balance in 2019 and 2020.

Modalities

While more than half (17) of the EMDE CBs announced interventions exclusively in the sovereign market, there are important exceptions: Brazil, Chile, China, Egypt, Ethiopia, and Thailand envisaged purchases of private securities; Colombia, Hungary, Mauritius, and Mexico announced programs involving both private and public securities. For instance, the Chilean central bank is not legally allowed to purchase sovereign bonds, and therefore only purchased bank bonds. Egypt purchased equities to “support asset prices amid sustained market volatility.” Hungary announced contemporaneously the purchase of sovereign bonds and mortgage securities. A month after starting the intervention in the sovereign bond market, Mexico announced a new APP targeting private securities, “to promote an orderly behavior of Mexico’s debt market.”27

The majority of countries among both EMDEs and AEs are conducting APPs exclusively in the secondary markets (Figure 11). Among the exceptions of APPs conducted in the primary markets are Bolivia, Indonesia, Ghana, Cabo Verde,28 Thailand, and Norway. Bolivia announced simultaneously the purchase of US$ 1 billion in sovereign bonds on the primary market, and of US$ 500 million on the secondary market, to “preserve the lives of Bolivian people and avoid the crisis state experienced by more developed countries, who delayed measures.” After a Presidential decree initially announced an APP of sovereign bonds on the primary markets, the CB of Indonesia announced the beginning of purchases on the primary market, as a “source of funding for the government in the context of national economic recovery, including maintaining the sustainability” of the government. The Norwegian government reinstated the Government Bond Fund, initially established in 2008, to purchase in the primary and the secondary market bonds issued by companies headquartered in Norway with credit rating above CCC-. South Africa clarified in their announcements that their APPs in secondary markets did not constitute direct financing.29

Figure 11.
Figure 11.

Asset Purchase Programs by Primary vs. Secondary Market Purchases

Citation: IMF Working Papers 2021, 014; 10.5089/9781513567211.001.A001

1/ Percentage of programs.Source: Staff Analysis.

Quantity-based programs prevail among both EMDEs and AEs (Figure 12). An exception is Chile, which announced a price-based APP targeting bank bonds (Swap de Tasas Promedio Cámara or SPC plus 2.5 percent for AAA-rated bonds, SPC plus 2.6 percent for AA-rated bonds, and SPC plus 2.8 percent for A-rated bonds). Australia pursued a similar approach and announced a target for the yield on government bonds of around 0.25 per cent.

Figure 12.
Figure 12.

Asset Purchase Programs by Quantity-Based vs. Price-Based 1/

Citation: IMF Working Papers 2021, 014; 10.5089/9781513567211.001.A001

1/ Percentage of programs.Source: Staff analysis.

While most CBs do not mention specific maturities, EMDEs are overall more likely than AEs to specify the maturities targeted (Figure 13).30 19 percent of the programs among AEs (Canada, Korea) targeted short-term maturities, compared to 3 percent among EMDEs: among EMDEs, Colombia is the only CB targeting short-term securities (up to three years, private securities). No AE targeted long-term securities, while 11 percent of the CBs among EMDEs did. Many countries in our sample operated APPs on the broad yield curve. For example, Hungary’s APP encompassed all private securities and sovereign bonds of three year or higher duration. Finally, some countries which conducted APPs along the broad yield curve started swap programs, selling bonds of short maturity and purchasing those of long maturity (Colombia, India, Mexico).

Figure 13.
Figure 13.

Maturity of the Securities Purchased under the APPs 1/

Citation: IMF Working Papers 2021, 014; 10.5089/9781513567211.001.A001

1/ Percentage of programs.Source: Staff analysis.

Other operational aspects: coordination and communication

While most central banks made announcements independently of other national authorities, some instances of coordination of the CB with the government occurred across all income groups. For example, in Egypt, the President announced the APP and no information was provided or announced directly by the central bank. Among EMDEs, joint announcements were made in Thailand, Mexico, China, and Indonesia. In the case of Indonesia, the introduction of the APP on March 31st was the result of a Presidential decree, and it was jointly announced with the Ministry of Finance. In Ethiopia, the announcement was made by the Prime Minister. Among AEs, the cases of Korea and Norway are worth mentioning. In Korea, the central bank set up an SPV with joint support from government-run financial institutions to purchase private securities. In Norway, the APP was conducted by the government rather than the central bank.

The extent of information made available by CBs when introducing an APP varies across the sample and is higher for advanced economies than for EMDEs. For each announcement in our database we collected up to 27 variables on the description of the APP at the moment of the announcement, ranging from the objective, the size of the program, the implementation timeframe, etc.31 The number of variables collected is a rough indicator of the communication strategy implemented by the different central banks.32 Canada’s central bank scores the highest on this metric of transparency and communication, providing information on all 27 variables collected in our database. At the opposite extreme, Romania provided information on 12 variables. However, it should be noted that this list excludes those countries for which we were not able to collect a minimum amount of publicly available information on the programs. Overall, by comparing the distributions on the amount of information collected between central banks of AEs and EMDEs (Figure 14), the latter on average provide less comprehensive information. Exceptions include Colombia, India, and Mexico, ranking respectively first, second, and third among EMDEs in the extent of information provided.33

Figure 14.
Figure 14.

Information Available in APP Announcements 1/

Citation: IMF Working Papers 2021, 014; 10.5089/9781513567211.001.A001

1/ Number of characteristics of the APP included in the announcements, ranging from 12 to 27.Source: Staff analysis.

IV. Effectiveness of UMP Measures in EMS

In this section, we assess empirically the effects of APP announcements on financial market variables as well as their channels of transmission. We confirm the findings of previous studies that APP announcements reduce bond yields and that this effect of the direct and signaling channels of transmission is larger in magnitude for APPs compared to that of policy rate cuts and stronger in EMs compared to AEs. We also find that the strength of the exchange rate channel depends on whether the APP announcement was made within a few days of a policy rate cut. In addition, as the objective of most APPs is to boost confidence and calm markets, we look at their second-round effects on external borrowing costs. 41 If APP announcements managed to boost the confidence of investors, then the cost of borrowing should decrease as well, or at least level off. However, we find that this is not always the case.

We use an event study methodology to study the effects of APP announcements in our sample of 15 EMs and 8 small AEs.42 This methodology allows us to study the one, two and three-day impact of the APP announcements, where we also look at announcements that coincide with a policy rate cut. In Annex IV, we present a second event study as well as a set of panel regressions to check whether the effects on the days of announcements hold when control variables are introduced. We do this because the impact of APP announcements on the main variables of interest might be influenced by other policy announcements as well as shocks that we have not captured in the analysis.

The first event study, following Hartley and Rebucci (2020), aims to assess the effect of APP announcements on sovereign bond yields, as well as the exchange rate. We also investigate the second-round effects of APP announcements and whether they managed to affect the cost of external borrowing. While Hartley and Rebucci (2020) consider only the impact of the first announcement of every new program, which are more likely to be a surprise, we consider subsequent announcement dates as well, similar to Gagnon et al. (2011). Considering all announcement dates is important because the majority of central banks that made consecutive APP announcements did so in order to adjust the size and/or scope of the program, target additional maturities, etc. and such information can provide valuable signals to the market participants.

The effects of announcements are studied using the following variables of interest: (i) yield curve—10-year bond yield; 5-year bond yield; 2-year bond yield; 6-month bond yield; and (ii) exchange rate (local currency per US dollar). In order to estimate the second-round effects on external funding costs, we look at the EMBI spread. Finally, we also consider the equity index and the corporate bond yields.43 The impact on the variables of interest are evaluated using daily data from Reuters, omitting weekend and official holiday dates.

We compute the 1, 2, and 3-day change for each of the variables of interest, starting on the day prior to the announcement date, and divide it by the corresponding unconditional standard deviations.44 These standard deviations are computed using daily data from January 1, 2017 through the day prior to the announcement, allowing for a sample of about 1,000 observations, long enough to obtain reliable measures. We then test the null hypothesis that the APP announcements have no significant impact on the variables of interest. We consider multi, rather than only one-day windows, because asset prices might not react instantaneously and in full following a policy announcement.

The main findings for EMDEs are shown in Figures 15–17 (see also Annex III). The results are rather heterogeneous across countries as well as announcements. However, on average, the estimated effect is statistically significant and broadly consistent with that found in the literature and in particular the results of Hartley and Rebucci (2020), Arslan et al. (2020), as well as Sever et al. (2020). In contrast to previous studies which focused only on the 10-year bond yield, we also analyzed the impact of APPs on other maturities. For other maturities between 2 and 5 years, the announcements also had a negative and statistically significant multi-day effect on bond yields suggesting effectiveness of APPs across the yield curve. However, APP announcements that coincided with FX intervention were not effective in lowering borrowing costs: they had a positive and statistically significant effect on the 10-year bond yield (Table III.1).

Figure 15.
Figure 15.

10-Year Bond Yield and Exchange Rate Results

Citation: IMF Working Papers 2021, 014; 10.5089/9781513567211.001.A001

Sources: Reuters and staff analysis.
Figure 16.
Figure 16.

Other Bond Yields and Financial Variables Results

Citation: IMF Working Papers 2021, 014; 10.5089/9781513567211.001.A001

Sources: Reuters and staff analysis.
Figure 17.
Figure 17.
Figure 17.

Results for the 10-Year Bond Yield by Type and Number of Announcements

Citation: IMF Working Papers 2021, 014; 10.5089/9781513567211.001.A001

Sources: Reuters and staff analysis.

Quantity-based programs as well as programs focusing on government securities seem to be more effective in terms of reducing the bond yields. No other specific patterns emerged when considering whether the announcements with the following characteristics were more effective: (i) made jointly with other national authorities; (ii) part of a policy package; (iii) had a detailed communication; (iv) the CB made a single or multiple announcement; (v) made by a credible CB; (vi) made by a transparent CB; (vii) the type of the exchange rate regime, (viii) capital account openness; (ix) the share of non-resident investors. In some cases, the first APP announcement had the expected statistically significant and negative effect on the 10-year bond yield, which was also more pronounced compared to the effect of consecutive announcements. However, in other cases, subsequent announcements were more effective, or the first announcement was not effective at all. Therefore, it remains inconclusive whether initial announcements are more effective due to their “surprise effect.”

The findings related to the effect of APP announcements on the bond yields, continue to hold when we exclude from the sample the announcements that coincide with a policy rate cut. Specifically, we exclude 9 APP announcements from the sample. As is illustrated in Figure 16, this is the case not only for the 10-year bond yield, but for other maturities as well.

The results related to the exchange rate indicate that the strength of the exchange rate channel depends on whether the APP announcement was made within a few days of a policy rate cut. When all APP announcements are taken under consideration, the results are predominantly positive and statistically significant, indicating that most APP announcements resulted in exchange rate depreciation. However, a few central banks announced policy rate cuts just the day before or the same that that they also announced an APP (Chile, Indonesia, Mexico, Poland, and Thailand). Therefore, it is very likely that the depreciation of the exchange rate after the APP announcement is a spillover effect from the policy rate cut announced the previous day or a few days before.45 When the APP announcement dates that are close to a policy rate cut are excluded (following the existing literature), the results for the exchange rate effect are inconclusive and thus suggest a weaker exchange rate channel. Furthermore, similarly to the results related to the 10-year bond yield, we could not identify any patterns (Table III. 2).46

Looking at the second-round effects, the announcements have predominantly a positive and statistically significant effect on the EMBI, although with a significant heterogeneity across the sample. This confirms the finding that the announcements were not able to calm the markets sufficiently and restore the confidence of international investors, thus reducing the cost of external borrowing. This finding still holds when the APP announcements that are close to a policy rate cut are excluded.

The same event study methodology applied to implementation instead of announcement dates yields broadly consistent results for the countries for which data is available. However, the sample is more limited as implementation dates are available only for 5 small EAs (Sweden, Canada, Australia, New Zealand, and Korea) and 6 EMs (Colombia, India, Thailand, Hungary, Croatia, and Poland).

Finally, the event study with control variables as well as the panel regression, presented in Annex IV, broadly confirm the findings above (Annex IV).47 In addition, the panel regressions also reveal that APP announcement made by credible central banks are more effective (in terms of their impact on bond yields). We do not find evidence of the role of central bank transparency, the non-residential investment share, or the monetary regime and the exchange rate regime on the effectiveness of the APP. Interestingly, we also find that larger programs, as measured by the announced size as a share of GDP, as well as programs announced in countries with low monetary space are associated with smaller effects on the government 10-year bond markets.

As central banks often implemented both conventional and unconventional policies, we run the same event study analysis to estimate the effects of policy rate cut announcements— including those that coincide with APP announcements. We find that these policy rate cut announcements had negative and statistically significant multi-day effects on the bond yields across the maturity curve, mostly over the period mid-March to end-April. Overall, the impact of conventional monetary policy transmission to bond yields is slightly less than that of APPs (Table III.3). Looking into the second-round effects, the results show that policy rate cut announcements managed to reduce the cost of external borrowing as well as have the expected depreciating effect on the exchange rate.

This analysis does not evaluate the effectiveness of APP programs in EMDEs related to the support of market functioning. Such an analysis is suitable for future work as this was the stated objective for most of the programs in our database. Also important for future work would be an analysis of the longer-term impact of these measures.

V. Conclusions

The COVID-19 crisis induced an unprecedented launch of unconventional monetary policy through asset purchase programs by emerging market and developing economies. Such measures have so far been largely used in advanced economies, and relatively little is known about their effectiveness in EMDEs, resulting in unease in policy circles regarding their use by these countries. This paper fills the gap by building a comprehensive database of APP announcements and implementations by a sample of 27 EMDEs and 8 small AEs from the onset of the crisis in March until August 2020. It also provides preliminary evidence of the effect of these APPs by country and announcement, as well as for the sample as a whole, in helping stabilize financial market prices.

Specifically, our findings suggest that the APP announcements were effective in reducing the bond yields across different maturities, to a greater degree than to the CMP implemented in mid-March–end-April. However, APPs had little effects on exchange rates or external borrowing costs, after controlling for other factors (other policies or global factors). The implementation effects were similar in their impact to the announcement effects. Programs that coincided with FX intervention had the opposite expected effect on bond yields (positive and statistically significant). Some other program characteristics (announcements made jointly with other authorities, multiple announcements, size) also seemed to worsen outcomes. A few country specific factors e.g. CB credibility, high monetary policy space, low share of non-resident holdings of government bonds, seemed to improve the outcomes of APPs while other (CB transparency, monetary or exchange rate regime, foreign investment share) did not seem to have an impact.

Our results suggest that APPs can be usefully deployed by EMDEs in support of their macro-financial stabilization objectives in the current COVID-19 crisis. However, more observations on their use, beyond the current crisis, and additional work will likely be needed in understanding the underpinnings and channels of the effectiveness of UMP in EMDEs to assess whether these tools could be also useful to EMDEs for normal times.

Annex I. Structure of the Database

A. Structure and Key Information in the UMP Database

The database is organized by several identifying factors, per program and per announcement, including the following factors:

1. Primary information related to the program, CB mandate, and involvement of other national authorities besides the CB:

  • a. Identify if the press release announces a new program or if the CB has already used APP in the past. We also record whether the press release mentions/allows a change in the CB mandate.

  • b. Record if the notice is released by the CB, national authorities, or jointly.

  • c. Record whether the press release mentions direct financing of the economy (captured by interventions on primary markets). We keep track of announcements in which the CB explicitly distances itself from direct financing.

2. Information on the program’s objective(s):

  • a. Record the program’s objective using direct quotes from the press release.

  • b. Classify the objectives into four main categories using textual analysis of the verbatims: “Support fiscal needs and measures to alleviate costs on the population”, “Boost confidence and tackle market dysfunctionalities”, “Provide monetary stimulus” and “Address exchange rate concerns”.48

  • c. For the first three categories, we also record more details of the objectives.49 For example, a program can have several objectives and sub-objectives falling under different categories.

3. Information on the program type (price/quantity):

  • a. Record whether the program is price-based and if the press release specifies/updates a price target.

  • b. Record whether the program is quantity based and if the press release specifies/updates a quantity target as well as the frequency of implementation.

  • c. Identify whether the press release refers to peculiar programs (swap, grant, SPV, etc.) and if the program is changing its operation procedure.

4. Information on the program’s main targets:

  • a. Record whether the press release indicates targeting public or private securities (type of asset). and whether the CB is intervening on the primary or secondary markets.

  • b. Record if specific maturities are mentioned in the press release.

  • c. Record whether the press releases provides/updates information on the counterparties/credit rating targeted as well as whether the transaction is defined (outright, auctions, bilateral etc.).

5. Information on the CB’s communication style:

  • a. Identify whether this is the first press release related to the APP. If no, we record the consecutive number of the announcement. We also record if the press release provides more information on an existing APP.

  • b. Record whether the press release provides information on expanding an existing the program or is announcing an APP in the future.

  • c. Record whether the APP announcement is done jointly with other CB measures:

    • i. Interest rate cuts;

    • ii. Repurchase agreements;

    • iii. Reduction in reserve requirements;

    • iv. Credit facilities/lending operations;

    • v. Foreign Exchange Interventions; and

    • vi. Capital flow measures.

6. Information on APP implementation

  • a. Record whether the press release specifies an implementation date, as well as whether the frequency of interventions is increased/scaled back.

  • b. Record whether the press-release announces more/less purchases, if so by how much and whether the press release terminates the program?

In addition, the database contains information on the APP transactions made by the central banks, when available (date, market, size, price, issuer, yields, demand etc.)

Table 1A

Taxonomy of Asset Purchases Programs’ Objectives

article image
Source: Staff analysis.

To construct the taxonomy, we identify, across all quotes recorded as stated objectives, groups of key words that structure the textual information.

We then group the selected key words by topic and recurring words/elements50. We then organize the groups into broader categories to reflect the objectives of UMP that are commonly discussed in the literature and are being presently addressed in the press releases.

The companion excel spreadsheet providing information at the program level, provides the information on all verbatim and how they are related to the layers of the taxonomy.

Annex II. Details on Taxonomy of Objectives

article image
article image
article image
article image
article image
article image
article image
article image
article image
article image
article image
article image
article image
article image

Annex III. Econometric Results

Table III.1.

1, 2, and 3-day Effect on the 10-year Bond Yield following APP Announcement Dates

article image
Note: Countries above the first bold line announced quantity-based programs and countries between the two bold lines announced price-based programs. Croatia did not specify. One dot indicates whether the APP focused on government securities, the CB is considered credible, and the APP was announced together with FXI. Two dots indicated that the APP focused both on government and private securities. For the March 22nd announcement for Thailand, which is on Sunday, we used March 20th instead, as data is not available on weekends. Sources: Reuters, Dincer/Eichengreen/Geraats revised Central Bank Transparency Data Set for 1998–2015, and staff analysis.
Table III.2.

1, 2, and 3-day Effect on the Nominal Exchange Rate following APP Announcement Dates

article image
Note: *De Facto Classification of Exchange Rate Arrangements, as of April 30, 2019. One dot indicates whether the APP announcement was around the time of a policy rate cut. For the March 22nd announcement for Thailand, which is on Sunday, we used March 20th instead, as data is not available on weekends. The Philippines announced a Php 300 billion purchase of government securities under repo on March 22nd, not included in the database. Source: Reuters and staff analysis; Chinn, Menzie D. and Hiro Ito (2006). “What Matters for Financial Development? Capital Controls, Institutions, and Interactions,” Journal of Development Economics, Volume 81, Issue 1, Pages 163–192 (October); 2019 Annual Report on Exchange Arrangements and Exchange Restrictions.
Table III.3.

1, 2, and 3-day Effect on the Nominal Exchange Rate following Policy Rate Cut Announcement Dates

article image
Sources: Reuters and staff analysis.
Table III.4.

1, 2, and 3-day Effect on the EMBI Spread following APP Announcement Dates

article image

Annex IV. Event Study with Controls and a Panel Regression

We enhance the event study methodology discussed in Section IV to control for global factors (e.g., periods of heightened volatility), and other factors (e.g., other CB policies, pandemic-related shocks) that may affect the variables of interest.51 We use country specific OLS regressions, with the variables of interest as the dependent variable. The right-hand side includes a dummy capturing the APP announcement dates as well as control variables such as the VIX (to gauge for financial market risk aversion and a general proxy for financial turmoil, economic risk and uncertainty), the domestic central bank policy rate change, the Fed’s policy rate change, as well as the Oxford index52 and the Google mobility index53 in order to control for effects related to the covid-19 pandemic. The following regression is estimated for every country using daily data from January 1, 2020 until the end of August:

ΔYit = α + β1ANNit + β2VIXit + β3ΔIit + β4ΔFEDt + β5Oxfordit + β6Gooogleit + εt

Where Ys are the variables of interest used in the previous event study, ANN is the dummy variable capturing the days of APP announcements, I is the domestic policy rate change, FED is the Fed’s policy rate change, VIX is the volatility index, Oxford is the government response to COVID-19 tracked by the University of Oxford, and Google is the Google mobility index.

The regression results for every country are presented in the table below, showing the coefficients on the announcement dummy variable only from every regression. They confirm that even when other factors are considered, the APP announcements have a statistically significant and negative effect on the bond yields and little effects on the exchange rate as well as the remaining dependent variables. This is probably because some announcements have a positive effect on the exchange rate while other have a negative one. Furthermore, the control variables such as the VIX and the FED’s policy rate change often had statistically significant coefficients, positive and negative, respectively in the regressions with the exchange rate. These effects of the VIX index and the FED policy rate change likely washes out the significant effect on the exchange rate. Using a dummy variable with the implementation dates did not yield any statistically significant results for the countries where data is available. However, these results should be taken as an illustration only and interpreted with caution as there is little variability in the data.

To further check whether the main findings of the event study methodology without controls regarding the effects of UMP hold, we use a panel regression as well. We run a panel regression with fixed effects, where the specifications of the dependent and independent variables are the same as the ones with the country-level regressions.54

The results, using the full sample in our database including both EMDEs and AEs, indicates that the APP announcements had a statistically significant and negative effect on the bond yield across maturities. Restricting the sample to EMDEs only confirms these and illustrates the stronger effect of APP announcements. In addition, the effect on the exchange rate is not statistically significant55, similar to the findings of Chapter 2 of the October GFSR. Finally, including a dummy variable for the implementation days did not yield any statistically significant results, which is most likely due to the limitation of the data.

Next, we explore whether there are any patterns in the effectiveness of APP announcements. To do so, we interact the announcement dummy variable with a number of structural characteristics. In particular, we explore the role of central bank credibility (measured by the variability of long-term inflation forecasts), central bank transparency, the share of non-residential investment, the monetary regime (inflation targeting or other regimes), the announced size of the APP, and the exchange rate regime (floating or not). We use the specification of the panel regressions above.

The coefficients of the announcement dummy interacted with the country structural characteristics are reported in Table IV.4 below. We find that APP announcement made by credible central banks are more effective (in terms of their impact on bond yields). We do not find evidence of the role of central bank transparency, the non-residential investment share, or the monetary regime and the exchange rate regime on the effectiveness of the APP. However, we ascribe the latter to the little heterogeneity in the sample, rather than a conclusive statement on the role of the exchange rate regime. Specifically, most of the countries in the sample have a floating exchange rate regime, with the exception of Croatia, Egypt, and Indonesia. Similarly, most countries in the sample follow an inflation targeting regime, with the exception of Croatia and Egypt.

Interestingly, we find that larger programs, as measured by the announced size as a share of GDP, are associated with smaller effects on the government 10-year bond markets. This can be interpreted as an indication that markets may not yet be reacting fully, or know how to absorb the information, of these new unconventional policies. It could also be attributed to the prevailing market uncertainty and strong risk aversion arising from the pandemic. Furthermore, announcements of the Chile price-based program targeting bank bonds had the opposite effect on sovereign bond markets than announcements of quantity-based programs. We also find that programs with multiple announcements as well as those announced in countries with low monetary space are associated with smaller effects on the government 10-year bond markets. Overall, conclusions on the role of the institutional framework and the APP characteristics are robust to the inclusion of AEs in the regressions as well as to the restrictions of the sample to the first announcements for each program.56

When the sample is restricted to the first announcement of each program (Table IV.5), in order to investigate the surprise effect, we find broadly similar results. One notable result is that when the first announcement is considered only, central bank transparency matters as well for the effectiveness of APPs. Finally, trying to distinguish between the effectiveness of APPs between EMDEs and the small AEs in our sample, we find a more significant impact for EMDEs (Table IV.6). However, the results related to the implementation dates may be biased due to the very small sample size for EMDEs.

Table IV.1.

Regression Coefficient on the Announcement Dummy Variable

article image
*** p<0.01, ** p<0.05, * p<0.1
Table IV.2.

Panel Regression Results, EMDE Sample

article image
Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1 Source: Staff analysis.
Table IV.3.

Average of Country-Specific Coefficients for the EMDE Sample

article image
Source: Staff analysis.
Table IV.4.

Panel Regression Coefficients with Interactive Terms for EMDEs

article image
Note: each cell corresponds to a different regression. The table reports the coefficients of the announcement dummy interacted by the variables listed in the first column. Dependent variables for each regression are the first difference of the variable whose name is reported in the first row. P-values are reported in parentheses. In blue, coefficients significant at 1, 5, or 10 percent. Sources: Reuters, IMF WEO, IMF IFS, Dincer/Eichengreen/Geraats revised Central Bank Transparency Data Set for 1998–2015, and staff analysis.
Table IV.5.

Panel Regression Coefficients with Interactive Terms for EMDEs, First Announcement Only

article image
Note: each cell corresponds to a different regression. The table reports the coefficients of the announcement dummy interacted by the variables listed in the first column. Dependent variables for each regression are the first difference of the variable whose name is reported in the first row. P-values are reported in parentheses. In blue, coefficients significant at 1, 5, or 10 percent. Sources: Reuters, IMF WEO, IMF IFS, Dincer/Eichengreen/Geraats revised Central Bank Transparency Data Set for 1998–2015, and staff analysis
Table IV.6.

Panel Regression Coefficients with Interactive Terms for EMDEs and AEs

article image
Note: each cell corresponds to a different regression. The table reports the coefficients of the announcement dummy interacted by the variables listed in the first column. Dependent variables for each regression are the first difference of the variable whose name is reported in the first row. Panel a reports results for all announcements. Panel b reports results only for the first announcement for each country. Panel c reports result on the date of implementation. P-values are reported in parentheses. Indonesia has been excluded. P-values are reported in parentheses. In blue, coefficients significant at 1, 5, or 10 percent. Sources: Reuters and staff analysis.

Annex V. Literature Review

article image
article image
article image

VI. References

  • Altavilla, C., D. Giannone, and M. Lenza, 2013. “The Financial and Macroeconomic Effects of OMT Announcements,” ECB Working Paper Series No. 1707.

    • Search Google Scholar
    • Export Citation

  • Arslan, Y., M. Drehmann, and B. Hoffman, 2020. “Central Bank Bond Purchases in Emerging Market Economies,” BIS Bulletin No. 20.

  • Briciu, L., and G. Lisi, 2015. “An Event-Study Analysis of ECB Balance Sheet Policies Since October 2008,” European Economy, Economic Brief 001.

    • Search Google Scholar
    • Export Citation

  • Chen, J., T. Mancini-Griffoli, and R. Sahay, 2014. “Spillovers from United States Monetary Policy on Emerging Markets: Different This Time?” IMF Working Paper No. 14/240 (Washington: International Monetary Fund)

    • Search Google Scholar
    • Export Citation

  • De Pooter, M., R. F. Martin, and S. Pruitt, 2018. “The Liquidity Effects of Official Bond Market Intervention”, Journal of Financial and Quantitative Analysis, Vol 53, No 1, Feb 2018

    • Search Google Scholar
    • Export Citation

  • Falagiarda, M., P. McQuade, and M. Tirpak, 2015. “Spillovers from the ECB’s Nonstandard Monetary Policies on Non-Euro Area EU Countries: Evidence from an Event-Study Analysis,” ECB Working Paper Series No. 1869

    • Search Google Scholar
    • Export Citation

  • Gagnon, J., M. Raskin, J. Remache, B. P. Sack, 2011. “Large-Scale Asset Purchases by the Federal Reserve: Did They Work?”, International Journal of Central Banking, Vol. 7, No1. pp. 3–41

    • Search Google Scholar
    • Export Citation

  • Gambacorta, L., B. Briciu, and G. Peersman, 2013. “The Effectiveness of Unconventional Monetary Policy at the Zero Lower Bound: A Cross-Country Analysis”

    • Search Google Scholar
    • Export Citation

  • Hartley, J. and A. Rebucci, 2020. “An Event Study of Covid-19 Central Bank QE in Advanced and Emerging Economies,” NBER Working Paper No. 27339.

    • Search Google Scholar
    • Export Citation

  • International Monetary Fund, 2013. “Global Impact and Challenges of Unconventional Monetary Policies,” Monetary and Capital Markets/Strategy and Policy Review Departments, IMF Policy Paper, October 18 (Washington)

    • Search Google Scholar
    • Export Citation

  • International Monetary Fund, 2016. “Cross Country Report on Spillovers”, IMF Country Report No 16/212

  • Jordà, Ò., 2005. “Estimation and inference of impulse responses by local projections”, American Economic Review, vol 95, no 1, pp 161–82

    • Search Google Scholar
    • Export Citation

  • MacDonald, M., and M. K. Popiel, 2017. “Unconventional Monetary Policy in a Small Open Economy”, IMF Working Paper 17/268

  • Rai, V., and L. Suchanek, 2014. “The Effect of the Federal Reserve’s Tapering Announcements on Emerging Markets,” Bank of Canada Staff Working Paper 2014–50

    • Search Google Scholar
    • Export Citation

  • Sever, C., R. Goel, D. Drakopoulos, and E. Papageorgiou, 2020. “Effects of Emerging Market Asset Purchase Program Announcements on Financial Markets During the COVID-19 Pandemic.” IMF Working Paper 20/292

    • Search Google Scholar
    • Export Citation

  • Swanson, Eric T., 2011. “Let’s Twist Again: A High-Frequency Event-Study Analysis of Operation Twist and Its Implications for QE2”, Brookings Papers on Economic Activity, Spring 2011

    • Search Google Scholar
    • Export Citation

  • Vissing-Jorgensen, A. and A. Krishnamurthy, 2011. “The Effects of Quantitative Easing on Interest Rates: Channels and Implications for Policy”, Brookings Papers on Economic Activity, Fall 2011. pp. 215–265

    • Search Google Scholar
    • Export Citation

  • Weale, M. and T. Wieladek. 2016. “What Are the Macroeconomic Effects of Asset Purchases?” Journal of Monetary Economics 79: 81–93.

    • Search Google Scholar
    • Export Citation
1

The authors gratefully acknowledge the useful discussions and suggestions by Rupa Duttagupta, Petya Koeva Brooks, participants of the EM Forum, the authors of Chapter 2 of the IMF’s October 2020 Global Financial Stability Report, and comments from the ED offices representing the countries in the sample.

2

All programs for EMs were newly announced during the COVID-19 crisis, except Indonesia, where it was an expansion of a pre-existing program. During and after the GFC, AEs introduced four main types of unconventional measures: (i) negative interest rate policies, (ii) lending operations or credit policies, where central banks expand their liquidity facilities to support credit flows to the private sector, (iii) large-scale asset purchase programs to address market dysfunctionalities that impaired the transmission channels of monetary policy and/or lower long term bond yields and ease broad financial conditions, and (iv) forward guidance.

3

The universe of countries discussed in this section comprises 27 EMDEs and 8 small AEs, which deployed UMP tools between January and August 2020. However, the empirical analysis is constrained at times to smaller sub-samples based on data availability.

4

At the same time, the weakening local currency should stimulate exports and have an offsetting positive effect on the economic outlook and inflation.

5

We included realized transaction data only when such information was publicly available and clearly linked to the asset purchase program announced. In the case of Philippines, purchases of government securities under the follow-up of the repurchase agreement program in the secondary market by the CB (announcement of April 10th, 2020) are excluded from the database.

6

Even though the paper focuses on CB interventions, several reasons justify not relying solely on the CBID: (i) some announcements were made jointly and announced, for example, by the Ministry of Finance; (ii) in some cases, the programs were directly announced by other authorities (Presidential decrees, Prime Minister Statement); (iii) finally in some cases the passing of a law allowing the CB to purchase assets served as the first mention to market participants of this shift in mandate of the CB.

7

For Egypt and Ethiopia, the CB communication did not discuss the introduction of the APP. We collected instead information from the news.

8

See next subsection for further details.

9

In the case of Norway, the CB was not involved in the conduct of the APP even though the measures were branded as CB asset purchases/sales (and thus included in the database). Rather, the government commissioned a dedicated fund to purchase private securities, e.g. equities.

10

The Bank of Thailand established the Corporate Bond Stabilization Fund to stabilize the corporate bond market.

11

Central Bank Press Release, March 2nd, 2020.

12

The sample of EMDEs is restricted to those countries for which we have data availability.

13

Whenever the amounts were expressed in local currency units, we used the 2020 end-of-period exchange rate from the last published WEO to convert the amounts. The amounts in percentage of GDP are computed using the 2019 GDP estimate from the last published WEO. The amount in USD for China is truncated for expositional purposes.

14

For Thailand, we present the announced size of both its private and public purchase programs, but we only have data on purchases of sovereign bonds. For Korea, we only report the size of the sovereign debt purchase program.

15

For Turkey, data on implementation refers only to purchases of government bonds from the Unemployment Insurance Fund.

16

The sample in the Figure is restricted to those countries for which we have an exact date of implementation.

17

Countries are counted more than once if they announced more than one program and/or stated more than one objective.

18

Among AEs, 80 percent of the programs had as stated objective boosting confidence and tackling market dysfunctionalities (Australia, Iceland, Israel, Korea, New Zealand, Sweden) and 20 percent to provide monetary stimulus (Israel, New Zealand).

19

“[…] all the above-mentioned measures will mitigate the negative economic impact of coronavirus spread, while further ahead they will contribute to recovery in domestic economic activity and will be conducive to inflation running at the level of the NBP inflation target in the medium term.” Narodowy Bank Polski, Information from the meeting of the Monetary Policy Council held on 17.03.2020. https://www.nbp.pl/en/aktualnosci/2020/mpc_2020_03_17.pdf

21

Central bank press release, April 21, 2020. The phrase cited in the text has been translated using Google translate. https://www.bi.go.id/id/ruang-media/info-terbaru/Pages/BI-Terbitkan-Ketentuan-Pelaksanaan-Lelang-SUN-dan-atau-SBSN-Jangka-Panjang-di-Pasar-Perdana.aspx

22

Bank of Mauritius, Covid-19 Support Programme: Supporting Systemic Economic Operations and Financial Stability. https://www.bom.mu/media/covid19-actions/covid-19-support-programme-supporting-systemic-economic-operators-and-financial-stability

23

Bank of Ghana, Monetary Policy Committee Press Release, May 15, 2020. https://www.bog.gov.gh/wp-content/uploads/2020/05/MPC-Press-Release-15th-May-2020-2.pdf

24

The information reported refers to announced policies. Policies implemented and not announced would not be captured in the dataset.

25

We consider that policy rate cuts consider with APP announcements if they occurred simultaneously, the day before, or a few days before.

26

The policy rate cut was on April 16th.

27

The Polish central bank purchased government-guaranteed debt securities besides sovereign bonds. In our database, it is classified as APP in the sovereign market.

28

In line with APPs’ objective to provide support to fiscal needs related to the COVID-19 outbreak (Figure 9).

29

The South African Reserve Bank stated “No, we are not monetising the government debt. The SARB is not giving money directly to government and is not buying bonds from government”. Czech Republic never announced an APP. However, the CB mandate was changed and the CB press releases stated: “The amendment to the Act on the CNB concerning CNB open market operations which was approved by the government today does not affect the prohibition of monetary financing.”

30

Countries are listed more than once if announced programs with different maturity targets. The information reported is related to the most updated information provided by the central bank.

31

For a description of the database and the variables included see Annex I.

32

The metric is imperfect as quantity of information may not necessarily capture the quality/impact of the communication but nonetheless helps to illustrate the diverse CB communication strategies across countries.

33

We also identify South Africa and Costa Rica as having a detailed communication strategy, with long and detailed CB announcements.

34

In the case of Iceland, represents an upper bound.

35

Realized size refers to the amount of realized transactions until August 2020, if data is available.

36

Percent of 2019 GDP.

37

On March 24th, the Magyar Nemzeti Bank stated they were assessing “the possibility of relaunching the MNB’s mortgage bond purchase program to increase bank liquidity, while the Financial Stability Council has temporarily loosened its regulation on mortgage bond funding.” Press Release on the Monetary Council Meeting of 24 March 2020.

38

These non-standard measures were implemented as early as March 2020, but no official announcement was made until April 10.

39

The APP transaction data include also THB 64,120 million in purchases of Bank of Thailand bonds.

40

This program involved the purchase of Government Domestic Debt Securities that they have bought from the Unemployment Insurance Fund by primary dealer banks.

41

Exploring the effect of APP announcements on stress indicators (e.g., liquidity risk), tail risks, and market liquidity indicators (bid/ask spreads, volatility, market turnover, etc.) would be desirable given the objectives of many programs of addressing market dysfunctions. However, doing so faces data limitations for most EMDEs.

42

The EM sample size is constrained by the availability of financial variables of interest. Nearly all EMs included in the sample announced quantity-based programs involving purchases of sovereign bonds (in a few cases, Colombia, Hungary, and Thailand, the programs involved purchases of both sovereign and private securities). Chile is the only country where the APP was price-based, targeting bank bonds. Croatia did not specify whether the APP was price- or quantity based.

43

The results obtained from these variables are available upon request. The results for the equity index and corporate bond yields indicate significant heterogeneity across the sample and an inconclusive average median effect (Figure 16).

44

This is similar to the approach employed by Hartley and Rebucci (2020). Vissing-Jorgensen and Krishnamurthy (2011) and Swanson (2011) use two-day windows, while Gagnon et al. (2011) use 1-day event windows.

45

The authors also appreciate the fact that other factors, such as the effect of the pandemic as well as the country’s macroeconomic fundamental may be at play, which we attempt to control for with the analysis in Annex IV.

46

Additional results are available upon request.

47

The results of panel regressions find no significant effect of first announcements on the 10-year bond yield (Table IV.6, panel b).

48

To that end, we identify groups of key words that structure the textual information. We group the selected key words by topic and recurring elements. We then organize the groups into broader categories to reflect the objectives of UMP from the literature that are being addressed in the press releases.

49

For category 1 (fiscal needs etc.) we detail whether the objectives aim at “budget financing” (and related key words) or if they concern “Covid 19 support measures to the government or the population”. For category 2 (DTC) we detail whether the objectives focused on boosting confidence and addressing market dysfunctionalities.

50

We allow for some degree of flexibility as, for example, some statements explicitly define targeted institutions for example banks or corporations, whereas other statements mention more broadly financial or non-financial institutions.

52

The Oxford COVID-19 Government Response Tracker (OxCGRT) systematically collects information on several different common policy responses that governments have taken to respond to the pandemic on 17 indicators such as school closures and travel restrictions.

53

Using mobility trends for places of work.

54

In a separate specification we also included lagged values of the dependent variable. As this did not yield results that are materially different, we do not discuss them here.

55

We exclude Indonesia both from the full as well as the EMDE sample as it is a large outlier.

56

Robustness checks are not shown in full and are available upon request.

Same Series

Other IMF Content

Other Publishers

  • Collapse
  • Expand
Cover IMF Working Papers
Unconventional Monetary Policies in Emerging Markets and Frontier Countries
Volume/Issue:
Volume 2021: Issue 014
Publisher:
International Monetary Fund
ISBN:
9781513567211
ISSN:
1018-5941
Pages:
71
DOI:
https://doi.org/10.5089/9781513567211.001
Figure 1.

Stress in Financial Markets, January–July 2020
Figure 2.

Policy Rate and Public Debt Levels Pre-COVID-19 Crisis in EMDEs
Figure 3.

CMP and UMP Transmission to the Economy
Figure 4.

APP Announcements by EMDEs by Type of Securities Purchased
Figure 5.

APP Announcements by EMDEs and Policy Frameworks
Figure 6.

Size of Announced and Realized CB Interventions
Figure 7.

Implementation of APPs Over Time 1/
Figure 8.

Realized Size of Interventions
Figure 9.

Stated Objective of APP Announcements in EMDEs
Figure 10.

APP Stated Objectives and Other Policy Measures 1/
Figure 11.

Asset Purchase Programs by Primary vs. Secondary Market Purchases
Figure 12.

Asset Purchase Programs by Quantity-Based vs. Price-Based 1/
Figure 13.

Maturity of the Securities Purchased under the APPs 1/
Figure 14.

Information Available in APP Announcements 1/
Figure 15.

10-Year Bond Yield and Exchange Rate Results
Figure 16.

Other Bond Yields and Financial Variables Results
Figure 17.

Results for the 10-Year Bond Yield by Type and Number of Announcements
Figure 17.

Results for the 10-Year Bond Yield by Type and Number of Announcements