Financial and Real Integration between Mexico and
the United States
⇤
Kristiana Rozite
†2018
Abstract
Mexico is a small open economy in need of a macro-prudential policy frame-work which is designed to safeguard itself from major US shocks. Hence it is important to quantify the contribution of the US factors in the fluctuations of Mexican macroeconomic indicators. In this paper, we separate the roles of US and MX financial and business cycles in describing MX indicator dynamics. Our results suggest that MX household leverage has domestic financial cycle. MX non-financial leverage has complex short and long-term links to the US economy. While MX non-financial leverage growth rates has negative association with the short-term US GDP cyclical component, it relates positively to the short-term and negatively to the long-term US investor sentiment cyclical components. We find that MX stock price index and GDP growth rates contain US business cycle components with increased role in the late NAFTA period.
⇤The views and conclusions presented in this paper are exclusively the responsibility of the author
and do not necessarily reflect those of Banco de Mexico. This version of the paper was developed during the summer internship program at the Financial Stability General Directorate of Banco de Mexico in the summer of 2017. I thank Tom Wansbeek, Alberto Romero, Irma Hindrayanto, Dirk Bezemer, Jan Jacobs and the participants of the Research seminar at Banco de Mexico for valuable comments.
†Correspondence to: Kristiana Rozite, Faculty of Economics and Business, University of Groningen,
1
Introduction
The aim of this paper is to quantify the exposure of Mexican economy to the US fac-tors. In particular, we decompose Mexican household and non-financial corporation leverage, GDP growth rates, and stock market price index returns into their domestic and US short and long-term cyclical components over 1981:Q1–2016:Q1. We also in-vestigate changes in US business cycle e↵ects on Mexican stock price index and GDP growth rates in the late North American Free Trade Agreement (NAFTA) period 2000– 2016. A recent policy debate stresses that whenever global factors explain a large part of local macroeconomic fluctuations, an independent monetary policy for small open economies is an implausible construct.1 Recognition of monetary policy limitations has lead to development of parallel frameworks such as macro-prudential regulations which are crucial to support the smooth operations of markets (see e.g., Kose, Otrok, and Whiteman, 2003). Knowledge of the extent and nature of international exposure is necessary to refine financial stability objectives. Current Mexican macro-prudential policy framework includes such instruments as countercyclical capital bu↵ers, loan to value ratios and limits on currency mismatches (see e.g., Upper, 2017 ).
There are several reasons to expect similar dynamics in the US and Mexico. Mexico is a small open economy with strong transmission of US shocks due to its real and financial integration with the US markets. At the same time, domestic structural re-forms and macroeconomic policies must be taken into account. Before 1996 domestic volatility swamped the role of US factors in the fluctuations of macroeconomic indica-tors (see e.g., Swiston and Bayoumi, 2008). The time period after 1995 is described as the Great Moderation of the Mexican business cycle with less volatility in key macroe-conomic indicators. One explanation for this change in dynamics is an improvement in 1These global factors include monetary policies of the Federal Reserve and of other major central
the monetary policy framework and a stronger fiscal position.
Mexico has strong real integration with US since it is its largest import and export market. In 1994 the NAFTA agreement established a tri-lateral trade block between Canada, Mexico and the US. Empirical evidence suggests that US influence on the Mexican economy increased in the post NAFTA period resulting in closely synchro-nized business cycles and increased cross-country correlations among major macroeco-nomic aggregates. As pointed out by Swiston and Bayoumi (2008), these correlations may result from similar responses to common shocks, idiosyncratic shocks that happen to be correlated across countries, or spillover e↵ects where one country is responding to shocks originating in other. Potential channels for spillovers include trade, wage remittances and Foreign Direct Investment (e.g. see L´opez-C´ordova, Hern´andez and Monge-Naranjo, 2003; Kose, Meredith, Towe, 2004; Arora and Vamvakidis, 2005; Sosa, 2008).
Financial integration plays a role in shock transmissions as well. Arora and Cerisola (2001) argue that the rise in US risk-free benchmark rates, can increase emerging market spreads through e↵ects on cost, availability of funds and creditworthiness. Uribe and Yue (2006) show that US interest rate shocks a↵ect emerging country spreads which first fall and then overshoot. The e↵ects of US monetary shocks on emerging markets are substantial and explain an important proportion of fluctuations in macroeconomic variables (see e.g., Canova, 2005 ; Mackowiak, 2007).
in 2000 (see Maudos and Solis, 2011). Bekaert and Harvey (1995) examine several economies which exhibit time-varying integration with international capital markets. They note that the Mexican stock market had been closed to foreign investment until 1981 and before 1991 only one American depositary receipt (ADRs) was traded. In 1989 the Mexican stock market opened to foreign investors with exception of some key sectors and in 1991 the dual exchange rate was abolished. Clark and Berko (1997) examine the Mexican equity market and note that the broadening in the investor base increases risk sharing and liquidity and should decrease expected returns and increase covariance with the US markets.2 Rey (2015) based on the dataset of more than 50 countries provides aggregate level empirical evidence that there is a US financial cycle in capital flows, asset prices and in credit growth which co-moves with VIX, a measure of uncertainty and risk aversion of the markets.
The current paper does not examine any specific transmission mechanism of US shocks to the Mexican economy, but investigates their aggregate e↵ect. Conceptually similar work to the present paper is by Kose, Meredith and Towe (2004). They use a dynamic latent factors model with regional and country-specific factors to analyze Mexican output, consumption and investment series over the period 1980–2002. This paper analyzes Mexico’s household, non-financial corporation leverage, GDP and stock price index growth rates. We relate these indicators to US investor sentiment, household leverage and GDP growth rates using the concepts of business and financial cycles. Traditionally business cycles have a period of two to eight years whereas financial cycles duration is more than eight years and have an average duration of around 16 years. The distinction between the two types of cycles is relevant since Mexico’s short-term and long-short-term links to the US may di↵er in nature. We compare how the role of a US business cycle component changed for Mexico’s GDP and stock price index growth 2A stock with a restricted investor base pays premium which is an increasing function of stock’s
rates from 1981:Q1–1999:Q1 to 2000:Q1–2016:Q1. We refer to the latter period as the late NAFTA subsample. Our subsample choice is motivated by the fact that from 2000 Mexico had opened its banking sector and equity market to foreign participation and achieved monetary stability.
The results suggest that Mexican household and non-financial corporation leverage contain a domestic financial cycle component. Interestingly, we find no evidence that Mexican household leverage would be related to US financial cycle components. We propose to relate this finding to low Mexican household financial inclusion determined by labor market informality, low social security and modest changes in real wages. In contrast, Mexican non-financial corporation leverage has complex short and long-term links to the US economy. First, there is a negative association with the short-term US GDP cyclical component. Using insights from Fernandez and Gulan (2015) we reason that positive foreign demand shock can increases equity through increased profits by so reducing a firm’s overall leverage. Second, Mexican non-financial corporation leverage relates positively to the short-term and negatively to the long-term US investor sentiment cyclical component. This may relate to the earlier findings by Morais, Pedro and Ruiz (2015) wherein they link softening of the foreign monetary policy to increased lending by foreign banks to risk-seeking Mexican firms. These firms typically have higher ex-post defaults which may explain the negative long-term e↵ect on leverage. Empirical evidence suggests that US business cycle e↵ects became more important for the Mexican economy in 2000:Q1–2016:Q1. In this sample period, main part of the cyclical movements in Mexican stock price index returns and GDP are due to the US business cycle components.
Section 6 the findings are discussed and Section 7 concludes.
2
Data
Our dataset contains four Mexican and three US quarterly-frequency economic activity indicators. The dataset spans 1981:Q1–2016:Q1 hence half of the sample observations relate to the NAFTA period. We analyze the role of short and long-term Mexican and US cyclical components in describing the dynamics of Mexican indicators. Short and long-term cyclical components are related to the concepts of business and financial cycles respectively. In a nutshell, a business cycle lasts from two to eight years whereas a financial cycle eight years and longer. Borio (2014) estimates that in the sample of seven industrialized countries a financial cycle takes on average 16 years (see also Drehmann, Borio and Tstasaronis, 2012).
For our model input we transform non-stationary indicators into stationary growth rates. The common practice is to use observations in levels and to model stationary components alongside non-stationary trends using for example model based filters. We deviate from this practice for several reasons. First, modeling stochastic trends are not of direct interest in our case since we are interested in growth rate cycles primarily. Second, the econometric model-based approaches to detrending frequently are prompt to calibration choices and can generate spurious cycles (e.g. Harvey and Jaeger, 1993). Since empirically it is difficult to separate a stochastic trend due to its weak signal to noise ratio, estimations of the state-space system are facilitated by calibrating the stochastic trend parameters.3 Third, we di↵erence our data because the process we observe might have higher frequency components than the frequency of sampling. In 3These choices are based on the objective to obtain smoothness and some persistence of the long
that situation an e↵ect known as aliasing occurs in which high frequency components get translated to low frequencies, which can be confused for actual trends.
Instead of modeling stochastic trends, one can use non-parametric detrending meth-ods such as Hodrick and Prescott (1980), the band-pass filters of Baxter and King (1999) and Christiano and Fitzgerald (2003) which require prior assumptions on the length of the cycle. As a result of these prior assumptions, estimation problems may be encountered. For example, low frequency cycles outside the pre-specified band can be classified as part of the long-term trend (see e.g., Comin and Gertler, 2006; Cog-ley and Nason, 1995; Igan, Kabundi, Simone, Pinheiro and Tamirisa, 2009; Galati, Koopman Hindrayanto and Vlekke, 2016). These considerations determine our choice to first-di↵erence non-stationary data. Having discussed the motivation to work with stationary data, we are now ready to introduce our selected indicators.
Traditionally, a financial cycle is described with information in asset prices and credit variables. Rozite, Bezemer and Jacobs (2016) investigate US financial stress and crisis moments and conclude that two factors extracted from household, non-financial leverage, real estate price index, stock market price index returns and purchasing man-ager index relate well to these developments. Namely, crisis moments usually coincide either with troughs in household or corporate sentiment.
local currency units. Household leverage is defined as credit to households scaled by wages and salaries. The second indicator we collect is investor sentiment index created by Baker and Wurgler (2006) (SENTG,t). The index is based on the first principal component of five (standardized) sentiment proxies: the closed-end fund discount which is the average di↵erence between the net asset values of closed-end stock fund shares and their market prices; value-weighted dividend premium; first-day returns on IPOs; IPO volume and equity share in new issues. This sentiment index is not orthogonal to macroeconomic conditions, hence it also is a business cycle indicator. We refer to either of these financial cycle indicators by FIG,t. In addition, the indicator for the US business cycle is real GDP growth rate (GDPG,t).
The selection of Mexican financial and real activity indicators is constrained by data availability. From Banco de Mexico we obtain the consumer price index (CPI) with reference year 2010; GDP in local currency units in nominal terms; loans to non-financial enterprises; consumption and mortgage loans to households in nominal terms and Mexican stock price index with reference year 2010.4 Mexican financial and real activity indicators are seasonally adjusted using the X-13 ARIMA-SEATS procedure. In addition, the OECD recession indicator for Mexico is collected. Using this information, we check the timing of peaks and troughs in the Mexican series. Mexican household leverage is computed as total loans to households scaled with nominal GDP, thereafter, quarter on quarter logarithmic growth rates are computed (HHLEVM,t). The same procedure is followed to construct growth of leverage of non-financial corporation leverage (NFLEVM,t).5 Mexican stock price index and nominal GDP are adjusted for inflation using CPI, and converted to quarter on quarter log growth rates (SPM,t, 4Unfortunately data on Mexico’s real estate price index are available only from 2005 hence this
variable is omitted from our analysis.
5In the data there is a sharp decrease in the credit levels from 1995 and onwards due to the
GDPM,t). Figure 1 shows the development of these indicators.
The US and MX indicators do not contain unit roots and hence we proceed with standardizing the data (see Appendix A for further details). In addition, since sud-den crisis episodes blur economic regularities these episodes are controlled for using dummy variables. To detect outliers we run regressions for each observed indicator on a constant. Outliers were detected inspecting t-statistic on a dummy variable defined to equal one for the date of interest. The results were robust to other outliers detection methods and were checked against formally documented sudden crisis episodes. The selected dates include the following episodes: the 1982:Q4 debt crisis present in the Mexican leverage growth rates; 1986:Q1 oil shock present in Mexican GDP and non-financial leverage; the 1988:Q1 stock market crash present in the Mexican stock market returns and non-financial leverage; the 1993:Q1 change in peso-dollar parity present in Mexican GDP and the 1995:Q2 Tequila crisis is evident in all Mexican indicators.
3
Stylized Facts
Figure 1: The US and MX financial and economic activity indicators 1981:Q1–2016:Q1.
(a) US household leverage GR (HHLEVG,t) (b) US investor sentiment (SENTG,t)
(c) US real GDP GR (GDPG,t)
(d) MX household leverage GR (HHLEVM,t)(e) MX non-financial leverage GR (NFLEVM,t)
(f) MX stock price index GR (SPM,t) (g) MX GDP GR (GDPM,t)
We observe the following trends in the data. Mexican household leverage has steadily increased since the early 2000s, reaching around 16% in 2016. This is significantly less than in the US where household leverage reached close to 100% just before 2008 when it dropped and gradually recovered over the late sample period to 80%. Mexican non-financial leverage in 2016 was 27% which is considerably less compared to the ratio in the early 2000s when it was 50% of GDP. In comparison, non-financial leverage in US has generally increased throughout the sample reaching 72 % just before crisis in 2008 thereafter it dropped to 63% and recovered by the end of our sample to pre-crisis level of 72%. The stock market capitalization to GDP in Mexico has generally increased since early 2000s reaching 34% of GDP at the end of our sample. In US the stock market capitalization reached its peak of 146% in 2000s thereafter in 2008 the level dropped to 93% and gradually increased to the pre-crisis level of 146% in the late sample period. At the end of our sample period Mexico had 143 on stock market listed companies whereas in US this number is around 4300. It follows that for Mexican economy leverage plays relatively smaller role as compared to US. If we were to rank Mexican financial indicators in their relevance relative to GDP then stock market is the most relevant, followed by non-financial leverage and finally the household leverage.
Table 1: Kendall rank correlations for Mexican and US economic activity indicators.
HHLEVG,t SENTG,t GDPG,t HHLEVM,t NFLEVM,t GDPM,t SPM,t
1981–2016 1 0.066 1 0.011 0.118b 1 0.032 0.183c 0.002 1 0.094a 0.132b 0.125b 0.438c 1 0.005 0.075 0.072 0.023 0.005 1 0.111b 0.058c 0.191c 0.075 0.094a 0.125b 1 1981–1999 1 0.038 1 0.148b 0.090 1 0.082 0.250c 0.065 1 0.037 0.163b 0.085 0.440c 1 0.004 0.089 0.040 0.024 0.107 1 0.196c 0.074 0.141a 0.030 0.059 0.066 1 2000–2016 1 0.1082 1 0.1067 0.058 1 0.0288 0.052 0.072 1 0.1615b 0.061 0.174b 0.377a 1 0.0327 0.050 0.218c 0.002 0.146a 1 0.0577 0.072 0.260c 0.133 0.179b 0.217c 1 Notes: a p < 0.10, b p < 0.05, c p < 0.01, SENT
G,t is US investor sentiment, HHLEVG,t is US
household sentiment.
period. Examining 1981–1999 and 2000–2016 subsamples we find no general indication that correlations between US and MX increased for real or financial indicators. One ex-ception is the correlation between US GDP and Mexican stock price index returns. US sentiment index correlates with Mexican leverage indicators in the full and 1981–1999 sample periods but is consistently uncorrelated with Mexican GDP. Lastly, Mexican and US GDPs are correlated only in the late NAFTA period.
For robustness, Table 2 shows the timing of peaks and troughs in the indicators. These are obtained using the BBQ turning point detection algorithm which as the input for estimations requires few parameters: a window length of surrounding points for a peak or trough, a minimum phase length and a minimum cycle length (see Harding and Pagan, 2002). These parameters are selected to capture well all peaks and troughs in a sample. For our sample we set the window length at 12 quarters, the minimum phase length at 2 quarters and a minimum cycle length at 12 quarters.
Table 2: The turning points of US and Mexican indicators.
Indicator Peaks Troughs
HHLEVG,t 1985-1 1994-1 2001-3 2013-1 1987-4 2000-1 2012-1 SENTG,t 1984-1 1993-1 2001-1 2007-1 1989-3 1998-4 2003-4 2009-4 GDPG,t 1987-4 2000-2 2003-3 1990-4 2001-3 2008-4 HHLEVM,t 1984-4 1989-1 2005-4 2012-3 1987-1 1996-4 2009-3 NFLEVM,t 1988-4 2008-4 1988-1 1998-4 2009-4 SPM,t 1987-1 1991-2 1999-2 2009-3 1987-4 1995-1 2008-4 GDPM,t 1993-1 2002-2 2011-4 1986-1 1995-2 2008-4 2013-1
Notes: The turning points are documented using the BBQ algorithm of Bry and Boschan (1971) extended by Harding and Pagan (2002). The parameters for the algorithm were set as follows: a window length is 12 quarters, min phase length 2 quarters, min cycle length 12 quarters.
crisis in 1995 but this episode does not reflect in the troughs of US indicators which suggests that specifying US indicators as exogenous to the Mexican economy is not an invalid assumption. Lastly, there are no clear patters regarding the peak-trough regularities between US and Mexican leverage cycles.
4
Model
This Section introduces a model to investigate the roles of US and MX financial and business cycles in describing the long and short term movements in Mexican macroe-conomic indicators. Similarly as in Koopman, Lit and Lucas (2016) we define a State-Space model such that indicators of a country can contain financial and business cycles (see also Koopman and Lucas, 2005). To perform plausible estimations, the system of a State-Space model is kept parsimonious in observed indicators, parameters and state variables. However, in order to analyze the cyclical dynamics of an emerging economy, the necessary extension is to link its indicators to the cyclical components of an ad-vanced economy. In the case of Mexico, the role of US is important and hence we need to include Mexican and US financial and business cycle components. To our current knowledge it is the novelty to link observed indicators to local and global financial and business cycles simultaneously.
Let the N -vector yt = [y0G,t, y0M,t]0 contain stationary financial and real activity US and Mexican indicators. For our full sample estimations, we consider two variants how US indicators enter the observation vector given by
and Mexican indicators are given by
yM,t= (HHLEVM,t NFLEVM,t SPM,t GDPM,t)0
where we introduced the indicators in Section 2. The observed indicators (yt) are related to unobserved cyclical components (↵t) defined by
yt = Z↵t+ et, et ⇠ N(0, H),
↵t = T ↵t 1+ ⌘t, ⌘t ⇠ N(0, Q) (1)
where the k-state vector ↵t = ( G,t0 , G,t0 , 0M,t, M,t0 )0 collects long-term and short-term cyclical components. The subscript G is used to distinguish between a US (business or financial) cycle as opposed to a local Mexican cycle denoted with M. Each cyclical component is of the form G,t = ( G,t, G,t⇤ )0. The components marked with a star result from writing the trigonometric components in a recursive form and have no real interpretation.
The (N ⇥ k) matrix Z = [ZG ZM]⌦ e01 contains factor loadings for the US and Mexican cyclical components. Based on univariate analysis to be discussed in Section 6.1, Z has two variants. If FIG,t is defined as HHLEVG,t then Z is given by
ZG= 0 B @1 0 ⇤ ⇤ 0 0 0 1 ⇤ ⇤ ⇤ ⇤ 1 C A 0 , ZM = 0 B @0 0 1 ⇤ 0 0 0 0 ⇤ ⇤ 1 ⇤ 1 C A 0 (2)
but if FIG,t is defined as SENTG,t then
where ⇤ denotes an unrestricted element and e0
1 = (1 0). ZG and ZM contain iden-tification restrictions. In order to fix the scale of a cyclical component one observed indicator is chosen as a base and is linked to one cyclical component with a fixed fac-tor loading equal to one (see e.g., Koopman et al., 2005); we assume that none of the US indicators is influenced by the Mexican indicators. Other restrictions are based on univariate analysis to be discussed in more detail in Section 6.1.
The (8⇥ 8) matrix T = blkdiag[R , R , R , R ] is a block diagonal and describes the state transition for long and short term US and Mexican cyclical components. The cyclical components are specified in the form of a stochastic trigonometric cycle with a state transition matrix given by
Rj = j 0 B @ cos j sin j sin j cos j 1 C A , j = , , (4)
where the persistence parameter j 2 (0, 1) and j 2 (2⇡/T, ⇡) is a frequency parameter (see Harvey, 1989). The period of a cycle expressed in years is given by Pj = (2⇡/ j)/4. In the model there are two sets of similar cycles. First, the Mexican and US business cycles are similar. Second, the Mexican and US financial cycles are restricted to be similar. Hence they share the same frequency and persistence parameter (see Harvey and Koopman, 1997).
The (8⇥ 8) matrix Q = diag[ 2
G, 2 G, 2 M, 2
M]⌦ I2 and the (N ⇥ N) matrix H
are both diagonal and positive definite.
To see the relative importance of each factor, the covariance of observed indicators can be decomposed into the cyclical variance resulting from each factor and the noise part given by
where P has elements given by vec(P ) = I64 (T ⌦ T ) 1Q, H is the noise part.
5
Estimation
Model (1) is estimated by maximizing the log-likelihood over a set of model param-eter values ( ). The paramparam-eters are then used to filter the unobserved states using the Kalman filter. Finally we apply the smoother which improves the Kalman filter estimates of unobserved states at time t by using the full sample information (see e.g., Durbin and Koopman, 2001).
Since all the state variables are covariance stationary, their initial values are set to have unconditional distributions. For example the initial state of the US long-term cyclical component is distributed as G,0 ⇠ N(0, 2G/(1 2G)).
To implement the optimization process without constraints some model parameters are re-parametrized (e.g., Koopman and Azevedo, 2007). A typical diagonal element of a covariance matrix is specified as exp(2ci) for some ci 2 R. A cyclical component t has a period P = exp(✓) where ✓ 2 R hence the corresponding frequency parameter is = 2⇡/P . A persistence parameter for a cyclical component is parametrized as
= exp(⌫)/(1 + exp(⌫)) for some ⌫ 2 R.
which is then used to obtain the standard errors for the model parameters ( ) taking the square root of the diagonal entries of: ⌦ = [ @2log L/@ 0] 1.
6
Results
In this Section we separate the roles of US and Mexican financial and business cycles in describing the dynamics of the Mexican macroeconomic indicators. Section 6.1 mo-tivates the specification choices for multivariate model described in Section 4. Section 6.2 estimates US financial and business cycle e↵ects for the Mexican economy during 1981–2016. In Section 6.3 we perform a robustness check for business cycle e↵ects. In two subsamples we investigate how the short-term cyclical dynamics of the Mexican indicators has changed from 1981–1999 to 2000–2016. These subsamples mark what we refer to as the prior/early NAFTA and the late NAFTA periods. Even though NAFTA was signed in 1994 market liberalization was gradual and took approximately 10 years to complete. From 2000 onwards the Mexican banking sector and stock markets were opened to foreign participation, thus if any liberalization e↵ects are to be detected our subsample choice should reveal those.
6.1
Univariate analysis
shrinking the number of parameters across the indicators in a multivariate model can be too restrictive if the univariate results do not support the parameter pooling.
In general, the literature classifies all cyclical components into two groups. Short-term cycles with periodicity from two to eight years describe a business cycle whereas components from eight to 30 years describe a financial cycle (see e.g., Drehmann, Bo-rio and Tstasaronis, 2012). For each indicator we estimate a univariate model which has two cyclical components with their own frequency, persistence and state-innovation parameters. These parameters are not restricted to be common across several indica-tors. A priori, we do not impose that univariate models should contain one financial and one business cyclical component. Hence it is possible that both extracted cyclical components correspond either to a financial or a business cycle frequency. The data generating process of an indicator yi,t is given by
yi,t= zi0ct+ ei,t, ei,t ⇠ N(0, e,i2 ) ct = Tict 1+ ⌘i,t, ⌘i,t ⇠ N(0, Qi)
where ct,i = (c0i,1, c0i,2)0 are the trigonometric components written in the recursive form as in the multivariate model Eq.(1)6, i 2 {G, M} is the country index, e
i,tand ⌘i,t, are mutually independent error and state innovation terms, zi = [1, 1]⌦ e01 contains factor loadings, Ti = blkdiag[Rc1,i, Rc2,i] is the state transition matrix with rotation matrices
specified as in Eq.(4), Qi = diag[ c21,i,
2
c2,i]⌦ I2 is the state innovation matrix.
Table 3 shows the univariate estimation results. We are able to estimate two cyclical components and noise for all indicators. Since SENTG,tis the first principal component of multiple financial market series it does not contain noise by construction. With 6To refer to the cyclical components, here we use a di↵erent notation from the one used in Eq.(1).
our data we estimate that the financial cycle is 16- 17 years long and the business cycle lasts 2-5 years. We find that US and Mexican GDP growth rates and Mexican stock price index returns only contain business cycle frequency components whereas US household leverage only contains financial cycle frequency components. These findings are incorporated in the specification of the factor loading matrices shown in Section 4. Table 3: Business and financial cycle components found in univariate model estimations.
Indicators cycle 1 cycle 2 noise
ˆ1 ˆ1 P1 ˆ2 1 ˆ2 ˆ2 P2 ˆ22 ˆ✏2 log L SENTG,t 0.887c 0.446c 4 0.028b 0.923c 0.095c 17 0.030c 0.000 -6 (0.046) (0.070) (0.015) (0.037) (0.052) (0.015) (0.000) HHLEVG,t 0.773c 2.859c 1 0.120 0.984 0.090 17 0.013 0.354c -168 (0.110) (0.111) (0.115) (0.024) (0.015) (0.014) (0.147) GDPG,t 0.996c 0.425c 4 0.002c 0.763c 0.346b 5 0.156c 0.347c -168 (0.021) (0.054) (0.010) (0.110) (0.108) (0.076) (0.071) HHLEVM,t 0.937c 0.352c 5 0.030c 0.983c 0.099c 16 0.011a 0.293 -147 (0.031) (0.037) (0.014) (0.014) (0.014) (0.007) (0.041) NFLEVM,t 0.820c 0.501b 3 0.048a 0.984c 0.092c 17 0.007a 0.322c -151 (0.096) (0.110) (0.030) (0.014) (0.014) (0.005) (0.052) SPM,t 0.979c 0.774c 2 0.003 0.768c 0.414c 4 0.087 0.289c -151 (0.021) (0.020) (0.004) (0.104) (0.108) (0.043) (0.051) GDPM,t 0.946c 0.862c 2 0.010 0.755c 0.334c 5 0.0383 0.326c -148 (0.041) (0.040) (0.0089) (0.216) (0.199) (0.0449) (0.0565) Notes: a p < 0.10,b p < 0.05, c p < 0.01;
i is a frequency parameter, i is a persistence parameter,
ˆ2
✏ is the idiosyncratic variance, ˆ2i, i = 1, 2 is a variance of the cyclical innovations, Pi, i = 1, 2 is a
cycle period in years calculated as Pi= (2⇡/ i)/4.
6.2
Multivariate analysis
Both US household leverage growth rates and investor sentiment contain financial cycle frequency components. Even though univariate analysis indicated that their cyclical components are similar, their e↵ects on Mexican indicators may have di↵erent e↵ects. Along the two measures for the US financial cycle we consider two business cycle com-ponents. Since HHLEVG,t does not contain business cycle frequencies, we analyze its financial cycle e↵ects in combination with a business cycle estimate obtained from US GDP growth rates. Alternatively, along a financial cycle we estimate a business cycle component from SENTG,t. Table 4 shows estimated e↵ects of US Household and in-vestor sentiment on Mexican indicators. Before estimations, we impose restrictions on the factor loading matrices as discussed in Section 4 and Section 6.1.
Table 4: US and Mexico’s financial and real integration.
Household sentiment investor sentiment
yt G,t G,t M,t M,t
2
✏,i G,t G,t M,t M,t 2✏,i
Factor loadings [ZG, ZM] H Factor loadings [ZG, ZM] H
FIG,t 1 0 0 0 0.559c 1 1 0 0 0 GDPG,t 0 1 0 0 0.520c NA NA NA NA NA HHLEVM,t 0.511 -0.033 1 -1.212c 0.274c -0.200 0.079 1 -1.264b 0.267c (0.440) (0.187) (0.309) (0.200) (0.255) (0.575) NFLEVM,t 0.161 -0.293b 0.769c -0.884c 0.364c -0.339b 0.696b 0.587c -1.224b 0.313c (0.342) (0.159) (0.090) (0.249) (0.152) (0.301) (0.096) (0.595) SPM,t 0 0.781c 0 1 0.304a 0 -0.041 0 3.217c 0.257c (0.214) (0.398) (1.085) GDPM,t 0 0.335c 0 -0.04 0.492c 0 0.250 0 1 0.494c (0.132) (0.255) (0.247) i 0.085c 0.338c 0.085c 0.338c 0.105c 0.481c 0.105c 0.481c Pi 18 5 18 5 15 3 15 3 i 0.992c 0.863c 0.992c 0.863c 0.939c 0.876c 0.876c 0.876c Q 0.010c 0.132c 0.008c 0.050c 0.081c 0.036c 0.082c 0.014c log L -941 -659
Notes: a p < 0.10, c p < 0.05, c p < 0.01. The standard errors for the coefficients are reported in
brackets. In the table a factor loading ”1” indicates which observed series were used as a reference for an unobserved cyclical component. Zero factor loadings not reported here are for G,t⇤ , G,t⇤ , M,t⇤ ,
Figure 2: US and MX financial and business cycles.
Household sentiment
(a) US long-term cycle ( ˆG,t) (b) US short-term cycle (ˆG,t)
MX long-term cycle ( ˆM,t) MX short-term cycle (ˆM,t)
investor sentiment
(c) US long-term cycle ( ˆG,t) (d) US short-term cycle (ˆG,t)
We find no statistically significant link between Mexican household leverage and US financial or business cycle components. One possible explanation is that global forces have not changed Mexican labor market conditions. Therefore, household borrowing still hinges on unstable income and employment history. According to the Federal Reserve, over 1980–1999 the median Mexican full-time salary in real terms was 320 USD per week and only 340 USD calculated over 2000–2018. OECD estimates that almost six in ten workers are employed informally in Mexico and are not covered by social security (see e.g., OECD Employment outlook 2017). Thus changes in income have been relatively modest, limiting the growth of household borrowing capacity.
Several interesting results describe the dynamics of Mexican non-financial corpo-ration leverage growth rates. First, the US business cycle, extracted from US GDP growth rates, is countercyclical to Mexican non-financial leverage growth rates. The origins of this observation are not immediately clear and several theories could explain this. One possible technical explanation is that the US business cycle leads Mexican non-financial leverage dynamics. A non-technical explanation relates to Fern´andez and Gulan (2015) who suggest that a positive productivity shock does not only increases output, but also increases the net worth of entrepreneurs, thereby reducing leverage. Following a positive productivity shock, an initially leveraged entrepreneur will expe-rience high profits, increase equity by more than debt and therefore deleverage. This implies that leverage and income move in opposite directions.
than improving real outcomes of firms in emerging markets. Thus quantitative easing increases more the supply of credit to borrowers with higher ex-ante loan rates and with substantially higher ex-post loan defaults.
Finally, Table 4 shows that the US business cycle extracted from US GDP growth rates is positively linked to Mexican stock price index and GDP growth rates. Since many Mexican companies are oriented towards the US consumer market this finding is not surprising. A higher US GDP implies a larger global demand and especially so for Mexico which in 2017 was the third largest US trade partner after China and Canada. In order to compare the relative importance of each cyclical component we decom-pose variances of Mexican indicators into US and Mexican financial and business cycle, and noise part components. For orthogonal cyclical components the variance decom-position is given by Eq.(5) with the input parameter values as shown in Table 4.
Table 5: Decomposition of variance for the Mexican indicators (%).
HHLEVG,t SENTG,t yM,t G,t G,t M,t M,t ✏i,t G,t G,t M,t M,t ✏i,t HHLEVM,t 14 0 39 24 23 3 0 63 9 25 NFLEVM,t 2 5 32 18 43 10 9 30 11 40 GDPM,t 0 11 0 0 89 0 2 0 11 88 SPM,t 0 39 0 24 37 0 0 0 71 29
Notes: Cyclical decomposition into Mexican and US business and financial cycle components is ex-pressed in percentages and cyclical components are constructed using the results of Table 4.
To sum up, while Mexican GDP and stock returns are driven mostly by the US business cycle, Mexican household leverage in short and long-term seems to follow mostly local dynamics. These results do not exclude the possibility that there exists an unaccounted US factor which guides the long-term dynamics of Mexican household leverage growth rates. In particular, one potential candidate indicator is VIX (volatility index). Under a closer inspection we find that it contains only 10 year and four year cyclical components hence these are not similar to Mexican long-term cyclical compo-nents. Another candidate indicator which we examine is US slope of the yield curve. This indicator contains eight year cyclical component which does not match either the Mexican financial cycle frequency parameter. Figure 2 shows the smoothed cyclical components.
6.3
Robustness check
As discussed by Kose, Meredith and Towe (2004), there are reasons to believe that the North American Free Trade Agreement (NAFTA) had an e↵ect on trade and financial flows between Mexico and US.7One may expect that in this period the role of US factors increased for the Mexican economy. However, it is still difficult to isolate the NAFTA e↵ects since part of these were anticipated before the official agreement ratification as early as in 1991. In addition, NAFTA lead only to a gradual market liberalization, that is, elimination of tari↵s and trade barriers took place over the first ten to fifteen years after the agreement was signed. Following the signing of NAFTA, Mexico and US signed also other bi-lateral trade agreements with several countries, thereby increasing their global exposure in general.
7For example trade barriers decreased, Mexico’s exports shifted towards manufactured good,
Keeping in mind that the late NAFTA subsample analysis provides only an indirect evidence on the trade liberalization outcomes, it is crucial to repeat the analysis of Section 6.2 for 1981:Q1–1999:Q4 and 2000:Q1–2016:Q1 subsamples. Since NAFTA was implemented only gradually, 2000 defines a good threshold year for a subsample analysis. By 2000 Mexico had opened up its banking sector and stock market to foreign participation, it had begun to issue long-term treasury bonds and had stabilized its inflation level. Hence it seems reasonable to assume that if there are any e↵ects of the integration to be found these should become visible from year 2000 onwards. From our sub-sample analysis, we omit indicators with financial cycle components. The number of observations in subsamples is not sufficient to make any reliable inference for financial cycles.
Table 6 shows the subsample di↵erences in cyclical components for Mexican stock price index and GDP growth rates. These two indicators contain business cycle and idiosyncratic components only. The idiosyncratic components of both indicators have reduced in the late NAFTA subsample.8 This finding is not surprising given that monetary stability with stable inflation and flexible exchange rate was achieved in the late 90s. From 1999 a fully fledged inflation target became the focal point of Mexican monetary policy; Mexico issued its local currency bonds and had reduced the external debt (see e.g., Carstens and J´acome, 2005). All of this helped to shield Mexico from exchange rate shocks. At the same time, in the late NAFTA period business cycle components of Mexican GDP growth rates changed its local for US origins. Mexican stock price index returns have contained the US business cycle in both subsample periods.
Following the same procedure as for the full sample results, we calculate the relative proportions of cyclical variances explained by US and Mexican business cycles. The 8Notice that this e↵ect is even stronger since we have removed sudden crisis episodes which happen
Table 6: Subsample di↵erences in real and financial integration.
1981–1999 2000–2016 1981–2016
yt G,t M,t
2
✏,i G,t M,t ✏,i2 G,t M,t 2✏,i
Factor loadings H Factor loadings H Factor loadings H
GDPG,t 1 0 0.461c 1 0 0.452c 1 0 0.427c SPM,t 0.597b 1 0.367c 0.474c 1 0.148c 0.535c 1 0.258c (0.278) (0.139) (0.138) GDPM,t -0.054 0.303a 0.618c 0.669c -0.606 0.198c 0.218b 0.227 0.506c (0.163) (0.187) (0.164) (-0.522) (0.166) (1.206) i 0.344c 0.344c 0.532c 0.523c 0.316 0.316 Pi 5 5 3 3 5 5 i 0.809c 0.809c 0.774c 0.774c 0.777c 0.777c Q 0.194c 0.187c 0.179c 0.010c 0.211 0.211c 0.124c log L -288 -175 -488
Notes: a p < 0.10, c p < 0.05, c p < 0.01. The standard errors for the coefficients are reported in
brackets. In the table a factor loading ”1” indicates which observed series were used as a reference for an unobserved cyclical component. Zero factor loadings not reported here are for ⇤
G,t , G,t⇤ , M,t⇤ , ⇤
M,t.
results are shown in Table 7. We find that both Mexican GDP and stock price index growth rates have increased their exposure to the US business cycle. These e↵ects may be attributed to NAFTA however, Mexican and US global exposure in general has increased through other bilateral trade agreements.
Table 7: Decomposition of variance for the Mexican indicators over the subsamples (%).
1981–1999 2000–2016
yM,t G,t M,t ✏i,t G,t M,t ✏i,t
SPM,t 18 49 33 37 9 54
GDPM,t 0 7 92 49 2 49
7
Conclusion
This paper investigates the cyclical properties of Mexican economic and financial ac-tivity indicators. The indicators are decomposed into country specific and US business and financial cycles to investigate the real and financial integration between US and Mexico. Historically, Mexico has had tight trade ties with US and since 2000 it has opened up its equity and banking sectors. Thus there are many channels through which foreign financial and business cycles can be transmitted. In this context, understanding the scope of the domestic versus foreign dynamics is important.
idiosyn-cratic component for Mexican GDP and stock price index has decreased in the late NAFTA subsample. This accords with the advent of monetary stability from 2000 on-wards when exchange rate risks were tackled through reduction of external debt, bonds denominated in local currency and the central bank’s commitment to low inflation target.
References
Arora, V. and M. Cerisola (2001), “How does U.S. monetary policy influence sovereign spreads in Emerging markets?”, Technical report, IMF, Washington D.C.
Arora, V. and A. Vamvakidis (2005), “How much do trading partners matter for eco-nomic growth?”, Technical Report 52, IMF, Washington D.C.
Baker, M. and J. Wurgler (2006), “Investor sentiment and the cross-section of stock returns”, Journal of Finance, 61, 1645–1680.
Baxter, M. and R.G. King (1999), “Measuring business cycles: Approximate Band-pass filters for economic time series”, The Review of Economics, 81, 575–593.
Borio, C. (2012), “The financial cycle and macroeconomics: What have we learnt?”, Technical Report 395, BIS, Basel.
Bry, G. and C. Boschen (1971), Cyclical analysis of time series: Selected procedures and computer programs, National Bureau of Economic Research, New York.
Canova, F. (2005), “The transmission of US shocks to Latin America”, Journal of Applied Econometrics, 20, 229–251.
Carstens, A. and L. H. J´acome (2005), “Latin American Central bank reform: Progress and challenges”, Technical Report 05/114, IMF, Washington D.C.
Christiano, L.J. and T.J. Fitzgerald (2003), “The Band pass filter”, International Eco-nomic Review, 44, 435–465.
Clark, J. and E. Berko (1997), “Foreign investment fluctuations and Emerging market stock returns: The case of Mexico”, Technical Report 24, Federal Reserve Bank of New York, New York.
Drehmann, M., C. Borio, and K. Tstasaronis (2012), “Characterising the financial cycle: Dont lose sight of the medium term!”, Working Papers 380, BIS, Basel.
Durbin, J. and S.J. Koopman (2001), “Time Series Analysis by State Space methods”, Oxford University Press.
Fern´andez, A. and A. Gulan (2015), “Interest rates, leverage, and business cyclces in Emerging economies: The role of financial frictions”, American Economic Journal: Macroeconomics, 7, 153–188.
Galati, G., S.J. Koopman, I. Hindrayanto, and Vlekke M. (2016), “Measuring financial cycles with a model-based filter: Empirical evidence for the United States and the Euro area”, Technical Report 495, Dutch National Bank, Amsterdam.
Harding, D. and A. R. Pagan (2002), “Dissecting the cycle: A methodological investi-gation”, Journal of Monetary Economics, 49, 365–381.
Harvey, A.C. (1989), Forecasting, Structural Time Series Models and the Kalman Filter, Cambridge University Press, Cambridge.
Harvey, A.C. and A. Jaeger (1993), “Detrending, stylized facts and the business cycle”, Journal of Applied Econometrics, 8, 231–247.
Harvey, A.C. and S.J. Koopman (1997), System dynamics in economic and financial models, Wiley, chapter Multivariate Structural time series models, 269–298.
Igan, D., A. N. Kabundi, F. N.-D. Simone, M. Pinheiro, and N.T. Tamirisa (2009), “Three cycles: Housing, credit, and real activity”, Working Paper 09/231, Interna-tional Monetary Fund, Washington, D.C.
Koopman, S.J., R. Lit, and A. Lucas (2016), “Model-based business cycle and finan-cial cycle decomposition for Europe and the U.S.”, Discussion Paper 16-051/IV, Tinbergen Institute, Amsterdam.
Kose, A., M.G. Meredith, and C.M. Towe (2004), “How has NAFTA a↵ected the Mexi-can economy? review and evidence”, Technical Report WP/04/59, IMF, Washington D.C.
Kose, A., C. Otrok, and C.H. Whiteman (2003), “International business cycles: World, region and country-specific factors”, American Economic Review, 93, 1216–1239. L´opez-C´ordova, E., G.E. Hern´andez, and A. Monge-Naranjo (2003), “NAFTA and
manufacturing productivity in Mexico”, Economia, 4, 55–98.
Mackowiak, B. (2007), “External shocks, US monetary policy and macroeconomic fluc-tuations in emerging markets”, Journal of Monetary Economics, 54, 2515–2520. Maudos, J. and L. Solis (2011), “Deregulation, liberalization and consolidation of the
Mexican banking system: E↵ects on competition”, Journal of International Money and Finance, 30, 337–353.
Morais, B., J.L. Peydro, and C. Ruiz (2015), “The international bank lending channel of monetary policy rates and QE: Credit supply, reach-for-yield, and real e↵ects”, International finance Discussion Papers 1137, Board of Governors of the Federal Reserve System, Washington DC.
OECD (2017), OECD Employment Outlook 2017.
Rey, H. (2015), “Dilemma not trilemma: the global financial cycle and monetary policy independence”, Technical Report 21162, NBER, Cambridge, MA.
Sosa, S. (2008), “External shocks and business cycle fluctuations in Mexico: How im-portant are U.S. factors?”, Technical Report WP/08/100, IMF, Washington D.C. Swiston, A. and T. Bayoumi (2008), “Spillovers across NAFTA”, Technical Report
WP/08/3, IMF, Washington.
Upper, C. (2017), “Macroprudential frameworks, implementation and relationship with other policies-overview”, BIS Paper 94, BIS, Basel.
A
UNIT ROOT TESTS
Table 8: Dickey Fuller unit root test
yi,t probability t-stats lags
SENTG,t 0.007 -3.584 1 HHLEVG,t 0.051 -2.873 2 GDPG,t 0.000 -4.707 1 HHLEVM,t 0.051 -2.873 2 NFLEVM,t 0.002 -4.002 1 GDPM,t 0.000 -8.701 1 SPM,t 0.000 -7.100 1
Notes: yi,tis a financial indicator where i = 1, . . . , 7. The lag order was detected using AIC information
