I. Introduction

“Securitization had introduced a higher degree of uncertainty, creating non-linearities along the yield curve that impaired monetary policy effectiveness.”

—Christian Noyer, Governor of the Banque de France (“Challenges for Monetary Policy from Financial Innovation and Globalization,” IMF (European Office and Monetary and Capital Markets Department)/Bank of England/Banque de France Conference, Paris, January 29, 2008)

Increased sophistication of financial intermediation can change the nature of macro-financial linkages, often with considerable policy implications. The influence of monetary policy on interest rates and the real economy depends critically on the structure of the financial system and the maturity of capital markets. In particular, asset securitization can affect or permanently alter the transmission mechanism of monetary policy that ultimately permeates to the real economy by influencing consumption and investment decisions.

Financial innovation, specifically through securitization, provides an essential impetus for a more efficient allocation of capital. For instance, asset securitization has evolved into a widely-used capital market-based funding mechanism in many developed countries to mitigate balance sheet mismatches, remedy financing constraints, and optimize funding costs. Securitization offers issuers more flexibility to create securities with distinct risk-return profiles across the maturity structure to facilitate the unbundling, transformation and diversification of financial risks associated with different (and mostly illiquid) asset types.² This customization of risks according to the preferences and tolerances of agents improves the capacity of the financial system to bear risk and intermediate capital (Jobst, 2007 and 2008). In emerging market countries, securitization can also support local capital market development, facilitate investments in largely unexplored areas of economic activity, and expand the spectrum of financing options to finance housing and consumer deficits.

That said, financial innovation is not without inherent risks, and recent events suggest some reservations about the positive assessment of securitization. The evolution of structured finance illustrates that the risks from financial innovation have the potential to transpire in greater and more harmful ways than many other sources of risk from conventional finance instruments. Different forms of financial innovation, such as asset securitization, are bound to challenge the existing market order and, thus, if proven sustainable, have the potential to permanently alter the transmission mechanism of monetary policy and the way they affect the real economy by influencing consumption and investment decisions. The ongoing market turbulence in the wake of the recent U.S. sub-prime mortgage crisis (and attendant dislocations in money markets in tandem with the failure of major brokerage firms and commercial banks with excessive bad debt write-offs) highlights how securitization mechanisms—together with poor credit origination standards, lack of regulatory foresight and lax supervision—can perpetuate market disruptions, with potentially adverse consequences for employment, growth, and welfare. These dislocations are created when regulation and transparency lag behind, while the originators incentives to ensure minimum standards are compromised.³

The current regulatory debate centers on considerable skepticism about the incentive structure of securitization, which might have encouraged issuers to cutback on screening and monitoring borrowers, resulting possibly in a systematic deterioration of lending and collateral standards. The fallout from the U.S. sub-prime market crisis fueled mounting regulatory unease about current risk measurement standards of structured finance products.⁴ While securitization has greatly enhanced the secondary mortgage market in many countries, giving originators greater balance sheet flexibility (The Economist, 2008), the ability to securitize mortgages has also affected lender behavior, with lending standards experiencing greater declines in areas with higher mortgage securitization rates (Dell’Ariccia, 2008). However, recent studies also find that financial innovation has contributed to rising default rates and increased credit supply fuelled by poor screening of mortgage lenders intent on selling on loans in structured finance transactions (Mian and Sufi, 2008; Keys et al., 2008).

Besides financial stability, a key concern for policymakers is that structured finance has complicated monetary policy. As financial intermediation becomes more competitive as a result of greater capital market sophistication (particularly in EM countries), the macro-financial linkages are also likely to change, often with considerable implications for the execution of monetary policy. Given the greater complexity of financial products, the diversity of financial institutions, and the growing interdependence of financial markets, policy makers are indeed challenged in deciphering the implications of the highly sophisticated new derivative and structured financial products. In particular, they are now beginning to worry about how asset shocks in the structured finance market propagate across the financial system and how possible knock-on effects on bank balance sheets might affect the execution of monetary policy. As indicated by Jan Marc Berk of the Dutch central bank, “financial innovation has increased the supply of substitute sources of financing, providing more hedging opportunities, and reducing the impact of changes in policy interest rates (IMF, 2008).” In fact, the availability of alternative funding sources through financial innovation has helped substitute capital market finance for traditional bank deposits, which has changed the traditional capital structure of banks to a point where the “originate-to-distribute” banking model encouraged a disproportionate expansion of total assets.⁵ This situation suggest greater resilience of credit supply to changes in monetary policy in an environment of cheap credit and excess liquidity at higher leverage.

In this paper, we examine how securitization—as an essential element of structural change in the financial sector over the last 25 years—has affected macro-financial linkages, notably in the transmission of monetary policy through both the interest rate and the credit channel.⁶ Proponents of the latter claim that the traditional (interest rate) channel of monetary transmission alone cannot explain the magnitude, timing, and compositional effects of monetary policy actions on the real economy (see Figure 1). Instead, credit market frictions, such as the enhanced role of capital market finance relative to the significance of the banking sector in a financial system engender a “financial accelerator” mechanism (Bernanke and Gertler, 1995) that increases the impact of monetary policy on market rates while expanding the amount of loanable funds to banks (due to higher balance sheet leverage) to satisfy money demand. Without securitization, banks would lack adequate substitutes on both the asset and liability side of their balance sheets to avert the contraction of loan supply in response to monetary tightening.⁷ However, faced with a more restrictive monetary policy, banks can sell assets through securitization (as off-balance sheet operation) to fund new lending. While securitization together with the emergence of “near-banks” and the shortening of the term of market liabilities does not appear to undermine monetary policy transmission a priori, it attributes greater importance to alternative funding channels, which debilitate the timeliness of output response to changes in monetary policy as financial institutions can lend more by increasing quantities rather than raising interest rates.⁸

View Full Size

Transmission Channels of Monetary Policy in the United States

Citation: IMF Working Papers 2009, 026; 10.5089/9781451871739.001.A001

• Download Figure
• Download figure as PowerPoint slide

More specifically, we test two basic hypotheses in this context:

• Securitization can enhance the transmission channel of monetary policy to short-term interest rates (“early and middle linkage”) (see Figure 1). Since the financing cost for securitized assets is likely to be more closely aligned with market interest rates than the capital cost of bank lending⁹, developing a liquid and tradable MBS market can potentially strengthen the link between the policy rate and market interest rates.

• Securitization can affect (quantity measures of) monetary policy transmission (“final linkage”) by weakening the impact of interest rate changes on aggregate demand through the credit channel. Securitization breaks the balance sheet link between the banks’ deposit base and lending activities by creating an additional window of funding (without increasing the balance sheet of lenders). If banks are able to resort to such alternative forms of external finance, securitization can weaken the transmission effect via the bank balance sheet, increase the resilience of banks to interest rate shocks, and eventually lead to a lower interest rate elasticity of real output. This effect is likely to be more pronounced in countries which have a bank-dominated financial sector (and rapidly expanding securitized issuance).

In this paper, we first analyze the influence of securitization on monetary policy in an economy with sophisticated financial markets and then apply a more tailored methodology to examine similar dynamics in the EM context. We focus on the U.S. mortgage market, where macro-financial linkages from securitization are generally better understood. Given the sophistication, depth, and liquidity of the U.S. securitization market (at least until mid-2007 before the start of the subprime mortgage crisis), we examine the market for residential mortgage-backed securities (RMBS) in order to understand the macro-financial implications associated with the rapid growth of securitization, and their significance for the execution of monetary policy. We conduct similar analysis in EM, although it was constrained by the scarcity of historical information and data availability (Jobst, 2006).

Previous research suggests that market rates respond more significantly to policy rates in the presence of securitization. Sellon (2000) finds that securitization affects the interest rate channel of monetary policy by causing market rates (e.g., mortgage rates) to respond more to the policy rate than without securitization. However, this effect might hold true only for mature economies with fully developed financial systems. Empirical evidence of such dynamics is still hard to discern from EM countries, given the limited history of securitization.

Other studies show that securitization dampens the impact of monetary policy on aggregate demand due to the availability of alternative funding sources. The level of securitization can have a significant impact on the degree to which a given change in monetary policy effects changes of aggregate demand and real output (Kuttner and Mosser, 2002). Since the bank balance sheet determines how monetary policy changes the demand or supply of credit (i.e., the way the policy rate is set to determine the market-clearing demand for funds), the credit channel is likely to be an important conduit of transmission.¹⁰ In particular, monetary contraction reduces money supply and reservable deposits available for banks. At the same time, a downward sloping money demand curve implies that higher interest rates dampen interest-sensitive investment and expenditure (Rosenberg, 2005). The broader balance sheet channel reduces cash flows of firms (or leads to the decline of asset values/collateral), which, in turn, increases the external finance premium and the associated cost of capital (Levin et al., 2004). Declining liabilities force banks to shrink their asset base by curtailing lending (via an adjustment of prices, loan volumes, or both). However, securitization helps banks undo this effect of tightening monetary policy by providing them access to alternative sources of funds that substitute traditional funding sources (e.g., deposit-taking)—and decreases the market-clearing rate of money demand.¹¹ Financial market deepening tends to increase the use of securitization as the availability of reference assets and investment instruments increases in response to greater capital market maturity and lower transaction costs (ADB, 2007). Amid greater pervasiveness of securitization, liability constraints are probably less binding on bank balance sheets and asset growth, resulting in greater efficiency of loan origination.

We find that securitization activity does indeed lower the interest rate elasticity of real output.¹² In both cases, the United States and South Africa, controlling for the relative share of securitized mortgages alters the traditionally negative relation between output gap and real interest rates. Muted interest rate sensitivity implies a weaker effect of higher interest rates on money demand and/or greater availability of credit at the given policy rate (while controlling for increasingly forward-looking monetary policy might preserve a negative impact of monetary policy shocks).¹³ In the United States, the negative effect of securitization on interest rate elasticity of output is unaffected by the general expansion of alternative sources of finance, which is a measure of financial sector deepening. While securitization dampens interest elasticity of real output, it also cushions the immediate impact of a positive interest rate shock to the bank’s balance sheet, as the substitutability of funding sources makes banks less likely to curtail lending in response to tightening monetary policy, particularly in EM markets like South Africa.

Securitization, however, increases the interest rate pass-through, indicating that the credit channel (and its effect on money demand) may more than offset the interest rate channel. Rising securitization in the United States has facilitated the transmission from monetary policy to market interest rates over time. This, however, is not inconsistent with a more muted interest rate elasticity of output, as the presence of securitization enhances the role of the credit channel (through the bank balance sheet). Rising securitized issuance allows banks expand credit supply by adjusting quantities rather than prices (assuming that the impact of monetary policy shocks has remained unchanged), leaving monetary transmission via the interest rate channel largely unaffected. Our analysis of securitization in the bank-dominated financial system of South Africa demonstrates this point. While securitization dampens interest elasticity of real output, it also cushions the immediate impact of a positive interest rate shock to the bank’s balance sheet and marginal loan origination. The substitutability of funding sources makes banks less likely to curtail lending in response to tightening monetary policy and a change in prevailing interest rates.

While securitization might affect the way in which monetary policy influences changes in interest rates, there is no clear evidence that securitization actually increases credit supply. Investors of securitized assets could alternatively have invested in other assets, such as bank deposits or capital enabling the banks to originate the same volume of loans in the traditional manner. In such a case, securitization would seem to involve the displacement of other forms of credit extension rather than an increase in the outright volume of credit.¹⁴

The paper is structured as follows. First, we provide evidence of the macro-financial linkages in the U.S. mortgage market by investigating the interest rate elasticity of output gap and the transmission channel of monetary policy based on different OLS-AR and VAR specifications in keeping with the existing literature. In the second part of the paper, we apply our approach to the financial sector in South Africa in order to assess whether our findings for the U.S. mortgage market apply to an advanced emerging market country with rapidly developing securitization activity. We then present our estimation results and discuss our findings. The fourth section provides an initial assessment of policy implications and concludes the paper.

II. Empirical Analysis

We analyze the secondary mortgage market in the United States in order to understand the macro-financial linkages of securitization in a well functioning financial system, which could provide meaningful lessons for the effects of securitization in EM.¹⁵ More specifically, we examine whether securitization (i) alleviates liability constraints of banks in times of monetary contraction (“credit supply effect”) and/or (ii) helps maintain money demand at a certain policy rate by limiting credit rationing (“credit demand effect”) (while allowing banks increase their balance sheet leverage).

After exploring the influence of securitization on the interest rate elasticity of GDP and the monetary transmission channel of interest rates in the United States (i.e., the interest rate pass-through from policy rates to market rates), we adapt our approach to the bank-dominated secondary mortgage market in South Africa, where securitization has morphed into the most vibrant capital market segment over a period of little more than five years. Although South Africa has not escaped the economic fallout from the U.S. sub-prime mortgage crisis, it still hosts one of the most active securitization markets in EM after having exceeded an annual issuance volume of US$4 billion in 2006 (see Box 1). Nonetheless, still only a fraction of mortgages (about three percent) are securitized in South Africa, whilst more than 60 percent of the mortgage market is refinanced via structured finance transactions of private sector banks, financing houses or government programs and agencies (see Figures 2 and 3).

View Full Size

Mortgage-Related Securitization (Outstanding and Issuance) in the United States, Emerging Markets, and South Africa

(in billions of U.S. dollars, 1999–2007)

Citation: IMF Working Papers 2009, 026; 10.5089/9781451871739.001.A001

1/ Includes GNMA, FNMA, and FHLMC mortgage-backed securities (MBS), collateralized mortgage obligations (CMOs) and private-label MBS/CMOs. Source: U.S. mortgage agencies, Thomson Financial, Bloomberg, Dealogic, Securities Industry and Financial Markets Association (SIFMA).

• Download Figure
• Download figure as PowerPoint slide

View Full Size

Stock of U.S. Mortgage-Backed Securities

(in billions of U.S. dollars, 1966–2006)

Citation: IMF Working Papers 2009, 026; 10.5089/9781451871739.001.A001

Source: U.S. mortgage agencies, Securities Industry and Financial Markets Association (SIFMA), Federal Reserve Board of Governors.

• Download Figure
• Download figure as PowerPoint slide

A. Macro-Financial Linkages in Mature Economies—Evidence from the U.S. Secondary Mortgage Market

The interaction between the housing sector financing and monetary policy has changed following deregulation, which prompted greater competitiveness in the U.S. mortgage market.¹⁶ In a bank-based financial system, credit supply and its effect on housing are largely driven by monetary policy. In general, lower money supply reduces deposit flows to banks (and other savings institutions), causing less mortgage lending due to limited funds and cutbacks in residential investment, home purchases, and housing starts. While housing credit activity is normally an early indicator of the influence of monetary policy on the economy, it has gradually lost much of its former significance as a result of the full integration of the primary mortgage market with the capital market (Kahn, 1989), creating more diversified sources of funding of mortgage loans, such as securitization. The supply of mortgage credit is no longer subject to sharp swings in availability of thrifts to secure deposits. Instead, mortgage credit is generally available at a going interest rate that is less affected by the transmission channel of monetary policy.¹⁷ This has lowered the chance that tightening monetary policy leads to non-price rationing of mortgage credit (McCarthy and Peach, 2000). Conversely, banks would adjust quantities (rather than prices) if securitization (or similar technologies) did not exist in times of contractionary monetary policy.

The growing use of mortgage securitization seems to have reduced the efficacy of U.S. monetary policy on the real economy but increased the interest rate pass-through (by creating higher liquidity and credit volume).¹⁸ Previous research shows that the reaction of mortgage rates to changes in the U.S. Federal Funds rate has strengthened, while the interest elasticity of GDP growth has become close to zero, suggesting that the efficacy of monetary policy may not be working through the interest rate channel but through the credit channel of a wider range of providers of mortgage loans (Estrella, 2001 and 2002). Interest channel-based funding of bank loans appears to have become less important as a determinant of GDP growth in light of increasing mortgage securitization (in lieu of deposits) as an alternative source of funds (Mayer, 2007), whose off-balance sheet transfer of credit risk fosters loan origination while leaving capital coverage unchanged. At the same time, more forward-looking and endogenous (rather than proactive) U.S. monetary policy might explain a muted response of output to monetary policy shocks under conventional identification methods of transmission mechanisms, which do not take into account these considerations. Other studies have found that there has been an increased response of consumer and mortgage loans to interest rate changes (Sellon, 2000). Kolari et al. (1998) find that mortgage interest rates have indeed declined with the growth of securitization in the mortgage markets. In other words, there may well be a need to undertake a larger monetary policy move to achieve its same intended objective in terms of output.

In this section, we examine the impact of securitization on both (i) the interest rate elasticity of real output and (ii) the dispersion of mortgage interest rates (i.e, interest rate pass-through) in order to gauge the efficacy of the monetary policy in the United States. Curiously, Bernanke and Gertler (1995) find that the potential effect of securitization on monetary policy transmission might be largest when the development of local capital market attributes less economic significance to the conventional credit channel via the banking system. In our study, we introduce a richer model specification that allows us to disentangle the influence of financial deepening (i.e., bond and equity market size and bank credit relative to GDP) from securitization per se, which tends to increase as capital markets develop. We also acknowledge rising financial disintermediation of mortgage loans in the U.S. financial system. However, this study largely focuses on the GSE issued MBS market and does not take into account the private label securitization market.

Monetary transmission in the presence of Securitization (OLS and VAR estimation)

We present two modeling approaches (OLS and VAR) to test the hypothesis that more pervasive securitization may limit the efficacy of monetary policy to change real economic variables as banks can substitute traditional liabilities to finance credit supply, with money demand more likely to be met at a given policy or target rate set my monetary authorities.

Our OLS-AR(2) model is based on the original dynamic IS curve by Rudebusch and Svensson (1999) as well as Johansen and Juselius (1994), which specify the marginal (contractionary) influence of the real interest rate on the output gap as

$\begin{array}{ccc}{y}_t=\alpha \left(w_t\right)+{\displaystyle \sum _{j=1}^p{\beta }_j{y}_{t-j}+{\beta }_3\left(w_t\right)\left({\overline{i}}_{t-1}-{\overline{\pi }}_{t-1}\right)+{\epsilon }_t,}& & (1)\end{array}$

where

$\begin{array}{ccc}\alpha \left(w_t\right)={\alpha }_1& & (2)\end{array}$

and

$\begin{array}{cc}{\beta }_3\left(w_t\right)={\beta }_3<0,& & (3)\end{array}$

with quarterly output gap y_t lagged by one (t-1) and two quarters (t-2), the four-quarter average of the current and lagged effective Federal Funds rate ${\overline{i}}_{t-1}$ (based on monthly weighted average), the average inflation ${\overline{\pi }}_{t-1}$ over the same four quarters, and non-autoregressive i.i.d. residuals ε_t, so that the long-term mean of the output gap is defined as

$\begin{array}{ccc}E\left(y_t\right)={\left(1-\left({\displaystyle \sum _{j=1}^p{\beta }_j}\right)\right)}^{-1}E\left(\alpha \left(w_t\right)-{\beta }_3\left(w_t\right)\left({\overline{i}}_{t-1}-{\overline{\pi }}_{t-1}\right)\right).& & (4)\end{array}$

Estrella (2002) extends this model to test the effects of mortgage securitization on the reaction of output to monetary policy moves by conditioning both the coefficient β₃ of the real interest rate and the intercept term on the ratio S_t of securitized home mortgages to the value of all home mortgages (in percent),¹⁹ so that

$\begin{array}{ccc}\alpha \left(w_t\right)={\alpha }_1+{\alpha }_2{S}_t& & (5)\end{array}$

and

$\begin{array}{ccc}{\beta }_3\left(w_t\right)={\beta }_{3,1}+{\beta }_{3,2}{S}_t{.}^{20}& & (6)\end{array}$

In extending equations (5) and (6)²⁰ above to equations (7) and (8) below, we follow Roldos (2006) and introduce the financing ratio F_t, which reflects the changing level of direct financing (i.e., equity, bonds, commercial paper or other capital-market based sources of external funding, with securitized issuance excluded) relative to indirect (or intermediated) financing (i.e., bank loans to the non-financial private sector, with household loans excluded) over the chosen sample period. By controlling for F_t, we allow interest elasticity and its interaction effect with the prevalence of securitization to vary with the degree of financial sector sophistication. In this way, we present a more refined approach to examine the influence of financial market deepening (i.e. bond and equity market size and bank credit relative to GDP) on the use of securitization, which tends to increase as capital markets develop. In addition, we include credit growth $K_t=\text{ln}\left(\frac{k_t}{GD{P}_t}/\frac{k_{t-1}}{GD{P}_{t-1}}\right)$ as the logarithmic difference between private sector credit k to GDP over two subsequent periods in order to control for the relative importance of the credit channel to growth of aggregate demand.

Inference procedures of standard parametric models, such as OLS, assume non-integration and level stationarity of endogenous variables. Thus, we first analyze the stochastic properties of each variable in our sample in order to ascertain mean reversion by ruling out stochastic behavior (with a constant forecast value and time-varying auto covariance). Based on the classical Augmented Dickey-Fuller (ADF) (Dickey and Fuller, 1979 and 1981) and Phillips-Perron (PP) (1988) unit root tests, we find that the output gap, the real interest rate (over an extended sample time period), and credit growth are all I(0) level stationary. Since the securitization ratio S_t is I(1) integrated, we introduce a time trend t in order to control for the linear effect of continuously increasing securitization on the relation between changes in output gap and monetary policy. Thus, based on equation (1), we estimate the following model (and subsets thereof):

$\begin{array}{ccc}\alpha \left(w_t\right)={\alpha }_1+{\alpha }_2{S}_{t}t+{\beta }_4{F}_t+{\beta }_5{K}_t+{\beta }_6{S}_{t}t{F}_t+{\beta }_7{S}_{t}t{F}_t{K}_t& & (7)\end{array}$

and

$\begin{array}{ccc}{\beta }_3\left(w_t\right)={\beta }_{3,1}+{\beta }_{3,2}{S}_{t}t+{\beta }_{3,3}{F}_t+{\beta }_{3,4}{K}_t+{\beta }_{3,5}{S}_{t}t{F}_t+{\beta }_{3,6}{S}_{t}t{F}_t{K}_t.& & (8)\end{array}$

with the financing ratio F_t and the logarithmic difference of relative credit growth K_t as control variables and a simple lag structure without gaps at a maximal order of p=2 in order to sufficiently accommodate delayed demand effects to interest rate changes.²¹

As an alternative model of investigating interest elasticity, we adopt an unrestricted VAR framework to study the response of output to monetary policy shocks—with and without controlling for securitization and financial sector depth as instrumental variables (Hamilton, 1994; Davidson and MacKinnon, 1993). As an useful alternative to structural modeling procedures, the linear simultaneous equation system of VAR measures the dynamic impact of random disturbances on contemporaneous changes of both output gap and interest rates as stationary endogenous variables, without imposing a cointegration restriction on their long-term intertemporal relation.

We model the five-dimensional VAR specification of vector X_t = (Δy, Δr_t, Δr_t × ΔS_t, Δr_t × ΔF_t, Δr_t × ΔS_t × ΔF)^′ of quarterly output gap Δy_t = y_t − Δy_t−1, real interest rate $\Delta r_t=\left({\overline{i}}_t-{\overline{\pi }}_t\right)-\left({\overline{i}}_{t-1}-{\overline{\pi }}_{t-1}\right)$, the securitization ratio ΔS_t = S_t − S_t−1, the financing ratio ΔF_t = F_t − F_t−1, and their interactive term with the real interest rate at first differences at time t as

$\begin{array}{ccc}{X}_t=C+{\displaystyle \mathbf{\sum }_{j=1}^p{{\Phi }}_t{X}_{t-j}+{{E}}_t},& & (9)\end{array}$

with C as a (2x1) vector of constants c₁ and c₂, Φ_t as (5x5) parameter coefficient vector matrix of jointly dependent past X_t values (compounded by lag structure p=2 number of lags), and Ε_t as (5x1) vector ${\left({\epsilon }_t^{\left(\Delta y_t\right)},{\epsilon }_t^{\left(\Delta r_t\right)},{\epsilon }_t^{\left(\Delta S_t\right)},{\epsilon }_t^{\left(\Delta F_t\right)},{\epsilon }_t^{\left(\Delta S_t\times \Delta F_t\right)}\right)}^{\prime }$ of non-autoregressive i.i.d. residuals ε_t ~ N(0,Σ) with variance-covariance matrix Σ.²² The assumption of not serially correlated residuals is not restrictive, since any residual serial correlation could be easily absorbed by a higher polynomial lag as long as normality holds. However, in this case statistical inference would be inefficient and would warrant regression weights to inform a correct estimation of statistical errors.

We focus primarily on the output gap equation and its impulse-response function in order to measure the interest rate elasticity of output and its sensitivity to interest rate shocks. The VAR specification defines the contemporaneous average quarterly change of the output gap as dependent on its previous changes and past changes of the average real interest rate. The model is estimated both statically for non-overlapping sample periods and dynamically (reverse recursive and ten-year rolling window with quarterly updates) based on a heteroskedasticity consistent coefficient covariance (White, 1980). There is a particular focus on the evolution of the coefficient of the real interest rate at one lag, which can reasonably be interpreted as the “short-term elasticity” of real activity to changes in the policy rate. Any innovations of the policy rate in the variance-covariance matrix show how monetary policy shocks impact on output.

Interest rate pass-through (OLS estimation)

We also apply a modified form of our original OLS specification above to test the hypothesis that securitization enhances the interest rate pass-through of monetary policy on the level of market rates, such as mortgage interest rates. In the case of the United States, we define the interest rate pass-through as

$\begin{array}{ccc}{m}_t=\alpha +{\displaystyle \mathbf{\sum }_{j=1}^p{\beta }_j{m}_{t-j}+{\displaystyle \mathbf{\sum }_{j=0}^p\left({\gamma }_{j+1,1}+{\gamma }_{j+1,2}{S}_t\right){\overline{i}}_{t-j}+{\displaystyle \mathbf{\sum }_{j=0}^p{\varphi }_{j+1}{\overline{\pi }}_{t-j}+{\epsilon }_t,}}}& & (10)\end{array}$

based on the structural identification of the policy rate affecting market rates (and not vice-versa), where m_t is the end-of-quarter 30-year mortgage rate, ${\overline{i}}_{t-j}$ is the four-quarter average of current and lagged effective Federal Funds rates (with and without “securitization control”), ${\overline{\pi }}_{t-j}$ is the average inflation over the same four quarters, and non-autoregressive i.i.d. residuals ε_t, in order to examine adjustments in mortgage rates in response to changes of policy rates for a simple lag structure p=2. The specification of equation (10) seems robust to potential inefficiencies caused by multi-collinearity. We use the traditional linear approach of Granger causality testing (Granger, 1969)²³ to examine the joint short-term dynamics of exogenous factors. While the mortgage rate and the output gap could Granger cause each other, changes in inflation or the federal fund rate are independent of variations in mortgage rates.

Findings

We find that securitization activity dampens the interest rate elasticity of output. Controlling for the relative share of securitized mortgages alters the traditionally negative relation between output gap and real interest rates.²⁴ Our OLS estimation results indicate that the interaction term of the securitization ratio and the real interest rate has a positive coefficient and remains statistically significant even after we control for the time trend of both the securitization ratio and the real interest rate (see Table 1). This observation vindicates the results from an earlier study by Estrella (2002) who finds that securitization tempers the sensitivity of output to monetary policy.²⁵ Also our VAR estimation results (see Table 2) are consistent with this finding, confirming that increasing securitization repeal the negative feedback effect of monetary policy on investment.

Table 1.

United States—OLS Estimation Results: Is Dynamic Equation of Output Gap with Instrumental Variable Controls for Securitization and Financial Depth (1970–2006)

Note: Sample time period: 1970 (3^rd qtr.)-2006 (4^th qtr.), 144 obs. The following cases have been estimated based on the general model specification $y_t=\alpha \left(w_t\right)+{\mathbf{\sum }}_{j=1}^p{\beta }_j{y}_{t-j}+{\beta }_3\left(w_t\right)\left({\overline{i}}_{t-1}-{\overline{\pi }}_{t-1}\right)+{\epsilon }_t$ (see equation (1)): (i) elasticity varying with S, F, and K, and joint effects of S and F as well as S, F, and K: α(w_t)= α₁ + α₂S_tt + β₄F_t + β₅K_t + β₆S_ttF_t + β₇S_ttF_tK_t and β₃(w_t) = β_3,1 + β_3,2S_tt + β_3,3F_t + β_3,4K_t + β_3,5S_ttF_t + β_3,6S_ttF_tK_t; (ii) elasticity varying with S, F, and K, and joint effects of S and F: α(w_t)=α₁ + α₂S_tt + β₄F_t + β₅K_t + β₆S_ttF_t and β₃(w_t) = β_3,1 + β_3,2S_tt + β_3,3F_t + β_3,4K_t + β_3,5S_ttF_t; (iii) elasticity varying with S and F, and joint effects of S and F: α(w_t) = α₁ + α₂S_tt + β₄F_t + β₆S_ttF_t and β₃(w_t) = β_3,1 + β_3,2S_tt + β_3,3F_t + β_3,5S_ttF_t; (iv) elasticity varying with S and F: α(w_t)=α₁ + α₂S_tt + β₄F_t and β₃(w_t) = β_3,1 + β_3,2S_tt + β_3,3F_t; (v) elasticity varying with S: α(w_t) = α₁ + α₂S_tt and β₃(w_t) = β_3,1 + β_3,2S_tt; (vi) base case: α(w_t) = α₁ and β₃(w_t) = β_3,1. Estimations are based on heteroskedasticity consistent coefficient covariance (White, 1980). The shaded areas highlight the coefficient values of the interest elasticity of the output gap. *, **, and *** indicate the 10 percent, 5 percent, 1 percent level of statistical significance (two-tailed).

Table 1.

United States—OLS Estimation Results: Is Dynamic Equation of Output Gap with Instrumental Variable Controls for Securitization and Financial Depth (1970–2006)

Model	(1)	(2)	(3)	(4)	(5)	(6)
α1 (std. error)	0.0138 0.0920	0.0249 0.1623	0.0691 0.1752	-0.5082 3.4683	-1.6473 0.3719	-2.2638 3.3242
α2 (std. error)	… …	-0.1249 0.2470	0.0008** 0.0004	0.0009 0.0008	0.0009 0.0008	-5.0976 5.4275
β1 (std. error)	1.1716*** 0.0913	1.1593*** 0.0909	1.1159*** 0.0941	1.1144*** 0.0932	1.1628*** 0.0806	1.1926*** 0.0790
β2 (std. error)	-0.2780*** 0.0896	-0.2762*** 0.0903	-0.2364*** 0.0932	-0.2463*** 0.0969	-0.2697*** 0.0865	-0.2884*** 0.08184
β 3,1 (std. error)	-0.0392 0.0285	-0.0820* 0.0460	-0.1924*** 0.0705	-0.5709 0.9975	-0.3439 1.0277	-0.2479 1.0051
β 3,2 (std. error)	… …	0.0002* 0.0001	0.0003*** 0.0001	0.0004** 0.0002	0.0003** 0.0002	0.0004** 0.0002
β3,3 (std. error)	… …	… …	… …	0.4406 1.0945	0.2303 1.1218	0.1323 1.0932
β3,4 (std. error)	… …	… …	… …	… …	1.9493 1.8344	-0.3811 2.4450
β 3,5 (std. error)	… …	… …	-0.0002*** 0.0001	-0.0003** 0.0001	-0.0003* 0.0002	-0.0003* 0.0002
β3,6 (std. error)	… …	… …	… …	… …	… …	0.0011** 0.0006
β4 (std. error)	… …	… …	… …	0.6078 3.6029	1.8033 3.4983	2.4693 3.4477
β5 (std. error)	… …	… …	… …	… …	-12.1610* 6.8252	-16.8156* 9.1726
β6 (std. error)	… …	… …	0.0003** 0.0002	0.0003 0.0004	0.0003 0.0003	0.0002 0.0004
β7 (std. error)	… …	… …	… …	… …	… …	0.0007 0.0012
Adj. R2	0.8571	0.8573	0.8615	0.8602	0.8619	0.8629

Note: Sample time period: 1970 (3^rd qtr.)-2006 (4^th qtr.), 144 obs. The following cases have been estimated based on the general model specification $y_t=\alpha \left(w_t\right)+{\mathbf{\sum }}_{j=1}^p{\beta }_j{y}_{t-j}+{\beta }_3\left(w_t\right)\left({\overline{i}}_{t-1}-{\overline{\pi }}_{t-1}\right)+{\epsilon }_t$ (see equation (1)): (i) elasticity varying with S, F, and K, and joint effects of S and F as well as S, F, and K: α(w_t)= α₁ + α₂S_tt + β₄F_t + β₅K_t + β₆S_ttF_t + β₇S_ttF_tK_t and β₃(w_t) = β_3,1 + β_3,2S_tt + β_3,3F_t + β_3,4K_t + β_3,5S_ttF_t + β_3,6S_ttF_tK_t; (ii) elasticity varying with S, F, and K, and joint effects of S and F: α(w_t)=α₁ + α₂S_tt + β₄F_t + β₅K_t + β₆S_ttF_t and β₃(w_t) = β_3,1 + β_3,2S_tt + β_3,3F_t + β_3,4K_t + β_3,5S_ttF_t; (iii) elasticity varying with S and F, and joint effects of S and F: α(w_t) = α₁ + α₂S_tt + β₄F_t + β₆S_ttF_t and β₃(w_t) = β_3,1 + β_3,2S_tt + β_3,3F_t + β_3,5S_ttF_t; (iv) elasticity varying with S and F: α(w_t)=α₁ + α₂S_tt + β₄F_t and β₃(w_t) = β_3,1 + β_3,2S_tt + β_3,3F_t; (v) elasticity varying with S: α(w_t) = α₁ + α₂S_tt and β₃(w_t) = β_3,1 + β_3,2S_tt; (vi) base case: α(w_t) = α₁ and β₃(w_t) = β_3,1. Estimations are based on heteroskedasticity consistent coefficient covariance (White, 1980). The shaded areas highlight the coefficient values of the interest elasticity of the output gap. *, **, and *** indicate the 10 percent, 5 percent, 1 percent level of statistical significance (two-tailed).

View Table

Table 1.

United States—OLS Estimation Results: Is Dynamic Equation of Output Gap with Instrumental Variable Controls for Securitization and Financial Depth (1970–2006)

Model	(1)	(2)	(3)	(4)	(5)	(6)
α1 (std. error)	0.0138 0.0920	0.0249 0.1623	0.0691 0.1752	-0.5082 3.4683	-1.6473 0.3719	-2.2638 3.3242
α2 (std. error)	… …	-0.1249 0.2470	0.0008** 0.0004	0.0009 0.0008	0.0009 0.0008	-5.0976 5.4275
β1 (std. error)	1.1716*** 0.0913	1.1593*** 0.0909	1.1159*** 0.0941	1.1144*** 0.0932	1.1628*** 0.0806	1.1926*** 0.0790
β2 (std. error)	-0.2780*** 0.0896	-0.2762*** 0.0903	-0.2364*** 0.0932	-0.2463*** 0.0969	-0.2697*** 0.0865	-0.2884*** 0.08184
β 3,1 (std. error)	-0.0392 0.0285	-0.0820* 0.0460	-0.1924*** 0.0705	-0.5709 0.9975	-0.3439 1.0277	-0.2479 1.0051
β 3,2 (std. error)	… …	0.0002* 0.0001	0.0003*** 0.0001	0.0004** 0.0002	0.0003** 0.0002	0.0004** 0.0002
β3,3 (std. error)	… …	… …	… …	0.4406 1.0945	0.2303 1.1218	0.1323 1.0932
β3,4 (std. error)	… …	… …	… …	… …	1.9493 1.8344	-0.3811 2.4450
β 3,5 (std. error)	… …	… …	-0.0002*** 0.0001	-0.0003** 0.0001	-0.0003* 0.0002	-0.0003* 0.0002
β3,6 (std. error)	… …	… …	… …	… …	… …	0.0011** 0.0006
β4 (std. error)	… …	… …	… …	0.6078 3.6029	1.8033 3.4983	2.4693 3.4477
β5 (std. error)	… …	… …	… …	… …	-12.1610* 6.8252	-16.8156* 9.1726
β6 (std. error)	… …	… …	0.0003** 0.0002	0.0003 0.0004	0.0003 0.0003	0.0002 0.0004
β7 (std. error)	… …	… …	… …	… …	… …	0.0007 0.0012
Adj. R2	0.8571	0.8573	0.8615	0.8602	0.8619	0.8629

Note: Sample time period: 1970 (3^rd qtr.)-2006 (4^th qtr.), 144 obs. The following cases have been estimated based on the general model specification $y_t=\alpha \left(w_t\right)+{\mathbf{\sum }}_{j=1}^p{\beta }_j{y}_{t-j}+{\beta }_3\left(w_t\right)\left({\overline{i}}_{t-1}-{\overline{\pi }}_{t-1}\right)+{\epsilon }_t$ (see equation (1)): (i) elasticity varying with S, F, and K, and joint effects of S and F as well as S, F, and K: α(w_t)= α₁ + α₂S_tt + β₄F_t + β₅K_t + β₆S_ttF_t + β₇S_ttF_tK_t and β₃(w_t) = β_3,1 + β_3,2S_tt + β_3,3F_t + β_3,4K_t + β_3,5S_ttF_t + β_3,6S_ttF_tK_t; (ii) elasticity varying with S, F, and K, and joint effects of S and F: α(w_t)=α₁ + α₂S_tt + β₄F_t + β₅K_t + β₆S_ttF_t and β₃(w_t) = β_3,1 + β_3,2S_tt + β_3,3F_t + β_3,4K_t + β_3,5S_ttF_t; (iii) elasticity varying with S and F, and joint effects of S and F: α(w_t) = α₁ + α₂S_tt + β₄F_t + β₆S_ttF_t and β₃(w_t) = β_3,1 + β_3,2S_tt + β_3,3F_t + β_3,5S_ttF_t; (iv) elasticity varying with S and F: α(w_t)=α₁ + α₂S_tt + β₄F_t and β₃(w_t) = β_3,1 + β_3,2S_tt + β_3,3F_t; (v) elasticity varying with S: α(w_t) = α₁ + α₂S_tt and β₃(w_t) = β_3,1 + β_3,2S_tt; (vi) base case: α(w_t) = α₁ and β₃(w_t) = β_3,1. Estimations are based on heteroskedasticity consistent coefficient covariance (White, 1980). The shaded areas highlight the coefficient values of the interest elasticity of the output gap. *, **, and *** indicate the 10 percent, 5 percent, 1 percent level of statistical significance (two-tailed).

Table 2.

United States—Estimation Results: VAR(2,2) Simultaneous Equation Model with Instrumental Variable Controls for Securitization and Financial Depth (1970–2006)

Note: Sample time period: 1970 (3rd qtr.)–2006 (4th qtr.), 144 obs. Estimations are based on heteroskedasticity consistent coefficient covariance (White, 1980). The shaded areas highlight the coefficient values of the first- and second order interest elasticity of the output gap (with and without controlling for degree of securitization activity and/or financial sector depth). *, **, and *** indicate the 10 percent, 5 percent, 1 percent level of statistical significance (two-tailed). Note that our specification of endogenous variables at first differences affects a positive interest rate elasticity of output gap. The progressive shading of the output gap equation indicates the critical values for each combination of the securitization and financing ratios.

Table 2.

United States—Estimation Results: VAR(2,2) Simultaneous Equation Model with Instrumental Variable Controls for Securitization and Financial Depth (1970–2006)

X1-5	Δyt	Δrt	Δrt x ΔSt	Δrt x ΔFt	Δrt x ΔSt x ΔFt
C (std. error)	0.0056 0.0740	-0.0069 0.0292	-0.0274* 0.0172	-0.0015*** 0.0005	-0.0006*** 0.0002
Φ1,1 (std. error)	0.1473* 0.0844	0.0437 0.0333	0.0174 0.0197	0.0012** 0.0010	0.0004 0.0003
Φ1,2 (std. error)	0.1819** 0.0858	0.0927*** 0.0338	0.0513*** 0.0200	0.0004 0.0010	0.0002 0.0003
Φ2,1 (std. error)	0.2612 0.2449	0.8934*** 0.0966	0.1533** 0.0571	-0.0003 0.00017	0.0005 0.0008
Φ2.2 (std. error)	-0.7391*** 0.2416	-0.2755*** 0.0953	-0.0915* 0.0563	0.0006 0.0016	-0.0007 0.0008
Φ3,1 (std. error)	0.4810 0.4258	0.0942 0.1679	0.1561 0.0992	0.0006 0.0029	0.0042*** 0.0014
Φ3,2 (std. error)	0.3272 0.4320	4.0543 6.9468	-3.2491 4.1052	0.2184* 0.1196	-0.0524 0.0570
Φ4,1 (std. error)	-10.2420 17.6152	4.0543 6.9468	-3.2491 4.1052	0.2184 0.1196	-0.0524 0.0570
Φ4,2 (std. error)	24.4643 17.2767	-8.8211 6.8133	2.6075 4.0263	0.0869 0.1173	0.0447 0.0559
Φ5,1 (std. error)	-14.1511 37.8327	23.2849 14.9198	28.8850*** 8.8168	-0.0670 0.2569	0.3271*** 0.1225
Φ5,2 (std. error)	-56.3022 38.7940	-10.6537 15.2990	-28.4559*** 9.0409	-0.1058 0.2634	-0.2766** 0.12558
Adj. R2	0.1194	0.5995	0.3798	0.0568	0.1553

Note: Sample time period: 1970 (3rd qtr.)–2006 (4th qtr.), 144 obs. Estimations are based on heteroskedasticity consistent coefficient covariance (White, 1980). The shaded areas highlight the coefficient values of the first- and second order interest elasticity of the output gap (with and without controlling for degree of securitization activity and/or financial sector depth). *, **, and *** indicate the 10 percent, 5 percent, 1 percent level of statistical significance (two-tailed). Note that our specification of endogenous variables at first differences affects a positive interest rate elasticity of output gap. The progressive shading of the output gap equation indicates the critical values for each combination of the securitization and financing ratios.

View Table

Table 2.

United States—Estimation Results: VAR(2,2) Simultaneous Equation Model with Instrumental Variable Controls for Securitization and Financial Depth (1970–2006)

X1-5	Δyt	Δrt	Δrt x ΔSt	Δrt x ΔFt	Δrt x ΔSt x ΔFt
C (std. error)	0.0056 0.0740	-0.0069 0.0292	-0.0274* 0.0172	-0.0015*** 0.0005	-0.0006*** 0.0002
Φ1,1 (std. error)	0.1473* 0.0844	0.0437 0.0333	0.0174 0.0197	0.0012** 0.0010	0.0004 0.0003
Φ1,2 (std. error)	0.1819** 0.0858	0.0927*** 0.0338	0.0513*** 0.0200	0.0004 0.0010	0.0002 0.0003
Φ2,1 (std. error)	0.2612 0.2449	0.8934*** 0.0966	0.1533** 0.0571	-0.0003 0.00017	0.0005 0.0008
Φ2.2 (std. error)	-0.7391*** 0.2416	-0.2755*** 0.0953	-0.0915* 0.0563	0.0006 0.0016	-0.0007 0.0008
Φ3,1 (std. error)	0.4810 0.4258	0.0942 0.1679	0.1561 0.0992	0.0006 0.0029	0.0042*** 0.0014
Φ3,2 (std. error)	0.3272 0.4320	4.0543 6.9468	-3.2491 4.1052	0.2184* 0.1196	-0.0524 0.0570
Φ4,1 (std. error)	-10.2420 17.6152	4.0543 6.9468	-3.2491 4.1052	0.2184 0.1196	-0.0524 0.0570
Φ4,2 (std. error)	24.4643 17.2767	-8.8211 6.8133	2.6075 4.0263	0.0869 0.1173	0.0447 0.0559
Φ5,1 (std. error)	-14.1511 37.8327	23.2849 14.9198	28.8850*** 8.8168	-0.0670 0.2569	0.3271*** 0.1225
Φ5,2 (std. error)	-56.3022 38.7940	-10.6537 15.2990	-28.4559*** 9.0409	-0.1058 0.2634	-0.2766** 0.12558
Adj. R2	0.1194	0.5995	0.3798	0.0568	0.1553

Note: Sample time period: 1970 (3rd qtr.)–2006 (4th qtr.), 144 obs. Estimations are based on heteroskedasticity consistent coefficient covariance (White, 1980). The shaded areas highlight the coefficient values of the first- and second order interest elasticity of the output gap (with and without controlling for degree of securitization activity and/or financial sector depth). *, **, and *** indicate the 10 percent, 5 percent, 1 percent level of statistical significance (two-tailed). Note that our specification of endogenous variables at first differences affects a positive interest rate elasticity of output gap. The progressive shading of the output gap equation indicates the critical values for each combination of the securitization and financing ratios.

Financial deepening has spurred increased securitization activity but has not significantly affected the direct capacity of securitization to dampen interest rate elasticity. Although securitization activity seems to increase with greater availability of alternative sources of finance (indicated by a higher financing ratio as a measure of financial sector development) amid persistent credit growth, greater financial deepening has not amplified the impact of securitization on the efficacy of monetary policy. In fact, greater market-based financing has slightly reduced the marginal effect of securitization on interest rate elasticity of output. The economic and statistical significance of the interaction term between the securitization ratio and the real interest rate decreases after we control for the financing ratio (see Tables 1–2).²⁶ Since the financing ratio does not include securitized issuance in its numerator, we implicitly control for the positive relation between financial deepening and the incidence of securitization in order to disentangle the contemporaneous effect of both factors on interest rate elasticity. Thus, lower interest rate elasticity of output seems to largely be attributable to securitization, which offers an important and unique channel of diversified funding that helps maintain steady credit supply. Credit growth, in turn, is not a cause of muted interest elasticity but appears to result from broader and deeper capital markets that foster an environment of increasing disintermediation amenable to securitization.

The diversified supply of external finance in the United States has fostered the surge of securitization without directly affecting the efficacy of monetary policy. Financial sector deepening in the United States demonstrates how a large variety of alternative funding channels has raised disintermediation to a point where securitization at the margin influences the efficacy of monetary policy materially. Pervasive securitized issuance by non-deposit taking financial institutions has long helped disintermediate large parts of the U.S. banking sector. While other forms of external finance are highly developed, they do not seem to significantly weaken the response of output to changes in monetary policy.

The dynamic estimation of our VAR model and the associated impulse-response analysis show that the influence of securitization on the efficacy of monetary policy has increased and seems to have an outsized impact relative to declining interest rate volatility over time. If we split the sample into two non-overlapping periods, we can distinguish between the long-term effect of securitization on interest rate elasticity at times when the financing ratio is low (from 1970 to 1990) and high (from 1991 to 2006) (see Figures 4–6). While low levels of securitization have hardly affected interest rate elasticity before 1991, rising securitization has gradually dampened the general sensitivity of output to monetary policy ever since. Although interest rate volatility had decreased by almost 25 percent (“great moderation”) between the two time periods, suggesting a structural decline in interest rate elasticity without any change in the sensitivity of real activity to changes in policy rates, we still find a significant impact of securitization. The increase of securitization activity has reduced the cumulative effect of interest rate shocks on output gap by more than 90 percent. Interest rate elasticity, on average, has remained unaffected by the greater availability of alternative forms of external finance other than securitization in the wake of increasing disintermediation and financial deepening.

View Full Size

United States—Impulse-Response Graphs of Interest Rate Elasticity: VAR(5,2) Simultaneous Equation Model with Instrumental Variable Controls for Securitization and Financial Depth (1970–2006)

Citation: IMF Working Papers 2009, 026; 10.5089/9781451871739.001.A001

• Download Figure
• Download figure as PowerPoint slide

View Full Size

United States—Impulse-Response Graphs of Interest Rate Elasticity: VAR(5,2) Simultaneous Equation Model with Instrumental Variable Controls for Securitization and Financial Depth (1970–1990)

Citation: IMF Working Papers 2009, 026; 10.5089/9781451871739.001.A001

• Download Figure
• Download figure as PowerPoint slide

View Full Size

United States—Impulse-Response Graphs of Interest Rate Elasticity: VAR(5,2) Simultaneous Equation Model with Instrumental Variable Controls for Securitization and Financial Depth (1991–2006)

Citation: IMF Working Papers 2009, 026; 10.5089/9781451871739.001.A001

• Download Figure
• Download figure as PowerPoint slide

However, the declining interest rate elasticity of real output should not be seen as evidence of impeded monetary transmission. In fact, the transmission from monetary policy to market interest rates is more effective in the presence of securitization despite. After having reached a critical mass, securitization facilitates the transmission of policy rates. Our OLS estimation results of the interest rate pass-through suggest that a higher share of securitized mortgages (which results in a higher securitization ratio) initially dulls the sensitivity of mortgage interest rates to changes of the policy rate (see Tables 3–4). Since the 1990s, however, among other things, a sustained increase of securitized issuance seems to have reversed this relation and has enhanced the sensitivity of mortgage rates to changes in the Federal Funds rate under a more transparent monetary policy regime. The more efficient interest rate transmission is not inconsistent with the declining efficacy of monetary policy (i.e., lower interest rate elasticity of output), as the presence of securitization (and the availability of alternative means of banks to fund their lending) increases the liquidity of mortgage markets while enhancing the role of the credit channel by helping banks satisfy money demand at a certain policy rate without credit rationing during times of monetary contraction. That said, the relation between policy rates and mortgage rates can be more complex due to the existence of other factors omitted in this analysis, such as long-term inflation expectations and risk premia (e.g., repayment risk) embedded in mortgage rates.

Table 3.

United States—OLS Estimation Results of the Mortgage Interest Rate Pass-through, with Instrumental Variable Controls for Securitization (1970–2006)

Note: Sample time period: 1970 (3rd qtr.)–2006 (4th qtr.), 144 obs. Estimations are based on heteroskedasticity consistent coefficient covariance (White, 1980). The shaded areas highlight the coefficient values of the first-order interest rate pass-through. *, **, and *** indicate the 10 percent, 5 percent, 1 percent level of statistical significance (two-tailed). Model 1 does not include the interaction terms of the Federal Funds rate and the securitization ratio.

Table 3.

United States—OLS Estimation Results of the Mortgage Interest Rate Pass-through, with Instrumental Variable Controls for Securitization (1970–2006)

Model	(1)	(2)
α (std. error)	0.4349** 0.1724	0.2555 0.1817
β1 (std. error)	0.9899*** 0.0879	0.9644*** 0.0924
β2 (std. error)	-0.1410* 0.0845	-0.1077 0.0885
γ1,1 (std. error)	0.2291*** 0.0441	0.1324** 0.0563
γ1,2 (std. error)	- -	0.0078*** 0.0028
γ2,1 (std. error)	-0.0935* 0.0557	0.1503* 0.0865
γ2,2 (std. error)	- -	-0.0186*** 0.0045
γ3,1 (std. error)	-0.0171 0.0471	-0.1836*** 0.0598
γ3,2 (std. error)	- -	0.0114*** 0.0022
φ1 (std. error)	0.2670 0.1688	0.3323* 0.1771
φ2 (std. error)	-0.4631 0.3259	-0.5119 0.3286
φ3 (std. error)	0.2276 0.1724	0.2349 0.1721
Adj. R2	0.9821	0.9841

Note: Sample time period: 1970 (3rd qtr.)–2006 (4th qtr.), 144 obs. Estimations are based on heteroskedasticity consistent coefficient covariance (White, 1980). The shaded areas highlight the coefficient values of the first-order interest rate pass-through. *, **, and *** indicate the 10 percent, 5 percent, 1 percent level of statistical significance (two-tailed). Model 1 does not include the interaction terms of the Federal Funds rate and the securitization ratio.

View Table

Table 3.

United States—OLS Estimation Results of the Mortgage Interest Rate Pass-through, with Instrumental Variable Controls for Securitization (1970–2006)

Model	(1)	(2)
α (std. error)	0.4349** 0.1724	0.2555 0.1817
β1 (std. error)	0.9899*** 0.0879	0.9644*** 0.0924
β2 (std. error)	-0.1410* 0.0845	-0.1077 0.0885
γ1,1 (std. error)	0.2291*** 0.0441	0.1324** 0.0563
γ1,2 (std. error)	- -	0.0078*** 0.0028
γ2,1 (std. error)	-0.0935* 0.0557	0.1503* 0.0865
γ2,2 (std. error)	- -	-0.0186*** 0.0045
γ3,1 (std. error)	-0.0171 0.0471	-0.1836*** 0.0598
γ3,2 (std. error)	- -	0.0114*** 0.0022
φ1 (std. error)	0.2670 0.1688	0.3323* 0.1771
φ2 (std. error)	-0.4631 0.3259	-0.5119 0.3286
φ3 (std. error)	0.2276 0.1724	0.2349 0.1721
Adj. R2	0.9821	0.9841

Note: Sample time period: 1970 (3rd qtr.)–2006 (4th qtr.), 144 obs. Estimations are based on heteroskedasticity consistent coefficient covariance (White, 1980). The shaded areas highlight the coefficient values of the first-order interest rate pass-through. *, **, and *** indicate the 10 percent, 5 percent, 1 percent level of statistical significance (two-tailed). Model 1 does not include the interaction terms of the Federal Funds rate and the securitization ratio.

Table 4.

United States—Summary Table of OLS Estimation Results for the Interest Rate Pass-Through (Different Time Periods)

Note: Sample time period: 1970 (3rd qtr.)–2006 (4th qtr.), 144 obs. Estimations are based on heteroskedasticity consistent coefficient covariance (White, 1980). The shaded areas highlight the coefficient values of the first-order interest rate pass-through. *, **, and *** indicate the 10 percent, 5 percent, 1 percent level of statistical significance (two-tailed).

Table 4.

United States—Summary Table of OLS Estimation Results for the Interest Rate Pass-Through (Different Time Periods)

Sample Period			Full Sample (see Tab. 3)	Before 1980Q4	1980Q4- 1993Q4	1994Q1- 2006Q4
Average securitization ratio $\overline{S}$ (In percent)			28.86	16.58	31.67	56.26
Sensitivity of mortgage rate m to Federal Funds rate $\overline{i_t}$
	without controlling for St
		γ1,1 (std. error)	0.2291*** 0.0441	0.2451*** 0.0460	0.2038** 0.0782	0.4556** 0.2088
	controlling for St
		γ1,1 (std. error)	0.1324** 0.0563	0.1381** 0.0568	0.0680 0.1285	10.511*** 3.1080
		γ1,2 (std. error)	0.0078*** 0.0028	0.0090** 0.0035	0.0010* 0.0059	0.1744*** 0.0527

Note: Sample time period: 1970 (3rd qtr.)–2006 (4th qtr.), 144 obs. Estimations are based on heteroskedasticity consistent coefficient covariance (White, 1980). The shaded areas highlight the coefficient values of the first-order interest rate pass-through. *, **, and *** indicate the 10 percent, 5 percent, 1 percent level of statistical significance (two-tailed).

View Table

Table 4.

United States—Summary Table of OLS Estimation Results for the Interest Rate Pass-Through (Different Time Periods)

Sample Period			Full Sample (see Tab. 3)	Before 1980Q4	1980Q4- 1993Q4	1994Q1- 2006Q4
Average securitization ratio $\overline{S}$ (In percent)			28.86	16.58	31.67	56.26
Sensitivity of mortgage rate m to Federal Funds rate $\overline{i_t}$
	without controlling for St
		γ1,1 (std. error)	0.2291*** 0.0441	0.2451*** 0.0460	0.2038** 0.0782	0.4556** 0.2088
	controlling for St
		γ1,1 (std. error)	0.1324** 0.0563	0.1381** 0.0568	0.0680 0.1285	10.511*** 3.1080
		γ1,2 (std. error)	0.0078*** 0.0028	0.0090** 0.0035	0.0010* 0.0059	0.1744*** 0.0527

Note: Sample time period: 1970 (3rd qtr.)–2006 (4th qtr.), 144 obs. Estimations are based on heteroskedasticity consistent coefficient covariance (White, 1980). The shaded areas highlight the coefficient values of the first-order interest rate pass-through. *, **, and *** indicate the 10 percent, 5 percent, 1 percent level of statistical significance (two-tailed).