The International Trade Journal ISSN: 0885-3908 (Print) 1521-0545 (Online) Journal homepage: http://www.tandfonline.com/loi/uitj20 The Mexican Interest Rate Pass-Through in the Post-U.S. Subprime Mortgage Crisis Era Chu V. Nguyen To cite this article: Chu V. Nguyen (2017): The Mexican Interest Rate Pass-Through in the Post-U.S. Subprime Mortgage Crisis Era, The International Trade Journal, DOI: 10.1080/08853908.2017.1360226 To link to this article: http://dx.doi.org/10.1080/08853908.2017.1360226 Published online: 16 Aug 2017. Submit your article to this journal Article views: 13 View related articles View Crossmark data Full Terms & Conditions of access and use can be found at http://www.tandfonline.com/action/journalInformation?journalCode=uitj20 Download by: [University of Florida] Date: 25 October 2017, At: 02:02 THE INTERNATIONAL TRADE JOURNAL https://doi.org/10.1080/08853908.2017.1360226 The Mexican Interest Rate Pass-Through in the Post-U.S. Subprime Mortgage Crisis Era Chu V. Nguyen Marilyn Davies College of Business, University of Houston-Downtown, Houston, Texas, USA Downloaded by [University of Florida] at 02:02 25 October 2017 ABSTRACT This study investigates the nature of the Mexican interest rate pass-through during the post-U.S. subprime mortgage crisis. The empirical results reveal a very high short-run and an almost complete long-run interest rate pass-through. The bounds test indicates a long-term relationship between countercyclical monetary policy and market rates. Notwithstanding the rigid inflation targeting set by the Mexican Central Bank in the very concentrated Mexican market and its openness to foreign competition, the Mexican open economy is very small compared to the U.S. economy. Despite these conditions, the Mexican Central Bank has been very effective in conducting its countercyclical monetary policy. KEYWORDS Central Bank policy-related rate; commercial banks; interest rate pass-through; lending rate; monetary policy I. Introduction Financial intermediation is a critical facilitator of investment and economic growth (McKinnon 1973; Patrick 1966; Schumpeter 1912). Commercial banks are an integral part of the monetary policy transmission mechanism, given their ability to change the lending rates in the economy through the interest rate pass-through. These financial institutions therefore play a critical role in transmitting the countercyclical monetary policy measures to consumption and investment activities in the economy. Changes in these two macroeconomic variables will change the macroeconomic policy target variables: unemployment, inflation, and GDP. Moreover, Illes and Lombardi (2013) articulated that the transmission of policy rates to lending (and deposit) rates—the interest rate passthrough—is far from mechanical and is affected by various factors. For instance, financial intermediaries may require higher compensation for risk due to slowing economic activity. In this case, a reduction in the policy rate would only be partially passed on to firms or households. CONTACT Chu V. Nguyen email@example.com Marilyn Davies College of Business, University of HoustonDowntown, 320 North Main St., Suite 410-D, Houston, TX 77002, USA. Color versions of one or more of the figures in this article can be found online at www.tandfonline.com/uitj © 2017 Taylor & Francis Downloaded by [University of Florida] at 02:02 25 October 2017 2 C. V. NGUYEN Conversely, low perceived risk can magnify the pass-through and lead to an overheating of the economy. Illes and Lombardi (2013) further argued that, over the last few decades, setting policy rates has been viewed as the standard tool of monetary policy. The implementation of the monetary policy stance via open market operations ensures that policy rates will influence the interest rates financial institutions use to refinance themselves. In turn, competition in the lending and funding markets should ensure that changes in the policy stance are also passed on to other interest rates. A reduction in the policy rate is thus expected to translate into a decline in lending rates for firms and households which should stimulate consumption and investment. This is the interest rate channel of monetary policy transmission. Theoretically, banks operating in a free market economy could be expected to consider all sources of risk in determining and setting the spread that separates the rate paid to lenders and the rate charged to borrowers. If banks set an intermediation premium too high or too low, market forces would normally force an adjustment back to the equilibrium spread. Three main hypotheses explain this rate-setting behavior: the bank concentration hypothesis, the consumer characteristic hypothesis, and the consumer reaction hypothesis. The bank concentration hypothesis theorizes that oligopolistic banks are slow to raise deposit rates but are quick to raise lending rates when market forces allow it. Conversely, banks in declining markets quickly adjust downward the rates paid to depositors and slowly reduce the rates charged to borrowers (Hannan and Berger 1991; Neumark and Sharpe 1992). The consumer characteristic hypothesis posits that banks can adjust rates to widen the spread and increase their profitability to the extent that consumers are unsophisticated and/or are saddled with higher costs of searching and switching (Calem and Mester 1995; Hutchison 1995; Rosen 2002). On the other hand, the consumer reaction hypothesis argues that asymmetric adjustments in lending rates may actually benefit consumers because the presence of asymmetric information can foster an adverse selection problem in lending markets such that higher interest rates will tend to attract riskier borrowers (Stiglitz and Weiss 1981). Therefore, even if the market rates rise, banks would be reluctant to raise lending rates because the expected cost to the banks of not raising the lending rates (when their marginal cost of funds increases) is offset by the risk-reduction benefits of not encouraging the higher-risk borrowers. The International Trade Administration (2016) reported that Mexico’s commercial banks now offer a full spectrum of services ranging from deposit accounts, consumer and commercial lending, corporate finance, trusts, and mutual funds to foreign exchange and money market trading. Additionally, 46 banks are currently operating in Mexico; seven of these (Bancomer, Banamex, Santander, Banorte, HSBC, Inbursa, and Scotia Bank) have 78% Downloaded by [University of Florida] at 02:02 25 October 2017 THE INTERNATIONAL TRADE JOURNAL 3 of the market share in terms of total assets and three banks are linked with retail stores. The report further articulated that Mexico’s commercial banking sector has been opened to foreign competition. The North American Free Trade Agreement (NAFTA) permits U.S. and Canadian banks or any other foreign bank with a subsidiary in the United States or Canada to establish wholly owned subsidiaries in Mexico. Furthermore, they are allowed to undertake financial intermediation and to solicit customers for their parent bank. Most importantly, almost all of the major banks, except for Banorte, are under the control of foreign banks. Similarly, Manning (2015) reported that a significant proportion of the banking industry’s biggest players in Mexico is now under foreign ownership; Banamex is now a unit of Citigroup Inc., Bancomer is under the ownership of Spain’s BBVA, SERFIN is part of Santander, Inverlat is now owned by Canada’s Scotiabank, and Bital operates under HSBC’s domain. Of the 45 banks currently operating in the private sector, the two biggest institutions— Banamex and Bancomer—hold 38% of the industry’s total assets, while the top five hold a sizeable 72%. Comparatively, 1,745 banks with more than $300 million in assets operate in the U.S., with the four biggest holding only 42% of the total assets. Manning (2015) further argued that this comparative lack of competition in Mexico’s banking sector induced the country’s Antitrust Commission to conduct a comprehensive review of the sector during the first half of 2014. Historically, Weisbrot, Lefebvre, and Sammut (2014) argued that the Mexican Central Bank’s form of rigid inflation targeting also adds a large element of unpredictability to the exchange rate. Additionally, NAFTA also increasingly tied Mexico to the U.S. economy at a time when the U.S. economy was becoming dependent on growth driven by asset bubbles. As a result, Mexico suffered a recession when the stock market bubble burst in 2000 to 2002 and was one of the hardest hit countries in the region during the U.S. Great Recession, experiencing a 6.7% decline in GDP. Weisbrot, Lefebvre, and Sammut (2014) posited that the vulnerability to developments in U.S. financial markets continues: in May of 2013, after the U.S. Federal Reserve announced a future “tapering” of its quantitative easing program (QE1, QE2, and QE3), there were fears of a repeat of the 1994 peso crisis and gross foreign portfolio inflows came to a sudden stop in the Mexican economy. Given the aforementioned, it is very interesting to learn the nature of the Mexican interest rate pass-through. More specifically, the objective of this study is to investigate how Mexican commercial banks passed changes in their cost of funds due to countercyclical monetary policy, as reflected in changes in the Central Bank policy rate, to their customers through the short-run and long-run interest rate pass-through processes 4 C. V. NGUYEN in the post-U.S.subprime mortgage crisis. The remainder of the article is structured as follows: Section 2 briefly reviews the literature. Section 3 describes the econometric methodology, and specifies an empirical model for the investigation and the estimation method to calculate the passthrough. Section 4 describes the data and some descriptive statistics. Section 5 presents estimation results. Section 6 briefly discusses the empirical findings. Finally, Section 7 summarizes and concludes the article. Downloaded by [University of Florida] at 02:02 25 October 2017 II. Review of literature Many approaches and methodologies have been used to study interest-ratesetting behaviors of lending institutions around the globe. Espinosa-Vega and Rebucci (2003) applied a standard Error Correction Model to consider whether interest rate pass-through in Chile’s experience was atypical compared to 10 other countries, including the United States. These authors found that the adjustment in the Chilean banking sector was incomplete—as in other countries—but generally faster than those in the rest of their sample. Also, Espinosa-Vega and Rebucci (2003) reported that the adjustment process was affected by institutional changes in the exchange rate regime and Chile’s monetary policy targeting. Hofmann and Mizen (2004) used 17 years of monthly data for rates on 13 deposit and mortgage products offered by UK financial institutions to empirically investigate the potential nonlinearity in adjustment of retail rates to base rates, due to menu costs. They reported that the speed of adjustment responded nonlinearly to the expected size of the gap between the base rate and retail rate in the near future. In other words, the perceived (expected) “aggressiveness” in base rate management was a significant factor in explaining the speed of pass-through effects. Sørensen and Werner (2006) performed Euro-area cross-country comparisons and reported empirical evidence of high-degree heterogeneity in passthrough of base rates to bank interest rates. Among other cyclical and structural factors, Sørensen and Werner (2006) found different degrees of competition in the national banking sector to be the most significant determinants of pass-through speed. Tonooka and Koyama (2003) searched for but found no relationship between interest rates on loans and market concentration in the Brazilian banking sector. Alencar (2003) estimated the speed of pass-through effects from changes in benchmark interest rates and compared them to those observed in retail banking. The revelation that the time lag for monthly average retail rates to fully adjust to changes in the opportunity cost of money is less than 12 months was pointed out as evidence of a significant degree of competition, driving banks to operate efficiently. Downloaded by [University of Florida] at 02:02 25 October 2017 THE INTERNATIONAL TRADE JOURNAL 5 Bernanke and Blinder (1992) investigated the response of credit aggregates to monetary policy shocks. Borio and Fritz (1995) and Cottarelli and Kourelis (1994) focused more specifically on the pass-through of policy rates to lending rates, which is also the focus of this investigation. Studies on the heterogeneity in the pass-through at the individual bank level are limited to a few country studies (Gambacorta 2008; Weth 2002). The bulk of the empirical literature has resorted to cointegrated time series models developed by Engle and Granger (1987) to account for co-movements of policy and lending rates. The European Central Bank (2003) focuses on major euro-area countries, reporting evidence of heterogeneity between core and peripheral countries. Additionally, Coelho, De Mello, and Garcia (2010) found that the pass-through is higher for larger banks using a sample from June 2000 to December 2006. Moreover, the monetary policy regime can affect adjustments and volatility of retail rates. For example, one would expect nominal prices to adjust faster or the pass-through to be larger when inflation is higher (Mojon 2000). The important factors of the country’s financial structure are bank competition, development of financial markets, and banking system ownership. If financial markets are well-developed, financially solid businesses tend to opt for alternative sources of financing when retail rates rise, increasing the overall risk to bank loan portfolios. In that case, banks increase lending rates to compensate for the higher risk instead of rationing credit (Sander and Kleimeier 2004). III. Methodology and model specification Structural break and its implication It is expected that long time series data will experience structural breaks. Failure to account for structural breaks may result in model misspecification. In monetary economics, when the structural breaks occur in an economy, usually as result of economic shocks, the Central Bank often reacts by implementing countercyclical monetary policy measures. Therefore, in reality, there is an interaction between the structural break and the policy rate. To account for possible structural breaks, this analysis defined the spread, denoted by SPt , as the difference between the two time series. To search endogenously for the possibility of any structural break in the relationship between the two time series, this study utilized Perron’s (1997) endogenous unit root test function with the intercept, slope, and the trend dummy to test the hypothesis that the spread has a unit root. Xk ψ ΔSPti þ υt (1) SPt ¼ μ þ θDU þ αt þ γDT þ δDðTb Þ þ βSPt1 þ i¼1 i where DU ¼ 1ðt > Tb Þ is a post-break constant dummy variable; t is a linear time trend; DT ¼ 1ðt > Tb Þ is a post-break slope dummy variable; DðTb Þ ¼ 1ðt ¼ Tb þ 1Þ is the break dummy variable; and υt are white- Downloaded by [University of Florida] at 02:02 25 October 2017 6 C. V. NGUYEN noise error terms. The null hypothesis of a unit root is stated as β ¼ 1. The break date, Tb , is selected based on the minimum t-statistic for testing β ¼ 1 (see Perron 1997). The common methodology to account for a structural break is to introduce a dummy independent variable dt with the value of 1 from the structural break date onward and 0 elsewhere. In the banking sector, a structural break is usually caused by a shock, which precipitates countercyclical monetary policy action by the Central Bank, which results in an interaction between structural break and countercyclical monetary policy measure. Let zt be an independent variable measuring the effect of the interaction between the structural break and policy rate. Econometrically, if the two regressors dt and zt are highly correlated, only zt is included in the model to account for both the structural break and the countercyclical monetary policy precipitated by the shock causing the structural break, and to avoid a single matrix in the estimation process. Model specification To investigate the reactions to or how the Mexican commercial banks responded to changes in countercyclical monetary policy measures by the Mexican Central Bank as reflected in changes in the policy-related rates—i.e., the Mexican interest pass-through—this study follows Wickens and Breusch (1988) and Pereira and Maia-Filho (2013) to specify and estimate an Autoregressive Distributed Lag [ARDL (n, m, s)] model hypothesizing the relationship between the endogenous variable it , and the independent variables rt and zt . The dummy variable dt is excluded from the model because the correlation between zt and dt is 93.52%. it ¼ μ þ n X j¼1 βj itj þ m X k¼0 δk rtk þ s X ρl ztl þ εt (2) l¼0 where “it ” is the lending interest rate and “rt ” is the Central Bank policy-related rate at time t. As defined earlier, zt is an independent variable measuring the effect of the interaction between the structural break and policy rate. δ0 þ ρ0 is the short-run effect—within the month after the Central Bank changes the policy-related rate. It is a priory expectation that 0 < δ0 þ ρ0 ≤ 1. δ0 þ ρ0 < 1 indicates sluggish adjustment, also known as lending rate stickiness. δ0 þ ρ0 = 1 represents a complete pass-through in the short run. Theoretically, the ARDL method proposed by Pesaran (1997) has been a valuable tool for testing for the presence of long-run relationships between economic time series. The advantage of the ARDL model is its ability to estimate both the long-term and short-term model parameters without requiring a pre-testing to determine the order of cointegration of variables; thus, avoiding the problems posed by non-stationary time series. This pre-testing is particularly problematic as Downloaded by [University of Florida] at 02:02 25 October 2017 THE INTERNATIONAL TRADE JOURNAL 7 noted in the unit-root cointegration literature, where the power of unit-root tests is typically very low, and where there is a switch in the distribution function of the test statistics as one or more roots of the right-hand-side variables process approach unity. Furthermore, the ARDL procedure is robust to small samples, allowing different optimal lags of variables. However, Pereira and Maia-Filho (2013) argued that the bounds test is based on the assumption that variables are either I(0) or I(1). Therefore, it is prudent to determine the stationarity of the time series data. The most common testing procedures to test for the stationarity of time series data are Kwiatkowski-Phillips-Schmidt-Shin and Phillips-Perron. As to the empirical estimation, Enders (2015) suggested that the process to estimate the coefficients for Equation (1) is to utilize the Akaike information criterion in selecting the largest values of n, m, and s deemed feasible; the CUSUM test is used to test for model stability. The Breusch-Pagan-Godfrey Heteroskedasticity Test and the Breusch-Godfrey Serial Correlation Lagrange (LM) Multiplier Test are then used as diagnostics to test the hypotheses that the residuals fεt g are white noise and that there is no correlation among independent variables. As articulated by Pereira and Maia-Filho (2013), given the estimation results for Equation (2), the long-run effect is calculated as: m P Φ¼ δk þ k¼0 1 s P l¼0 n P j¼1 ρl (3) βj As articulated by Berstein and Fuentes (2003), Φ should be positive and close to 1. Φ = 1 implies a complete pass-through in the long run, which can be considered evidence of significant competition in the banking system. If Φ< 1 or Φ > 1, it implies either stickiness (less than perfect pass-through) or overshooting, respectively, of retail rates with respect to changes in the policy rate. Explanatory factors include monetary policy regime and the country’s financial structure (Sørensen and Werner 2006). It is therefore important to study the long-run relationship between countercyclical monetary policy and market rates. To this end, this investigation follows Pereira and Maia-Filho (2013) to use the bounds testing approach (Pesaran, Shin, and Smith 2001) for the following error correction representation of the Autoregressive Distributed Lag model: Δit ¼ ϕ þ n X ηj Δitj þ j¼1 þ λ3 zt1 þ εt m X k¼0 πk Δrtk þ s X ωl Δztl þ λ1 it1 þ λ2 rt1 l¼0 (4) 8 C. V. NGUYEN where Δ is difference operator and the null hypothesis of “non-existing of the long-run relationship” is stated as λ1 ¼ λ2 ¼ λ3 ¼ 0. The relevant F-statistics for the joint significance of the λ0 s are calculated and compared with the critical values tabulated by Pesaran, Shin, and Smith (2001). If the estimated F-statistic is greater than the upper-bound critical value, the variables are cointegrated. If it is below the lower bound, the null hypothesis cannot be rejected; i.e., there is no support evidence for a long-run relationship between countercyclical monetary policy and market rates. Downloaded by [University of Florida] at 02:02 25 October 2017 IV. Data and descriptive statistics To empirically discern the aforementioned issues, this study uses the Mexican lending rates it and Central Bank policy rate rt from January 2008 through November 2016 to estimate the autoregressive distributed lag model (2). All time series data are collected from the International Financial Statistics, published by the International Monetary Fund. The mean of the Mexican monthly lending rate during the sample period was 5.17%, and ranged from 3.19% to 10.98% with a standard error of 1.79%. The mean Central Bank policy rate over the same period was 4.57%, and ranged from 3.00% to 8.25% with a standard error of 1.45%. Their correlation was 88.50%. The mean Central Bank policy rate and monthly lending rate spread over the same period were 0.60%, and ranged from 0.05% to 2.73% with a standard error of 0.57%. Finally, the mean of interaction measure between the Central Bank policy rate and the dummy variable is 3.61 and ranged from 0.00% to 5.72% with a standard error of 1.86%. Their correlation is 93.52%. V. Empirical results The empirical results for this investigation are reported as follows. The degree of cointegration The bounds test is based on the assumption that variables are either I(0) or I (1). The most common testing procedures to test for stationarity of time series data are Kwiatkowski-Phillips-Schmidt-Shin and Phillip-Perron. The results of Kwiatkowski-Phillips-Schmidt-Shin and Phillip-Perron tests of the Vietnamese lending rates it and Central Bank policy-related rate rt are summarized in Table 1. The unit root tests reveal lending rates it is an I(1) and Central Bank policy-related rate rt is also an I(1). Pereira and Maia-Filho (2013) argued that it is appropriate to use the bounds test to check for cointegration. THE INTERNATIONAL TRADE JOURNAL 9 Table 1. PP and KPSS test results, Mexican monthly data, January 2008–November 2016. Phillip-Perron Series it rt Note: “n” Level -1.784589n -2.029186n and “y” Kwiatkowski-Phillips-Schmidt-Shin Differencing -5.9191865y -5.736109y Level 0.816914n 0.775821n Differencing 0.179212y 0.318780y indicate whether the series is non-stationary and stationary 5 percent level. Downloaded by [University of Florida] at 02:02 25 October 2017 Structural break The estimation results for Perron’s (1997) endogenous unit root test function with the intercept, slope, and the trend dummy are summarized in Table 2. The estimation results reveal that the post-break intercept dummy variable, DU, is negative and the post-break slope dummy variable, DT, is positive with both statistically significant at the 1% level, while the break dummy, DðTb Þ, is positive but insignificant at any conventional level. The time trend, t, is positive and is significant at the 1% level. These results suggest a stationary trend process. Moreover, strength of the test statistic, tðα ¼ 1Þ ¼ 6:36226, confirms the structural break in September 2008, which may be attributable to the volatilities in the financial sector caused by the U.S. subprime crisis and the plummeted oil price in the first half of 2009 (Sidaoui, Ramos-Francia, and Cuadra 2011). ARDL model As discussed in the methodology section and based on the Akaike information criterion, the estimation process indicates that the optimal values are n = 8, m = 12, and s = 5; as the reported values for AIC in Table 4 suggests, the ARDL (8, 12, 5) model has the lowest AIC value; therefore, it will be used for this investigation. The estimation results and diagnostic statistics for the autoregressive model, ARDL (8, 12, 5), are summarized in Tables 3 and 4 and Figure 1. The left panel of Table 4 reports the diagnostic testing for the correlation among the independent variables and the possibility for the variance of the error term to depend on regressors included in the estimated model. The right panel of Table 4 reveals the AIC values of the five best-estimated models. Figure 1 illustrates the graph of the CUSUM Test over the sample period. Table 2. Perron’s endogenous unit root test, Mexican data: January 2008–November 2016. IPt ¼ 5:63530 5:56819DU 0:28411t þ 0:28502DT þ 0:19416DðTb Þ þ 0:60155IPt1 þ υt (12.08529*) (−11.54002*) (−10.01433*) (9.96765*) (1.12777) (9.60546*) Number of augmented lags: k ¼ 9 Break date: September 2009 tðα ¼ 1Þ ¼ 6:36226 Notes: Critical values for t-statistics in parentheses. Critical values based on n = 100 sample for the break date (Perron 1997). “*” indicates significance at the 1% level. 10 C. V. NGUYEN Table 3. Estimation results for ARDL (8, 12, 5) model and bounds test, January 2008–November 2016. Downloaded by [University of Florida] at 02:02 25 October 2017 ARDL (8, 12, 5): it is a dependent variable Variables i1 i2 i3 i4 i5 i6 i7 i8 r0 r1 r2 r3 r4 r5 r6 r7 r8 r9 r10 r11 r12 z0 z1 z2 z3 z4 z5 Constant R2 = 0.996295 and F-value = 667.3645* ARDL bounds test: Δit is a dependent variable Coefficient t-statistic Variables Coefficient t-statistic 0.287225* 4.263051 Δi1 −0.480539* −4.934355 0.441075* 5.213420 Δi2 −0.039464 −0.564789 −0.072163 −0.883760 Δi3 −0.111627 −1.633997 −0.024536 −0.273524 Δi4 −0.136164*** −1.844220 0.362957** 2.457595 Δi5 0.226793* 3.070623 −0.398171* −3.784385 Δi6 −0.171378** −2.037235 −0.167164 −1.534953 Δi7 −0.338542* −4.349196 0.338542* 4.235678 Δr0 0.657410* 6.948421 0.657410* 6.414960 Δr1 0.542567* 4.419274 0.174239 0.901436 Δr2 0.326645** 2.582475 −0.215922 −1.560865 Δr3 0.236846*** 1.857989 −0.089799 −0.516744 Δr4 0.199007 1.550069 −0.037839 −0.259804 Δr5 0.092635 0.811191 −0.106373 −0.656249 Δr6 0.024456 0.191429 −0.068178 −0.411701 Δr7 0.112956 0.933731 0.088500 0.619441 Δr8 0.258467** 2.463336 0.145511 0.690306 Δr9 −0.013061 −0.126089 −0.271528 −1.336288 Δr10 −0.111615 −1.048256 −0.098554 −0.890684 Δr11 −0.139741 −1.483039 −0.028127 −0.320029 Δz0 −0.001548 −0.039052 0.139741* 2.778805 Δz1 0.094262** 2.455860 −0.001548* −0.043366 Δz2 −0.010874 −0.284760 0.023176 0.442365 Δz3 −0.093306* −2.710082 −0.105135* −2.724873 Δz4 −0.050235 −1.666767 −0.082432* −3.345222 i1 −0.232237* −3.564418 0.043071* 2.596915 r1 0.289082* 3.379084 0.050235* 2.983332 z1 −0.072633* −3.031418 0.136642*** 1.835842 Constant 0.136642*** 1.809604 R2 = 0.994803 R2 = 0.876842 R2 = 0.912217 and and AIC = −1.666088 F-value = 25.78703* and Bounds Test F = 4.269490, k = 2 Note: “*”, “**” and “***” indicate 1%, 5%, and 10% significance levels, respectively. Critical values for bounds tests at 10%: I(0) = 2.63, I(1) = 3.35; 5%: I(0) = 3.10, I(1) = 3.87; 1%: I(0) = 4.13, I(1) = 5.00. Table 4. Diagnostic tests and five best models according to AIC criteria. Diagnostic tests Breusch-Godfrey serial correlation LM test: H0 : There is no serial correlation in the residuals. F(2,65) = 2.124741, p-value = 0.127 Breusch-Pagan-Godfrey Heteroskedasticity Test: H0 : The residual’s variance is constant. F(27,67) = 0.981164, p-value = 0.5051 Model selection criteria Five Best Models ARDL (8, 12, 5) ARDL (9, 12, 5) AIC -1.6661 -1.6552 ARDL (11, 9, 5) ARDL (8, 11, 5) ARDL (12, 9, 5) -1.6550 -1.6548 -1.6540 Note: Data are from calculations by the author. An analysis of the overall estimation results indicates that there is no serial correlation and that the model exhibits strong predictive power, as evidenced by the strength of the Breusch-Godfrey Serial Correlation Lagrange Multiplier Test F(2,65) = 2.124741, with the p-value being 0.1277 which also fails to reject THE INTERNATIONAL TRADE JOURNAL 11 20 10 0 -10 CUSUM Test 5% Significance -20 II III IV I II 2013 III 2014 IV I II III IV I 2015 II III IV 2016 Downloaded by [University of Florida] at 02:02 25 October 2017 Figure 1. Graphic illustration of CUSUM test to test for stability of model’s estimated parameters. the null hypothesis that there is no serial correlation in the residuals. Also, Breusch-Pagan-Godfrey Heteroskedasticity Test, F(27,67) = 0.981164 with the p-value being 0.5052, which fails to reject the null hypothesis that the variance of the residual is constant or no heteroskedasticity. As Figure 1 illustrates, the CUSUM test statistic falls entirely in the band of the 5% level of significance. This empirical finding indicates the stabilities of the estimated parameters of the model over the sample period. Overall, the diagnostic analysis indicates that the estimated ARDL (8, 12, 5) model is very reliable. As reported in Table 3, the estimated sum of δ0 þ ρ0 is 0.6558620 (0.657410–0.001548 = 0.6558620). Also, using Equation (3), the following calculation indicates that the estimated long-run interest rate pass-through rate in the Mexican banking system is Φ = 0.932021. m P Φ¼ δk þ k¼0 1 s P l¼0 n P j¼1 βj ρl ¼ 0:289081 0:07263 ¼ 0:932021 1 0:767765 Finally, to test the null hypothesis of “non-existing of the long-run relationshipH0 : λ1 ¼ λ2 ¼ λ3 ¼ 0” the calculated value of the relevant F-statistic being 4.269490 for the joint significance of the hypothesis is compared to the critical upper values bounds at the 5% level of significance. Comparing the value of the F-statistic of 4.269490 to the critical value of the upper bound I(1) = 3.87 indicates that the null hypothesis of “non-existing of the long-run relationship” in the Mexican banking sector should be rejected at the 5% level of significance, suggesting that there is a long-term relationship between the Central Bank policy rate and the lending rate in the Mexican commercial banks’ lending market. 12 C. V. NGUYEN Downloaded by [University of Florida] at 02:02 25 October 2017 VI. Discussion of the empirical results The endogenous search process for breaks in the Mexican interest rate structure using Perron’s (1997) endogenous unit root test function with the intercept, slope, and trend dummy found that the relationship between the Central Bank policy-related rate and the commercial bank’s lending rate experienced a break in September 2009. To account for this structural break, this investigation introduced a dummy variable and assigned the value of 1 from September 2009 onward and 0 elsewhere over the sample period. Econometrically, this introduction of the dummy variable precipitated the generation of the interaction term between the dummy variable and the Central Bank policy rate. Due to the high correlation between the two regressors, only the interaction term was included in the modelling process. An analysis of the overall estimation results indicates that there is no serial correlation and that the model exhibits strong predictive power, and also confirms that the estimated residuals are white noise. The estimation results of the Autoregressive Distributed Lag, ARDL (8, 12, 5) model, represented by Equation (2) and the derived long-run rates of pass-through, reveal that the Mexican short-run rate of pass-through, δ0 þ ρ0 = 0.6558620, is relatively high as compared to empirical magnitudes reported in the literature relating to emerging and advanced economies (Alencar 2011; Nguyen 2017; Pereira and Maia-Filho 2013; Wickens and Breusch 1988). This finding is consistent with the observed lack of competition and highly concentrated operating environment in the Mexican banking industry. The empirical results are also consistent with reports by Sørensen and Werner (2006) that degrees of competition in the national banking sector are the most significant determinants of passthrough speed. Based on the Akaike information criterion, the longest lag retained by the estimation process for the commercial lending rate is 8 (i8 ) and for the Central Bank policy-related rate is 12 (r12 ). These findings suggest that Mexican commercial banks considered their lending rate eight months back in determining their current lending rate, while these lending institutions took up to 12 months to respond to monetary policies completely. The calculated long-run pass-through rate in the Mexican banking industry is Φ = 0.932021. Finally, one way to measure the credibility of the Central Bank is the existence of the long-run relationship between its countercyclical monetary policy, as reflected in changes in the Central Bank policy-related rate, and the commercial banks’ lending rate in the lending markets. In this investigation, this issue is addressed by testing the previously stated null hypothesis H0 : λ1 ¼ λ2 ¼ λ3 ¼ 0. The testing procedure indicated that this hypothesis should be rejected at the 5% level of significance, suggesting a THE INTERNATIONAL TRADE JOURNAL 13 long-run relationship between the countercyclical monetary policy and the commercial banks’ lending rate in the lending markets. Downloaded by [University of Florida] at 02:02 25 October 2017 VII. Concluding remarks Historically, the Mexican open economy is relatively very small as compared to (although strongly tied to) the U.S. economy, even more so after the implementation of NAFTA. Additionally, the Mexican Central Bank’s form of rigid inflation targeting unintentionally adds a large element of unpredictability to the exchange rate. Even though Mexico’s commercial banking sector has been open to foreign competition, it is still very concentrated and lacks any natural competition. Financial intermediation is a critical facilitator of investment and economic growth. Commercial banks are an integral part of the monetary policy transmission mechanism since, through their interest rate pass-though, commercial banks change the lending rates in the economy which, in turn, transmit the countercyclical monetary policy measures to consumption and investment activities of the economy. Changes in these two macroeconomic variables will change the macroeconomic policy target variables: unemployment, inflation, and GDP. These aforementioned facts of the Mexican banking sector motivated this investigation to learn the nature of the Mexican interest rate pass-through. To achieve this objective, this study utilizes an Autoregressive Distributed Lag model to empirically investigate the nature of interest rate pass-through in the Mexican economy. Estimation results suggest that, based on the Akaike information criterion, the ARDL (8, 12, 5) model best fits the data. Estimation results of the ARDL (8, 12, 5) model reveal that the Mexican short-run rates of pass-through (δ0 þ ρ0 is 0.6558620) are relatively high as compared to empirical magnitudes reported in the literature relating to emerging and advanced economies. This finding is consistent with the observed lack of competition in the Mexican banking industry. The empirical results also indicate that the Mexican commercial banks considered their lending rate eight months back in determining their current lending rate, while these lending institutions took up to 12 months to completely respond to the monetary policies. The calculated long-run pass-through rate in the Mexican banking industry is Φ = 0.932021, which indicates almost complete pass-through in the Mexican banking industry. The procedure testing the null hypothesis H0 : λ1 ¼ λ2 ¼ λ3 ¼ 0 indicated that this hypothesis should be rejected at the 5% level of significance, suggesting a long-run relationship between the countercyclical monetary policy and the commercial banks’ lending rate in the lending markets. Downloaded by [University of Florida] at 02:02 25 October 2017 14 C. V. NGUYEN Rejection of this hypothesis suggests that the Mexican Central Bank has very good credibility. Finally, one way to measure the credibility of the Central Bank is to examine the long-run relationship between countercyclical monetary policy, as reflected in changes in the Central Bank policy-related rate, and the commercial banks’ lending rate in the lending markets. The rejection of the earlier null hypothesis of no long-run relationship is an indication of credibility for the Mexican Central Bank. Notwithstanding the rigid inflation targeting set by the Central Bank, a very concentrated market, the openness to foreign competition, and its relatively very small size as compared to (although strongly tied to) the U.S. economy, these aforementioned findings suggest that the Mexican Central Bank has been very effective in formulating and implementing its countercyclical monetary policy. References Alencar, L. S. 2003. “O Pass-Through da Taxa Básica: Evidência Para as Taxas de Juros Bancárias.” In Relatório De Economia Bancária E Crédito, 90–100. Banco Central do Brasil. Retrieved from https://www.bcb.gov.br/ftp/rel_economia_bancaria_credito.pdf Alencar, S. L. 2011. “Revisiting Bank Pricing Policies in Brazil: Evidence from Loan and Deposit Markets.” Working Paper Series 235, Edited by Benjamin Miranda Tabk, Research Department, Banco Central Do Brasil. Retrieved from https://www.bcb.gov.br/pec/wps/ ingl/wps235.pdf Bernanke, B. S., and A. Blinder. 1992. “The Federal Funds Rate and the Channels of Monetary Transmission.” American Economic Review 82:901–21. Berstein, S., and R. Fuentes 2003. “From Rate to Bank Lending Rate: The Chilean Banking Industry.” LACEA Papers and Proceedings, Mexico. Retrieved from http://citeseerx.ist.psu. edu/viewdoc/download?doi=10.1.1.198.4135&rep=rep1&type=pdf Borio, C., and W. Fritz. 1995. “The Response of Short-Term Bank Lending Rates to Policy Rates: A Cross-Country Perspective.” BIS Working Papers, No 27. Money and Economic Department, Bank for International Settlements Basle. Retrieved from http://www.bis.org/ publ/work27.pdf Calem, P., and L. Mester. 1995. “Consumer Behaviour and the Stickiness of Credit-Card Interest Rates.” The American Economic Review 85 (5):1327–36. Coelho, C., J. De Mello, and M. Garcia. 2010. “Identifying the Bank Lending Channel in Brazil through Data Frequency.” Economía 10:47–79. doi:10.1353/eco.2010.0004. Cottarelli, C., and A. Kourelis. 1994. “Financial Structure, Bank Lending Rates, and the Transmission Mechanism of Monetary Policy.” IMF Staff Papers 41 (4):587–623. doi:10.2307/3867521. Enders, W. 2015. Applied Econometric Time Series, 4th ed. Hoboken, NJ: John Wiley and Sons. Engle, R., and C. Granger. 1987. “Cointegration and Error Correction: Representation, Estimation, and Testing.” Econometrica 251–76. doi:10.2307/1913236. Espinosa-Vega, M., and A. Rebucci 2003. “Retail Bank Interest Rate Pass-Through: Is Chile Atypical?” IMF Working Paper 03/112. Western Hemisphere Department, International Downloaded by [University of Florida] at 02:02 25 October 2017 THE INTERNATIONAL TRADE JOURNAL 15 Monetary Fund, Washington, D.C. Retrieved from https://www.imf.org/external/pubs/ft/ wp/2003/wp03112.pdf European Central Bank. 2003. “New ECB Statistics on MFI Interest Rates.” Monthly Bulletin, December. www.ecb.europa.eu/pub/pdf/other/mb200312_focus02.en.pdf? c43d0eb21799464e33dfbf13969fd061 Gambacorta, L. 2008. “How Do Banks Set Interest Rates?” European Economic Review 52:792–819. doi:10.1016/j.euroecorev.2007.06.022. Hannan, T., and A. Berger. 1991. “The Rigidity of Prices: Evidence from the Banking Industry.” American Economic Review 81 (4):938–45. Hofmann, B., and P. Mizen. 2004. “Interest Rate Pass-Through and Monetary Transmission: Evidence from Individual Financial Institutions’ Retail Rates.” Economica 71:99–123. doi:10.1111/ecca.2004.71.issue-281. Hutchison, D. 1995. “Retail Bank Deposit Pricing: An Intertemporal Asset Pricing Approach.” Journal of Money, Credit, and Banking 27 (1):217–31. doi:10.2307/2077860. Illes, A., and M. Lombardi 2013. “Interest Rate Pass-Through Since the Financial Crisis.” Bank of International Settlements Quarterly Review. Accessed September 2013. http:// www.bis.org/publ/qtrpdf/r_qt1309g.pdf. International Trade Administration. 2016. “Mexico–Banking Systems, International Trade Administration.” Washington, DC: U.S. Department of Commerce. https://www.export. gov/article?id=Mexico-Banking-Systems. Manning, J. 2015. “International Banker: The Outlook for Mexico’s Banking System.” Accessed July 7, 2015. https://internationalbanker.com/banking/the-outlook-for-mexicosbanking-systems. McKinnon, R. I. 1973. Money and Capital in Economic Development. Washington, DC: Brookings Institute. Mojon, B. 2000. “Financial Structure and the Interest Rate Channel of ECB Monetary Policy.” European Central Bank Working Paper Series 40. Retrieved from https://www.ecb.europa. eu/pub/pdf/scpwps/ecbwp040.pdf?a1cb4280848b9c3557120a146468f3ab Neumark, D., and S. A. Sharpe. 1992. “Market Structure and the Nature of Price Rigidity: Evidence from the Market for Consumer Deposits.” Quarterly Journal of Economics 107:657–80. doi:10.2307/2118485. Nguyen, C. V. 2017. “The Brazilian Interest Rate Pass-Through.” Journal of Business and Economics Perspectives 44 (1):29–39. Patrick, H. T. 1966. “Financial Development and Economic Growth in Underdeveloped Countries.” Economic Development and Cultural Change 14 (2):174–89. doi:10.1086/ 450153. Pereira, C. M., and L. F. Maia-Filho. 2013. “Brazilian Retail Banking and the 2008 Financial Crisis: Were the Government-Controlled Banks that Important?” Journal of Banking & Finance 37 (7):2210–15. doi:10.1016/j.jbankfin.2012.03.009. Perron, P. 1997. “Further Evidence on Breaking Trend Functions in Macroeconomic Variables.” Journal of Econometrics 80:355–85. doi:10.1016/S0304-4076(97)00049-3. Pesaran, M. H. 1997. “The Role of Economic Theory in Modelling the Long Run.” The Economic Journal 107 (44):178–91. doi:10.1111/1468-0297.00151. Pesaran, M. H., Y. Shin, and R. J. Smith. 2001. “Bounds Testing Approaches to the Analysis of Level Relationships.” Journal of Applied Econometrics 16:289–326. doi:10.1002/jae.616. Rosen, R. 2002. “What Goes Up Must Come Down? Asymmetries and Persistence in Bank Deposit Rates.” Journal of Financial Services Research 21 (3):173–93. doi:10.1023/ A:1015085826129. Sander, H., and S. Kleimeier. 2004. “Convergence in Euro-Zone Retail Banking? What Interest Rate Pass-Through Tell Us about Monetary Policy Transmission, Competition Downloaded by [University of Florida] at 02:02 25 October 2017 16 C. V. NGUYEN and Integration.” Journal of International Money and Finance 23 (3):461–92. doi:10.1016/j. jimonfin.2004.02.001. Schumpeter, J. A. 1912. “Theorie der WirtschaftlichenEntwicklung.” Revised English translation by Redvers Opie. In The Theory of Economic Development. Cambridge, MA: Harvard University Press, 1934. Retrieved from http://www.hup.harvard.edu/catalog.php?isbn= 9780674879904 Sidaoui, J., M. Ramos-Francia, and G. Cuadra 2011. “The Global Financial Crisis and Policy Response in Mexico.” http://www.bis.org/publ/bppdf/bispap54q.pdf. Sørensen, C. K., and T. Werner 2006. “Bank Interest Rate Pass-Through in the Euro Area: A Cross Country Comparison.” European Central Bank Working Paper Series 580. Retrieved from https://www.ecb.europa.eu/pub/pdf/scpwps/ecbwp580.pdf?630. c54c4f654bbce1b9cdfbe2bea9a10 Stiglitz, J., and A. Weiss. 1981. “Credit Rationing in Markets with Imperfect Information.” American Economic Review 71 (3):393–410. Tonooka, E. K., and S. M. Koyama. 2003. “Taxa de Juros e Concentração Bancária no Brasil.” Trabalho Para Discussão Banco Central Do Brasil 62:35. Weisbrot, M., S. Lefebvre, and J. Sammut. 2014. Did NAFTA Help Mexico? An Assessment after 20 Years. Washington, DC: Center for Economic and Policy Research. Retrieved from http://cepr.net/documents/nafta-20-years-2014-02.pdf Weth, M. 2002. “The Pass-Through from Market Interest Rates to Bank Lending Rates in Germany.” Volkswirtschaftliches Forschungszentrum der Deutschen Bundesbank, Discussion Paper 11/02. Retrieved from https://www.bundesbank.de/Redaktion/EN/ Downloads/Publications/Discussion_Paper_1/2002/2002_05_12_dkp_11.pdf?__blob= publicationFile Wickens, M., and T. Breusch. 1988. “Dynamic Specification, the Long Run and the Estimation of Transformed Regression Models.” Economic Journal 98:189–205. doi:10.2307/2233314.