The sensitivity of long-term interest rates to global factors and domestic credit growth

(1)

The sensitivity of long-term interest rates to

global factors and domestic credit growth

Master Thesis

International Economics and Business (M.Sc. & M.A.)

University of Groningen – Faculty of Economics and Business

Corvinus University of Budapest – Faculty of Economics

Abstract

Financial globalization made it easier for financial imbalances to be transmitted across countries and regions. The bond market can serve as a stage for the transmission of financial conditions, where term interest rates exhibit a considerable co-movement. The paper investigates the sensitivity of long-term interest rates to global factors such as global market uncertainty (VIX) and US monetary policy (US federal funds rate), and finds a statistically significant effect. Additionally, changes in the financial system have both enhanced the role of credit and increased the susceptibility of domestic credit to global influences. The paper also investigates the sensitivity of long-term interest rates to domestic credit growth, but finds no conclusive evidence.

Key words: Long-term interest rates, global factors, domestic credit growth

Name: Anika van der Vaart Student number: S3192822

E-mail: A.Van.der.Vaart.3@student.rug.nl Supervisor: Professor D. Bezemer

(2)

1. Introduction

It has long been argued that capitalist economies are prone to financial instability (Minsky, 1977). This fact became more apparent than ever during the global financial crisis of 2008/09. Not only did it become evident that ”[...] it is simply not possible to understand business fluctuations and their policy challenges without understanding the financial cycle [...]” (Borio, 2014, p.183), but that financial imbalances in another country and region can be transmitted throughout the global economy. It appears that financial globalization, which has linked financial markets throughout the world, has interconnected the fate of economies. Financial globalization has also led to the emergence of the so-called global financial cycle, which exposes economies to global funding conditions, complicating domestic monetary policy (Rey, 2013).

The global financial cycle can be defined in terms of co-movements in credit creation, gross capital flows, bank leverage and risky asset prices. There are several stylized facts that describe the global financial cycle (Passai and Rey, 2015). Firstly, gross capital flows are positively correlated across countries and regions. Secondly, these gross cross-border flows tend to co-move negatively with measures of market uncertainty and risk aversion, such as the VIX1_{. The} VIX also tends to be negatively correlated with banking leverage and domestic credit creation. And finally, risky asset prices appear to be driven by a ‘global factor’ that is strongly negatively related to the VIX (Miranda-Agrippino and Rey, 2015). Interestingly, domestic credit conditions also appear to be influenced by cross-border capital flows, indicating that the domestic financial cycle might be partly subsumed by the global financial cycle.

US monetary policy has been found to be an important driver of the global financial cycle. Changes in United States (US) monetary policy are internationally transmitted and affect market uncertainty and risk appetite, gross credit and capital flows and global bank leverage (Rey, 2013). US monetary policy does not only influence financial flows, but also appears to have an impact on gross domestic product (GDP), inflation and consumer sentiment, especially in Europe (Miranda-Agrippino and Rey, 2015). The fact that US monetary policy seems to spill over into other countries, suggest that there is an international transmission channel. Shin (2013) argues that until the 2008/09 crisis, financial conditions were transmitted through banking sector capital flows. And indeed, global bank leverage and banking flows have been found to be vital in channeling US financial conditions to other markets. However, after the crisis, from about 2010, the transmission of financial conditions appears to take place in bond markets (Shin, 2013). Long-term interest rates appear to be driven by a ‘global factor’. However, stronger international links between long-term interest rates could hamper the effectiveness of monetary policy, as monetary policy uses its influence on short-term interest rates to affect long-term interest rates.

1_{The VIX is the Chicago Board Options Exchange (}_{CBOE)'s Volatility Index and it is a measure of the implied volatility of S&P}

(4)

Following this line of argument, the aim of this paper is to gage the sensitivity of long-term interest rates to global factors. Specifically it looks at the sensitivity to the global financial cycle, US monetary policy and domestic credit conditions. Examining the sensitivity of long-term interest rates to domestic credit conditions may be of interest as well, as domestic financial cycles, which can be proxied by domestic credit growth, appear to be influenced by global conditions. Domestic liquidity conditions may also influence investors’ appetite for long-term government bonds. Using cross-country panel data on 24 Organization for Economic Co-operation and Development (OECD) countries over the period of 2000 to 2015, this paper intends to answer the following research questions:

(1) Are long-term interest rates sensitive to the global financial cycle (proxied by the VIX) and US monetary policy (proxied by the US federal funds rate)? Do gross cross-border capital flows enhance the effect of US monetary policy on long-term interest rates?

(2) Do domestic liquidity conditions, i.e. credit growth, have an impact on long-term interest rates?

The paper contributes to the existing literature in that it combines the stylized facts about the global financial cycle and the increasing importance of credit cycles, with the literature on determinants of long-term interest rates. Instead of focusing on domestic macroeconomic determinants of long-term interest rates, the paper aims to expand the existing literature by gauging the impact of the global financial cycle (proxied by the VIX) and its main determinant (US federal funds rates) and domestic credit growth, on long-term interest rates. The results suggest that long-term interest rates are statistically significantly influenced by the VIX and the US federal funds rate. Additionally, the sensitivity of long-term interest rates to the US federal funds rate appears to be influenced by cross-border credit flows. The evidence on domestic credit growth, however, is inconclusive.

(5)

2. Literature review

2.1 Pre financial crisis

The 2008/09 financial crisis revealed a previously largely overlooked circumstance: financial imbalances in the global financial cycle can lead to a crisis, even if output growth is stable and inflation is low. Before the crisis, literature studying the business cycle relied on macroeconomic models that largely abstracted from the financial sector.2 Given that financial factors were often excluded from business cycle analyses, the resulting monetary policy only relied on macroeconomic and not financial variables. In theory, monetary policy is set according to the Taylor rule3_{, which only accounts for domestic factors (inflation gap and output gap) and} disregards financial flows and stocks (Taylor, 1993). However, the financial crisis revealed that domestic and global financial cycles are more important for macroeconomic dynamics than previously assumed.

Despite the aforementioned general exclusion of financial considerations from business cycle studies, a few researchers considered them early on. The so-called credit view of money and finance (Schumpeter, 1934; Kindelberger, 1978) argued that rule based monetary policy should go beyond the focus on output gaps and inflation. Instead monetary policy should observe credit movements and incorporate them in a broader monetary policy framework. The financial instability hypothesis (FIH) by Minsky (1977) draws on this view, arguing that capitalist economies are prone to financial bubbles - which inevitably burst. The financial bubbles are formed due to the procyclical nature of asset prices. For example, as property prices rise, more people want to invest in property, further driving up the price until eventually an asset bubble forms. What adds to this financial fragility, according to Minsky, is the increase in speculative and Ponzi finance, i.e. borrowing at low short-term interest rates and lending at high long-term rates. Consequently, given that the capitalist economy inevitably goes through periods of financial stability and instability, governments should acknowledge and regulate this market failure (Minsky, 1977).

Going one step further, Wolfson (2002) reasons that, given the institutional changes in the global economy, Minsky’s FIH should be modified to include the international setting. The key issue is that money flows globally, where one country can invest in another and can also lend and/or borrow globally. Wolfson (2002) refers to the Asian financial crisis in the 1990s, where a large amount of lending and investment flowed into emerging Asian markets. Financial institutions

2_{For example, in examining new perspectives on monetary policy Gali (2002) only considers the linkages between monetary}

policy inflation and the business cycle. Gürkaynak et al. (2005) look at the effects of economic news on long-term interest rates and disregard any international dimension.

3_{The Taylor rule is an approximation of the sensitivity of nominal interest rates set by central banks to output, inflation and other}

(6)

made use of speculative finance (or carry trade), borrowing in countries with low interest rates (Japan) and lending in other Asian countries where interest rates were higher. These financial movements then, caused considerable instability in both the domestic and the global context.

2.2 The financial cycle

2.2.1 Characterizing the financial cycle

As has been alluded to by Minsky, financial fragility in capitalist markets is closely related to the procyclicality of credit and asset prices. Strong episodes of credit growth tend to coincide with large increases in equity and property prices (Borio et al., 2001). This procyclicality is amplified by the positive feedback loop between credit growth, asset price inflation and spreads. A lower risk premium, for example, amplifies credit growth. In turn, rising asset prices make balance sheets look healthier. And in combination with lower measured risk, value-at-risk constraints relax, which in turn creates space for more lending and credit, due to higher leverage, and so on. This feedback loop then suggests that a rapid build up of credit and usually sharp increases in asset prices during an economic expansion, can sow the seeds for a more severe recession. Empirically, Gourinchas and Obstfeld (2012) find that regardless whether a country is advanced or emerging, a rapid buildup of domestic credit expansion and real currency appreciation are the most significant predictors of financial crises. Aikman et al. (2015) argue “[c]redit lies at the heart of the crisis” (p. 1072). The authors show that the large deviations of credit, specifically bank lending, from trend in boom phases are highly correlated with banking crisis. They also emphasize that credit cycles and the business cycle are distinct from each other and differ in their amplitude and frequency. Similarly, Borio and Drehmann (2009) find evidence that unusually large increases in asset prices and credit are fairly successful in providing a signal for bank system distress.

(7)

sheets and thus leverage ratios, fuelling financial crisis.

2.2.2 Financial cycle as determinant of the business cycle

Building on Minskys’ and Wolfsons’ insights, research since the 2008/09 financial crisis goes beyond looking at financial aspects of the business cycle.4 Instead, it looks at the financial cycle as a determinant of the business cycle. Some research investigating the sources of macroeconomic instability have singled out financial and uncertainty shocks as drivers of economic instabilities. For example, Caldara et al. (2016) provide evidence that financial and uncertainty shocks negatively influence economic outcomes and that these shocks were a source of business cycle fluctuations over the past four decades. Likewise, Borio et al. (2001) argue that the excessive procyclicality of the financial cycle amplifies fluctuations in the real economy, implying that financial developments reinforce the momentum of the underlying business cycle. Claessens et al. (2012) empirically confirm the hypothesis that the financial cycle amplifies the business cycle. More specifically, using an extensive dataset, the authors find that recessions coinciding with episodes of financial disturbance, such as house price busts, are longer and deeper. Recovery from recession on the other hand is faster when credit and house prices grow rapidly, thus indicating the importance of financial factors for the real economy. These findings are consistent with Jorda et al. (2011), who show that so-called financial-crisis recessions, i.e. those recessions that coincide with financial crisis, are more costly in terms of lost output. Moreover, credit intensive growth periods seem to be followed by both deeper normal recessions and financial-crises recessions and slower recoveries. Similarly, Drehmann et al. (2012) establish that recessions that coincide with the downturn of the financial cycle, i.e. a contraction of credit and property prices, are particularly severe. GDP tends to drop by 50% more in a recession coinciding with the downturn of financial factors than a recession during an upturn. The authors reason that this reflects the financial liberalization and changes in monetary policy frameworks since the mid-1980s.

2.3 The global financial cycle

The financial liberalization of the 1980s also ‘globalized’ the financial cycle. Therefore an important characteristic of the financial cycle that has emerged from recent literature is its global nature. Borio (2014) argues that since the global economy has become more integrated with respect to financial, product and input markets, the financial cycle now extends across borders. This period of financial globalization is characterized by the existence of what Rey (2013) calls

4_{Some literature that looks at financial aspects is, for example Bernanke et al. (1999), who develop a dynamic equilibrium model}

(8)

the ‘global financial cycle’.5 This global financial cycle (GFC) can be defined in terms of co-movements in gross capital flows, banking sector leverage, credit creation and risky assets prices such as corporate bonds and stocks.

2.3.1 Existence of a GFC in international capital flows, leverage and credit cycles and asset prices

Looking at four subsets of capital flows, i.e. foreign direct investment (FDI), portfolio equity, portfolio debt and credit, Rey (2013) finds positive patterns of correlations between most types of capital flows (with the exception of FDI flows) into different geographical regions. The commonality in capital flows is particularly strong for credit and portfolio debt inflows. These co-movements can also be associated with global factors. Specifically, movements in the VIX are strongly associated with these flows. Even after conditioning on push factors such as the world short-term real interest rate and world growth rate, Rey (2013) finds a significantly negative correlation between the VIX and capital flows (except FDI which is positive). Forbes and Warnock (2012) investigate this relationship more formally. Analyzing waves of international capital flows, they find that global risk, which can be measured by the VIX, is significantly associated with extreme capital flow periods. Sudden stops in capital inflows, for example, follow an increase in global risk. Conversely, a decrease in global risk predicts surges in capital inflows. Similarly, Passari and Rey (2015) find that cross-border credit flows tend to move in tandem across countries, regardless of the exchange rate regime, and that credit flows tend to rise in periods of low volatility and risk aversion.

Cross-border credit and investment flows also seem to have an impact domestic credit growth. Avdjiev et al. (2012) show that when a crisis arises, higher levels of international credit are linked to a larger contraction in total domestic credit growth and real output in those economies. This suggests that countries financial cycles, specifically credit cycles, are linked across borders and are thus influenced by global factors. Rey (2013) also finds that fluctuations in the VIX seem to be negatively correlated with changes in domestic credit creation and leverage. Certain features of economies, in turn, appear to enhance the influence of capital inflows on domestic credit. Using a dynamic structural model, Gauvin et al. (2017) look at a sample of 31 advanced and emerging economies and find that cross-border capital flows have a procyclical effect on domestic credit to the private sector. Specifically, domestic credit is more sensitive to cross-border capital inflows in countries with lower flexibility of exchange rates, lower GDP per capita and a larger foreign bank presence.

Fluctuations in asset prices also seem to follow the GFC. One would expect that prices of equities, corporate bonds and commodities would be largely determined by country, industry and

5_{Note that the global financial cycle as described by Rey (2013) is related but different from national financial cycles. National}

(9)

company specific factors. However, Miranda-Agrippino and Rey (2015) show that a significant part of the variance of risky asset returns is influenced by a global factor. Using a dynamic factor model with one global factor and a set of regional factors, they show that the global factor reflects the volatility of asset markets as well as the degree of risk aversion in the market. Risk aversion or appetite, in turn, can be related to the leverage of financial intermediaries. The authors find that when global banks are the main investors, aggregate risk aversion and risk premia are low. This implies that risky asset prices are driven by both bank leverage and credit flows and are negatively related to the VIX.

In sum, there is a global financial cycle in both cross-border investment and credit flows, as well as in risky asset prices, domestic credit growth and leverage. Furthermore, the VIX appears to be strongly negatively correlated to capital flows, bank leverage and asset prices. It therefore makes a good proxy of the global financial cycle present in flows and prices. Given that there is a GFC in flows and prices, the question becomes what drives the GFC?

2.3.2 Determinants of the GFC: Monetary policy spillovers

(10)

This conclusion builds on the international role of the US dollar as a funding and investment currency. A large portion of international financial intermediaries portfolios are denominated in dollars, which implies that US monetary policy conditions can be transmitted via capital flows. Particularly global banks and asset managers play an important role in the transmission of financial conditions across borders, as they account for a significant part of these flows. Adrian and Shin (2010) find evidence that financial intermediaries actively manage their balance sheets, so that changes in net worth (through fluctuations in asset prices) result in changes in leverage in the same direction. If, for example, there is an increase in asset prices (or a decrease in measured risk), financial intermediaries increase their leverage by adjusting their balance sheets, contributing to the procyclicality of leverage cycles.

This mechanism also applies to the international context, where global bank leverage is vital in channeling US dollar liquidity to other world markets. Bruno and Shin (2015) show that looser financial conditions are associated with an increase in cross-border capital flows and adjustment in the leverage cycle of international banks. Similarly, Cetorelli and Goldberg (2012) find that global banks respond to changes in US monetary policy by reallocating funds between the head office and its foreign offices, thus contributing to the international propagation of domestic liquidity shocks. Especially European banks have been found to play a pivotal role in the transmission of global liquidity (Shin, 2012). Empirically, Hale and Obstfeld (2016) find that large European monetary union banks’ lending to periphery borrowers increased after the introduction of the Euro and that their lending was linked to their increased borrowing from outside the Euro area. These findings reinforce the importance of cross-border gross capital flows in transmitting monetary and financial conditions and provide evidence for the vital role of global banks in shaping financial conditions.6

Rey (2013, 2016) argues that the fact that US monetary policy appears to be transmitted to the rest of the world invalidates the trilemma, which proposes that if capital is mobile then independent monetary policy is only feasible if exchange rates are floating. She suggests that cross-border flows and leverage of global institutions transmit monetary conditions globally, even when exchange rates are floating. Passari and Rey (2015) come to the same conclusion. Using VAR analysis, they find that US monetary policy has an effect on the external finance

6_{This international transmission channel can be classified into the ‘credit channel’ and the ‘risk-taking’ channel (Rey, 2016). The}

channels can operate parallel to domestic monetary policy and thus affect the independence of monetary policy. Empirically, Ciccarelli et al. (2015) find that through effects on the balance sheets of banks, firms and households, the credit channel amplifies monetary policy shocks on prices and GDP in the US and Euro area. Other studies that find evidence of the credit channel include, amongst others, Ciccarelli et al. (2013), De Fiore and Tristani (2013) and Iacoviello and Minetti (2008). The risk-taking channel, on the other hand, works through the influence of changes in policy rates on risk perceptions and risk tolerance of financial intermediaries. It can operate through the impact of interest rates on asset and collateral values and the impact on investor behavior, i.e. cross border arbitrage (Borio and Zhu, 2012).

(11)

premium of the United Kingdom, although it has a flexible exchange rate. This indicates that a floating exchange regime does not insulate economies as well as previously assumed and as such, foreign monetary policy can be transmitted across borders. Obstfeld (2014) takes a more moderate stance and suggests that exchange rates do insulate economies from foreign financial and monetary shocks to a certain extent. However, he argues that financial globalization does worsen the trade-offs that monetary policy faces and raises the value of additional financial and macroeconomic tools.

2.6 Bond markets and global factors

2.6.1 Bond markets as transmission channel

It appears that international bond markets have become an important transmission channel of foreign financial conditions and global liquidity. Shin (2013) argues that the first phase of global liquidity transmission, from about 2003 to 2008, had global bank leverage and cross-border banking flows at the heart of co-movements in financial conditions. However, he argues that ever since 2010, the second phase of global liquidity uses the bond market as the main stage. This seems especially apparent for emerging market debt securities. Shin (2013) reasons that ‘searching for yield’ by international investors has added to the decline in debt securities’ risk premiums and a rise in international debt issuance. Feyen et al. (2015) investigate this empirically using panel regressions and conclude that global factors have a significant impact on bond issuance by both corporate and sovereign lenders. Additionally, these global factors also seem to have an impact on maturities and yields of bonds at issuance.

The impact of financial globalization on monetary independence of countries has also been explored through the lens of interest rate co-movements. Co-movement of short-term interest rates across countries generally increases with exchange rate pegs, open capital accounts and a higher presence of foreign banks (Goldberg, 2013). However, when exchange rates are flexible there is scope for short-term interest rate independence. Obstfeld (2014) confirms that there is a considerable independence at the short end of the term structure, however long-term interest rates appear to be significantly correlated, especially in advanced economies. The co-movement at the long-end of the term structure could imply a weakening of the effectiveness of domestic monetary policy and a channel for monetary spillovers. For example, yields in all bond markets have the tendency to rise whenever US yields jump. He and McCauley (2013) find that US quantitative easing (QE) policies that reduce the term premium spill over into a reduction of term premiums abroad.

(12)

the policy rate are not significant (both in terms of levels and changes) in explaining changes in interest rates from 2007. This co-movement in term premiums could then possibly be driven by global factors. Consequently, if long-term rates are subject to global forces, the power of short-term rates could diminish. This has important implications for monetary policy.

Monetary policy tries to influence long-term interest rates through its power over short-term interest rates. At the lower end of the yield curve, policy rates are largely independently set by the central bank. However, interest rates at the longer end of the maturity spectrum are a reflection of current short rates, market expectations about future short rates plus a term premium. At longer maturities, monetary policy indirectly influences long-term interest rates through shaping expectations about future rates and inflation, and thus influencing the risk premium on the riskless curve. Importantly, Borio and Disyatat (2011) argue that market interest rates are not only determined by monetary factors but also by financial factors and economic agents’ expectations. This means that international monetary and financial factors could have a significant effect on the determination of market rates. Consequently stronger international linkages between rates, through a global factor, could hamper monetary independence, possibly resulting in the need for sharper adjustments of short rates to achieve desired outcomes.

2.6.2 Determinants of long-term interest rates

Investigating the determinants of term interest rates has its basis in the importance of long-term interest rates in the delong-termination of business and housing investment as well as household spending. Interest rates have an important influence on the business cycle and capital formation, and are therefore a key link in monetary policy transmission (Howe and Pigott, 1991). Existing models and literature on determinants of long-term interest rates have focused on a wide range of variables, but particularly emphasize the importance of fiscal balances and government finance variables as important determinants of interest rates. Most studies in this context tend to find a positive relationship between government debt levels and long-term bond yields.

(13)

over to long-term yields in both emerging and advanced countries.7 This spillover effect operates mostly through changes in the global risk free rate and local spreads.

Although sovereign debt has been found to have a positive impact on government bond yields, some researchers do not find a significant or positive relationship between fiscal variables and long-term interest rates. Contrary to most empirical literature that use static estimation methods, Dautovic (2017), also uses dynamic estimates in the empirical estimations. Using a panel of 20 OECD countries, he contrasts the findings of the static and dynamic models. The findings for the static model are in line with the previous literature, however the dynamic model does not show a significant relationship between fiscal balances/debt-to-GDP ratio and long-term interest rates. Furthermore, the fact that US government bond yields, for example, have remained low despite rising indebtedness following the financial crisis also challenges previous findings. Investigating this relationship, using a single country setting, Akram and Li (2017) show that, contrary to pervious findings, higher government debt in the US has in fact a negative effect on long-term sovereign bond yields in the long run.8 However, they do find a positive relationship in the short run.

Besides sovereign debt levels, some studies have also found significant effects of exchange rate regimes and exchange rate risks on long term interest rates. Focusing on a sample of Scandinavian countries, Hol (2006) finds that long-term interest rates are also determined by currency risk and exchange rate regimes. International conditions in Europe, such as international debt and unemployment, also appear to be significant in explaining long-term interest rates. Gadanecz et al. (2014) investigate the effect of exchange rate risk on sovereign bond yields in emerging market economies and find a significant causal effect from exchange rate risk to local currency government yields. An increase in uncertainty about future exchange rates, leads investors that are exposed to these currencies to require a larger compensation in form of higher bond yields. The authors also include international factors such as the VIX and US term premia in the model and find a significant effect of these global factors on emerging market local currency bond yields.9 Thus given that long-term interest rates may also be subject to global forces, it could be of interest to utilize a cross-country panel framework to investigate the sensitivity of long-term interest rates to these global conditions.

7_{The model also controls for a range of domestic factors, such as short-term interest rates, measures of financial development,}

real GDP growth, and exchange rate regime amongst others.

8_{According to the authors, a plausible explanation, for this finding is that the US central bank credits reserves in accordance with}

an increase in government spending, which leads to an increase in the amount of banks reserves. A rise in the amount of reserves held by banks puts downward pressure on government bond yields, if monetary policy does not counter this effect with changes in short-term rates.

9_{Their results are consistent with Miyajima et al. (2015) who find a negative relationship between the VIX and emerging market}

(14)

2.7 Hypotheses

Against this backdrop and extending the scope of long-term interest rates literature, this paper seeks to investigate the sensitivity of long-term interest rates to domestic credit conditions, global financial cycle and US monetary policy, while controlling for country specific macroeconomic determinants. The hypotheses build on the main aspects gained from the literature: (1) There is a GFC in capital flows, credit growth, leverage and asset prices, (2) US monetary policy can be transmitted across borders, (3) domestic credit is influenced by cross-border credit flows and global bank leverage cycles and is thus not fully under the control of domestic monetary policy, and (4) the bond market may be an important stage for the international transmission of financial conditions.

The global nature of the financial cycle and its transmission into domestic markets implies that domestic macroeconomic variables may be partly determined in world markets and are thus essentially beyond the reach of domestic policymakers. And indeed, the high correlation of long-term interest rates across countries suggests that global factors may be of importance. The paper hypothesizes that if long-term interest rates are sensitive to the global financial cycle (proxied by the VIX) and US monetary policy, monetary policy may be constrained and requires global considerations when setting policy. Also if cross-border capital flows enhance the sensitivity of long-term interest rates to US monetary policy, central banks may not be able to run fully independent monetary policies when capital flows are freely mobile.

Domestic credit is a variable of interest, as credit is an indication of domestic liquidity conditions and can be used as a proxy for the domestic financial cycle (Borio, 2014). Domestic liquidity conditions could provide a signal to investors about present and future economic conditions and therefore affect their appetite for government bonds. Domestic financial cycles also appear to have been partially subsumed by the global financial cycle and are thus affected by global conditions (Rey, 2013). As monetary policy not only affects the general level of interest rates but also has an impact on credit demand and supply through the so-called credit channel10, the influence of global factors on domestic credit could hamper the effectiveness of monetary policy. Consequently, the paper aims to test the following hypotheses:

(1) Long-term interest rates are sensitive to global factors, i.e. global market uncertainty (VIX) and US monetary policy (US federal funds rate).

(2) Long-term interest rates are sensitive to changes in domestic liquidity conditions, i.e. domestic credit growth, which is partly influenced by global conditions.

10_{According to the credit channel theory, direct effects of changes on policy rates are amplified by endogenous changes in the}

(15)

3. Methodology

This section describes the econometric approach used to analyze the proposed research questions. The paper will loosely follow the framework of Feyen et al. (2015), who look at the impact of global factors on the pricing and maturity of bonds at issuance, and Dautovic (2017), who looks at the effect of fiscal policy on long-term interest rates. Instead of examining the impact of global factors on sovereign bond prices at issuance or fiscal policy on long-term interest rates, this paper will attempt to use the same panel fixed effects setting to gage the sensitivity of long-term interest rates to global market uncertainty, US monetary policy and domestic credit growth.

3.1 Econometric technique

Generally, papers that attempted to uncover the determinants of long-term interest rates have used panel fixed effects models with clustered standard errors.11 This specification allows for variation within individual countries and thus captures country specific effects that might influence long-term interest rates not controlled for in the model. Feyen et al. (2015) use panel fixed effects regression with clustered standard errors to evaluate the impact of global factors on the pricing and maturity of bonds at issuance. Dautovics’ (2017) analysis is set within a panel data framework with both country and time fixed effects, where the estimates are acquired through the least squares dummy variable estimator (LSDV). Gadanecz et al. (2014), for example, is another paper that uses a panel framework with country fixed effects and clustered standard errors. The authors examine the sensitivity of local currency sovereign bond yields in emerging economies to exchange rate risk, while controlling for a range of domestic factors, but also international factors, namely the VIX and the US 10-year term premia.

The empirical method has also been used to test the effects of global variables on various other domestic factors. For example, Rey (2013) uses a panel fixed effects model to gage the effect of the global financial cycle (proxied by the VIX) on domestic asset markets and bank leverage. She also examines if capital flows matter for the sensitivity of these domestic variables to the global financial cycle. Bruno and Shin (2015) on the other hand investigate the global banking channel. Using panel regressions with country fixed effects, they examine if the growth in cross-border loans is affected by global factors, in this case global bank leverage. Although these studies focus on different variables of interest, the same framework could be used to gage the impact of global and domestic financial factors on long-term interest rates.

11_{Other papers that have used this empirical method include, for example Alper and Forni (2011), who examine the possible}

(16)

3.2 Limitations of the specification

This approach to modeling is subject to criticisms. In general, panel data studies add the cross-sectional dimension to the data, which improves statistical inference. However, a disadvantage is that panel regression assumes homogeneity, so that long-term interest rates across countries are assumed to respond similarly to changes in economic fundamentals and global influences. This assumption is somewhat relaxed by introducing country fixed effects. Including time fixed effects would allow to control for the possibility of unobserved global shocks that induce higher interest rates everywhere. However adding time fixed effects in long-run regressions may induce multicollinearity, as long term rates tend to be highly correlated among themselves (Obstfeld, 2014). Thus the specification will only include country fixed effects.

Some empirical studies examining the global financial cycle and spillover of US monetary policy to domestic markets have used VAR analysis to make causal statements.12 Panel regression of the nature used in this paper however cannot make strong causal statements, due to the endogeneity between long-term interest rates and especially domestic variables. The endogeneity problem stems from both the possible reverse causality running from the dependent to the independent variables or due to a (unknown) common causal factor that could influence both the dependent and independent variables. Especially domestic variables and interest rates may be negatively linked through the business cycle. For example, during a recession long-term interest rates fall due to monetary easing and at the same time fiscal conditions worsen due to automatic stabilizers. This endogeneity may cause biased coefficients.

One method to correct for this bias and to be able to make causal statements is to use Instrumental Variables (IV). However, it may be quite difficult to find instruments for financial variables that satisfy the IV conditions of exclusion restriction and strong correlation. Another way to correct for the bias and reduce endogeneity concerns is by lagging independent variables by one year as done by Bruno and Shin (2015). Laubach (2009) on the other hand suggests using forecast variables, and argues that using 1-year ahead forecasts of domestic macro variables partly solves the endogeneity problem that arises when estimating reduced form bond yield equations. This method is also applied in the more recent literature looking at determinants of long-term interest rates.13 Furthermore, including forecast variables also seems natural since financial market participants use forecasts of various determinants of bond prices to decide how many government bonds to hold (Ichiue and Shimizu, 2015). Nevertheless a common causal factor could still influence both forecasts and bond yields, since the current state of the business cycle likely still impacts forecasts to a certain extent.

12_{See for example Rey (2013), Passari and Rey (2015), and Miranda-Agrippino and Rey (2015).}

(17)

With these limitations in mind, the specification will use forecasts where data is available. Thus in order to reduce endogeneity concerns, the paper uses 1-year ahead forecasts of short-term interest rates, inflation, real GDP growth and government debt. Although reverse causality between global factors and long-term interest rates is unlikely, all other independent variables are included as lags.14 This specification also displays the lowest overall magnitude of multicollinearity.15 Multicollinearity may be an issue, as financial and economic variables tend to move together in systematic ways. For example, capital flows, and credit and leverage cycles tend to move with the VIX and are closely correlated (Rey, 2013). Also, movements in the US federal funds rate tend to change the VIX, global bank leverage, cross-border credit and domestic credit cycles (Miranda-Agrippino and Rey, 2015). The complexity of the relationships between variables has to be kept in mind when interpreting results using this linear specification.

3.3 Specification

The regression will use two sets of explanatory variables. The variables included in the specification are extracted from the reviewed literature. The first set includes global factors that proxy for global financial conditions. These include the VIX, the effective US federal funds rate, the number of global banks, and a subset of capital flows. The US federal funds rate has been shown to be a determinant of the global financial cycle. Also US monetary policy seems to spill over into other countries, through cross-border capital flows and global banks. Global banks are included, as they seem to propagate global financial conditions, particularly through cross-border banking flows (Bruno and Shin, 2015). The VIX is a measure of market uncertainty and risk appetite and can be used as a proxy for the global financial cycle due to its strong negative correlation with capital flows, bank leverage and domestic credit growth (Rey, 2013). The VIX is also said to be an indirect key indicator of the willingness to provide funding (CGFS, 2011). The second set of variables includes domestic macroeconomic and financial factors. The domestic control variables include real GDP growth, the government debt to GDP ratio, nominal effective exchange rate and a financial development measure. Additionally, short-term interest rates and inflation are included to account for domestic monetary policy. The change in credit to the non-financial private sector is included to provide information about domestic liquidity conditions. Structural changes in the financial system have enhanced the role of credit (Schularick and Taylor, 2012) and large credit growth seems to be linked to deeper and longer lasting recessions (Aikman et al., 2015; Drehman et al., 2012 & Jorda et al., 2011). Thus credit growth can be seen as the outcome of the global transmission channel, i.e. it is affected by global

14_{This specification is preferred after checking multiple combinations of variable transformations (using lags on all independent}

variables vs. using forecast and only lagging domestic variables vs. using forecast and lagging all other independent variables). Coefficients between the specifications vary slightly, however the significance of the variables of interest does not change. The size of the coefficients in the last specification is smallest and more reasonable.

15_{The variance inflation factor (VIF) is the lowest, where the mean VIF is 1.72, well below the commonly used threshold of 10}

(18)

leverage and market uncertainty (Rey, 2013), and also cross-border capital flows (Gauvin et al., 2017) and may thus serve as a link to long-term interest rates.

The research is conducted within a sample of 24 OECD countries16. The paper will use annual data, which covers the period from January 2000 to December 2015. The country coverage and time series sample are largely determined by data availability and the reasoning that co-movements in long-term interest rates are more pronounced for advanced economies. The panel regressions are with country fixed effects and clustered standard errors at the country level17. All variables are lagged by one period, except those domestic variables where forecasts were available. The benchmark specification is as follows:

IRLit = αi + εVIXt-1 + ηUSIRt-1 + ΩBANKi,t-1 + γΔCREDITi,t-1 + κIRSi,t+1 +

πINFLATIONi,t+1 + λGDPi,t+1 + µDEBTi,t+1 + θFDi,t-1 + ξNEERi,t-1 + uit

, where i denotes the country, t is time and uit denotes the error term. To test if cross-border capital flows enhance the sensitivity of long-term interest rates to the US federal funds rate, the paper adds ϒFLn=3,i,t-1, denoting a subset of three types of capital flows, to the benchmark

specification and interacts them with the US federal funds rate.

In the model the dependent variable IRLit is a vector of long-term interest rate estimates, for

country i in year t. The interest rates are implied by prices at which 10-year government bonds are traded on financial markets, thus measuring the appetite of investors for government bonds. The dummy variable αi captures the country fixed effects.18 Country fixed effects control for any

country level specific effects not captured by the control variables, such as the convergence of bond yields in the Euro area following the introduction of the common currency.19

(A) Global factors

VIX: Measure of the implied volatility of S&P 500 index options, i.e. the equity market. The VIX is often used as a proxy for global financial conditions, specifically for global market uncertainty and risk aversion. Higher levels of the VIX, thus higher level of market uncertainty, are generally

16_{The countries are: Australia, Austria, Belgium, Canada, Czech Republic, Denmark, Finland, France, Germany, Greece,}

Hungary, Ireland, Italy, Japan, Korea, Luxembourg, Netherlands, Norway, Poland, Portugal, Spain, Sweden, United Kingdom, United States

17_{This is the preferred specification. The Hausman test indicates that fixed effects are appropriate. Furthermore residual scatter}

plots of variables indicate that the standard errors should be clustered at the country level.

18_{The regression does not include a period dummy to control for the financial crisis of 2008/09. In unreported regressions, a}

dummy variable for the financial crisis years of 2008 and 2009 is included in the baseline specification, but it does not change the significance of the results.

19_{The term structure of Euro-zone countries was highly correlated with those of Germany during the period from the introduction}

(19)

associated with a flight to quality, specifically to long-term government bonds. As investors protect their portfolio against volatility and allocate more investments to bonds, the market price

of bonds will rise. The expected sign of the coefficient will therefore be positive.

USIR: United States (US) effective federal funds rate (%, annual average). The US federal funds rate is used as a proxy for US monetary policy, which can be said to represent global monetary policy conditions. This selection is supported by the special status of US dollar in international finance and the evidence that US monetary policy tends to spill over into domestic monetary policy. Expansionary US monetary policy, i.e. a lower policy rate, for example could signal improving global economic conditions, as investment is expected to pick up and credit becomes cheaper. Investors then prefer to hold riskier assets that provide them with higher returns, causing them to switch away from safe assets, such as government bonds. This causes bond prices to fall. The expected sign of the coefficient is therefore positive.

BANK: The percentage of foreign banks amongst total banks (%). Global banks20_{are included, as} they seem to be the propagators of global financial conditions. In general, banks hold interest earning securities such as government bonds or other safe securities as secondary reserves. Next to required reserves, these secondary reserves allow banks to pay their liabilities and maintain a relatively balanced balance sheet, thus are important in managing risk. A higher number of global banks may bid up government bond prices, as foreign banks would want to hold more government bonds to compensate for the increased risk of operating in a foreign market. The coefficient of global banks is consequently expected to be positive.

FL: Denotes a subset of capital flows: portfolio equity, portfolio debt, and cross-border credit flows (as % of GDP). These are included to test if the sensitivity of long-term interest rates to US monetary policy is enhanced by cross-border capital flows. Large cross-border credit, equity and bond portfolio flows integrate world asset markets and are expected to transmit and intensify US monetary policy spillovers.

(B) Domestic factors

CREDIT: Change (log difference) 21in total credit from all sources to the private non-financial sector. Private sector credit is considered a financial variable that gauges domestic liquidity conditions, but is also an indicator of economic activity. Improved economic conditions usually go hand in hand with greater credit growth, which can reduce bond prices. Greater credit growth could increase investors’ willingness to hold riskier assets, reducing their demand for fixed income assets such as long-term government bonds. Thus the expected sign of the coefficient will be negative.

20_{Global banks are banks with active operations spanning across multiple countries, regions and/or continents.}

21_{The log difference is an approximation for the rate of change of the variable. The paper will henceforth refer to the log}

(20)

IRS: The short-term interest rate (1-year ahead forecast, %) is included to control for the impact of monetary policy on long-term interest rates. Long-term rates are determined by short-term rates, the expected future level of short-term rates and market uncertainty. Expansionary domestic monetary policy (lower short-term rate) is expected to increase investment, thus signaling improving economic conditions. Investors’ appetite for long-term government bonds will decrease as they prefer to allocate a larger portion of their portfolios to more risky assets with higher returns, such as equities. It is therefore expected that the sign is positive.

INFLATION: Higher consumer price index (CPI) inflation (1-year ahead forecast, %) reduces the purchasing power of a bonds future cash flow. Thus with higher current rates of inflation and higher expected inflation investors demand higher yields to compensate for inflation risk, reducing bond prices. The expected sign of the coefficient is negative.

GDP: The real GDP growth (1-year ahead forecast, %) rate is used as a proxy for investment opportunities. Investors may prefer to hold long-term government bonds if lower economic growth is predicted to safeguard their portfolios, resulting in higher demand and prices of long-term government bonds. The expected sign of the coefficient is negative.

DEBT: General government gross financial liabilities as a percentage of GDP (1-year ahead forecast, %). Generally, as government debt levels rise and fiscal conditions worsen, investors require a larger premium to compensate for the increased default risk, which increases yields and reduces bond prices (Ichiue and Shimizu, 2015). The expected sign of the coefficient is negative. FD: Financial development index (Index from 0 to 1, where 1 is the highest measure of financial development)22. This index is included to control for the financial development of a country. In general, higher financial market development boosts countries resilience and promotes financial stability, where deep and liquid financial systems can help reduce the adverse impact of shocks (Sahay et al., 2015). One could expect then that investors prefer to hold the government bonds of countries with developed financial markets, which drives up sovereign bond prices of these countries. Thus the expected sign of the coefficient is positive.

NEER: The nominal effective exchange rate (yearly averages; 2010=100) is included to control for exchange rate risk. A rise in the NEER represents a depreciation of the domestic currency. As the domestic currency depreciates against the US dollar, bond yields are expected to rise to compensate for higher currency risk, causing bond prices to fall. The expected sign is negative.

22_{The new broad based Index of financial development developed by the International Monetary Funds (IMF) prepared by}

(21)

4. Data and descriptive statistics

4.1 Data Sources

The data are drawn from multiple sources. The long-term interest rates data series are from the OECD database. The data series of the VIX and the US federal funds rate are from FRED (Federal Reserve Bank of St. Louis). The data series for global banks are from two IMF papers, Van Horen and Claessens (2012) and Claessens and Van Horen (2015). The series from 2000 to 2004 are taken from the first paper, and the series from 2005 to 2013 from the second.23 The data series of global banks does not necessarily only represent large global banks but can also include branches of smaller regional banks that might not propagate the global financial climate to the same extent to their host countries. The series of annual external capital flows data (portfolio equity and debt as % of GDP), are taken from the paper by Alfaro et al. (2014) who extract data from the International Financial Statistics (IFS) database of the IMF. Cross border credit flows data are extracted from the Bank for International Settlements (BIS) locational statistics database and are calculated as a % of GDP.

The data series of total credit to the private non-financial sector are extracted from the BIS database. The financial development index, prepared by Svirydzenka (2016), is extracted from the IMF database. The data ranges from 2000 to 2014, thus this paper assumes the value for 2015 is the same as in 2014. The short-term interest rate and government debt forecasts are extracted from each December edition of the OECD economic outlook. Inflation and real GDP growth forecasts are drawn from the IMF World Economic Outlook24. The nominal effective exchange rate series are taken from the BIS broad effective exchange rate indices database. Data series that are given in monthly or quarterly frequency have been averaged (simple average) to obtain annual frequency.

4.2 Descriptive statistics and Correlations

The analysis uses annual panel data of multiple financial and macroeconomic variables from 24 OECD countries over 16 years, i.e. from 2000 to 2015. The dataset is unbalanced, as data points for some country specific variables are missing. Table A1 provides summary statistics. The average long-term interest rate is 4.1%, with the minimum being at 0.34% and the maximum at 22.49%. The highest long-term interest rates belong to Greece in years 2011 and 2012, after the start of the European debt crisis in 2010. The distribution of long-term interest rates is negatively skewed, where the bulk of interest rates range between 3% and 6%. The mean of the US federal funds rate is 1.8%, where the minimum is 0.8%. The US federal funds rate fell to zero territory

23_{The data series for Poland is mismatched between the two papers, thus instead data from FRED is used, which fits better with}

the first paper of Van Horen and Claessens (2012).

(22)

after 2008 and has remained low for the rest of the sample period. These low values are in line with monetary easing after the global financial crisis of 2008/09. The mean of the VIX is 21 index points, with a minimum of 13 and maximum of 33, which is in 2008, the year of Lehman Brothers collapse. While risk sentiment has somewhat recovered after 2009, the VIX remains quite volatile, with relatively high spikes in 2010 and 2011. The number of foreign banks as percentage of local banks has a large spread, where the minimum is at 0% and the maximum is 99%. This shows that there is quite a substantial difference within the sample of countries. Turning to domestic variables, the change in domestic credit is only slightly negatively skewed, but exhibits a relatively large variance, probably due to the global financial crisis. The US has the largest positive changes in domestic credit from 2005 to 2007, in line with massive inflows in the pre crisis period. The largest reductions in domestic credit in the US were in 2009 and 2010. Japan has the largest reduction in domestic credit within the sample period, between the years of 2013 and 2015. The mean of the government debt ratio forecasts is 72%, with a minimum and maximum values of 3.7% and 233%, respectively. The highest government debt ratio belongs to Japan and the lowest to Luxembourg. Inflation rate forecasts range between -2.5% and 7.5%. The distribution is relatively normally distributed around the mean of 2.08%. The median value is at 2.00%. Short-term interest rates on the other hand are negatively skewed, where the bulk of observations are at the lower end of the spectrum with only a few high observations, belonging to Hungary and Poland. GDP growth forecasts have a mean of 2.29%, with a minimum of -4% and a maximum of 7%. After the financial crisis however most countries had relatively low or negative growth expectations.

Table A2 reports correlations. Short- term rates have the highest correlation with the dependent variable, as expected. The US federal funds rate is also quite strongly positively correlated with long-term interest rates. Domestic credit growth also appears to be relatively highly correlated with the dependent variable. Looking at pairwise correlations of the independent variables reveals that the US federal funds rate and short-term interest rates are quite highly correlated, possibly indicating that domestic monetary policy in the sample countries is set in accordance with US policy. This may not be surprising, seeing as Miranda-Agrippino & Rey (2015) find that a contractionary move in the US is likely to be followed by stricter monetary policy in both the UK and Euro area, which make up the majority of the sample. Domestic credit growth is also quite strongly correlated with the US federal funds rate, possibly indicating its cross-border influence on domestic credit growth.

(23)

(24)

5. Regression results

This section empirically addresses the research questions posed in the previous sections. It summarizes and discusses the main empirical results for the sensitivity of long-term interest rates to the VIX, US federal funds rate and domestic credit growth. To control for heteroskedasticity and autocorrelation of the residuals all estimations are run with cluster robust standard errors, clustered around countries. All variables are logged. The estimation results are summarized in Table 1 and the benchmark specification is given in Model 2.

The first model only includes domestic factors and reveals that short-term interest rates and inflation significantly determine long-term interest rates. The signs of the coefficients are as expected, where lower expected short-term rates and higher expected inflation levels reduce long-term interest rates. These variables remain statistically significant throughout Models 2 to 5, which include global factors, indicating that domestic monetary policy has a significant effect on long-term interest rates. However, the size of the coefficients reduces somewhat when global factors are added. This is possibly due to correlations with the global variables or the combination of global factors. For example, short-term interest rates are relatively highly correlated with the US federal funds rate. The adjusted R2 rises from 0.601 to 0.730 when including global factors in Model 2, indicating that the benchmark model fits the data better. The government debt ratio becomes statistically significant in Model 2. The sign of the coefficient of the government debt ratio is positive and is therefore not in line with general predictions. This result may be explained by the fact that long-term interest rates and government debt levels are endogenous variables, connected through the business cycle. For example, debt levels are not only determined by fiscal policy but could also be influenced by automatic stabilizers in a downturn. When a downturn in economic activity occurs, central banks would apply expansionary policy and thus decrease short-term rates. If, however, economic conditions are expected to remain low in the near future, this will also reduce long-term interest rates, which in turn would automatically lower government debt levels (Dautovic, 2017).25

In the benchmark model (column 2) domestic credit growth becomes statistically significant in explaining long-term interest rates. This change in significance probably partly relates to the fact that credit growth is correlated with the combination of these global factors. As the literature revealed, domestic credit growth can be influenced by US monetary policy and global financial conditions through the behavior of global banks. An improvement in international conditions, for example, could increase leverage and credit supply abroad by global banks, which would raise

25_{Although the inclusion of forecasts is expected to reduce the endogeneity issues somewhat (Laubach, 2009), controlling further}

(25)

collateral values and reduce risk perceptions in recipient countries, resulting in increased domestic credit supply. Model 2 shows that a rise in domestic credit growth is associated with a reduction of long-term interest rates, indicating that improving domestic liquidity conditions reduce investors appetite for government bonds, as investors prefer to hold riskier assets.26 The benchmark regression, Model 2, points towards the importance of global factors. The VIX is highly statistically significant in explaining long-term interest rates. The positive coefficient of the VIX suggests that higher market uncertainty and risk aversion in the market increase bond prices as investors allocate more of their portfolio to safe assets such as long-term government bonds.27 A 1% rise in the VIX increases long-term interest rates by about 0.34%. This suggests that long-term interest rates are statistically significantly sensitive to changes in market uncertainty and risk appetite. The results imply that while central banks can set short-term rates to influence long-term interest rates, developments in global markets also appear to have a significant effect, which may constrain domestic monetary policy.

Long-term interest rates also seem to be sensitive to US monetary policy. The coefficient of the US federal funds rate is positive and statistically significant, indicating that a 1% decrease lowers long-term interest rates by about 0.15%. This suggests that a lower US policy rate signals improving global economic conditions to investors, which reduces their appetite for long-term government bonds and rather seek out riskier assets with higher returns. The result implies the existence of international monetary spillovers from the US to other countries, which is in line with the findings of Miranda-Agrippino and Rey (2015), Rey (2013) and Passari and Rey (2015). The fact that US monetary policy appears to influence long-term interest rates makes it difficult for domestic monetary policy to insulate their economy from foreign monetary shocks. This worsens the trade-offs that monetary policy faces and may require sharper adjustments of short-term rates to achieve desired outcomes, possibly raising the value of additional monetary tools. Asset market uncertainty (VIX) and US monetary policy also seem to be closely linked, where expansionary US policy has been found to decrease risk aversion (Bekaert et al., 2013; Rey, 2013). To control for the combined effect of the VIX and US monetary policy, an interaction term is included in Model 3. The combined effect of the VIX and US federal funds rate is statistically significant, indicating that a 1% rise in the VIX decreases the transmission of the US federal funds rate to long-term interest rates by 0.16%, and vice versa. Including the interaction effect significantly raises the magnitude of the US federal funds rate, indicating that the coefficient is biased downwards when the estimation does not control for their combined effect in Model 2.

26_{One has to be cautious with the interpretation and rather see it as proof of a correlation and not as a causal statement due to the}

possible endogeneity. The robustness test using system GMM in section 6.1 provides more clarity.

27_{It is worth noting at this point that the sample choice, i.e. advanced OECD countries, may partly drive the sign of the VIX. As}

(26)

Global banks also seem to have a statistically significant effect on long-term interest rates, albeit of lower significance than the other two global factors. A larger number of foreign banks seem to increase long-term interest rates, where a 10% rise in the amount of foreign banks to total banks in the economy increases bond prices by 1.77% (Model 2). One could speculate that due to their global reach and therefore somewhat higher risk exposure, foreign banks prefer to hold a larger amount of government bonds to safeguard their balance sheets, which bids up the prices of bonds. Their significance could also imply their importance in channeling global conditions to domestic markets.

Both global bank leverage and cross-border banking flows have been found to be vital in transmitting financial conditions (Bruno and Shin, 2015). Cetorelli and Goldberg (2012), for example, find that global banks reallocate funds between the head office and its foreign offices in response to changes in US monetary policy, which propagates liquidity shocks internationally. Also, Rey (2013) finds that global banks adjust their leverage in response to changes in the VIX, which in turn affects cross-border and domestic credit growth. To test if a large number of global banks increase the sensitivity of long-term interest rate to global conditions, Model 4 includes interaction terms between global banks and the VIX and US federal funds rate respectively. While the interaction effects are not significant, isolating the combined effect of global banks and the VIX and US federal funds rate appears to increase the magnitude of the VIX and the US federal funds rate, as well as global banks. Not controlling for their combined effects in Model 2 biases the coefficients of these global factors downwards.

Model 5 tests if cross-border capital flows enhance the sensitivity of long-term interest rates to US monetary policy, i.e. the US federal funds rate. To this end, the benchmark specification includes three subsets of capital flows: portfolio equity, portfolio debt, and cross-border credit flows (as % of GDP) and interacts them with the US federal funds rate. Only cross-border credit flows seem to significantly change the transmission of US federal funds rates to long-term interest rates.28 However, contrary to what was predicted, i.e. credit flows enhance transmission, the interaction effect is negative and statistically significant.This indicates that larger cross-border credit flows reduce the sensitivity of long-term interest rates to the US federal funds rate. A 1% increase in cross-border credit flows dampens the transmission of the US federal funds rate to long-term interest rates by about 0.04%. One could speculate that investors’ appetite for government bonds is less sensitive to US monetary policy, if there is an increase in domestic liquidity, via cross-border credit flow. 29

28_{This is in line with the findings of Rey (2013), Miranda-Agrippino and Rey (2015) & Passari and Rey (2015) who find that}

especially cross-border credit flows transmit US monetary policy conditions.

29_{This result may also be partly driven by the negative relationship between the US federal funds rate and cross-border credit}

The sensitivity of long-term interest rates to global factors and domestic credit growth