II.1. PVAR model

We assumed that the economy i (i = 1, 2,…, N) is described by the following equation:

(1)

where G(L) and C(L) are polynomial matrices in the lag of L; yti is a vector of M × 1 data of endogenous variables for country i at time t; x_t is a vector of K × 1 data of exogenous or global variables; dⁱ is a constant matrix M × 1; M and K represent the endogenous and exogenous variables in the model, respectively; and eti expresses a vector of structural disturbances. Considering that these disturbances are mutually unrelated, var ( eti ) is a diagonal matrix in which diagonal elements are the variance of structural disturbances. The fixed individual effect, dⁱ , was introduced to control country-specific factors that are not taken into account in the model (^{Kim & Mehrotra, 2018}).

According to ^{Kim and Mehrotra (2018)}, a reduced form of the PVAR model was estimated with individual fixed effects, represented by the following equation:

(2)

where cⁱ is a constant vector M × 1; B(L) and D(L) are polynomial matrices in lag L; and uⁱ is a vector M × 1 of reduced-form residuals and var (uⁱ )=Ʃt.

As the variance-covariance matrix of errors was not diagonal, and the errors of the reduced PVAR model presented contemporary relationships with each other, it was necessary to decompose the errors so that they became orthogonal. Therefore, Cholesky identification was introduced adopting a particular ordering of the variables of the system. In this case, the ordering took the variables that initially appeared in the system as weakly exogenous, while those that were arranged in the sequence were estimated as more endogenous (^{Vinhado & Divino, 2015}).

All data were stable. Granger causality was also tested. In the construction of probability bands for impulse responses, we applied the Monte Carlo integration method. In addition, we verified the ^{Dickey-Fuller} unit root test (1981), the ADF-GLS test elaborated by ^{Elliott, Rothenberg, and Stock (1996)}, and the ^{Phillips-Perron} test (1988). For this work, we used two and three lags considering the Akaike parameter for identification.

II.2. Data description

The following charts show the behavior of the variables real GDP, inflation rate, credit to the private sector, interest rate, and the implementation of macroprudential policies, respectively.

Source: Own elaboration based on the data obtained from ^{IMF (2020)} and ^{CEIC (2020)}.

Chart 1. Performance of real GDP (in percentage) for the sample of countries in the ITR and non-ITR groups between the first quarter of 2007 and the fourth quarter of 2019 

As can be seen in Chart 1, all the countries studied showed a weak economic performance as of the third quarter of 2008, which reflects the deleterious impact of the GFC. This fact is a consequence of the strong link between the Latin American economies and the United States, the epicenter of the crisis. Argentina, given its various internal crises, had other moments of negative real GDP, specifically in 2012, 2014, 2016, and between 2018 and 2019. Brazil, with the political and economic crisis of 2015, also experienced negative percentages of real GDP until the beginning of 2017.

Source: Own elaboration based on data obtained from ^{IMF (2020)} and the ^{central banks of Peru,} Argentina, and ^{Uruguay (2020)}.

Chart 2. Performance of inflation rate for the sample of countries in the ITR and non-ITR groups between the first quarter of 2007 and the fourth quarter of 2019Note: Non-ITR countries are represented on the secondary axis. 

Source: Own elaboration based on data obtained from ^{BIS (2020)} and the ^{central banks of Peru}, ^{Costa Rica}, ^Ecuador, ^Bolivia, and ^{Uruguay (2020)}.

Chart 3. Performance of credit to the private sector (in millions of dollars) for the sample of countries in the ITR and non-ITR groups between the first quarter of 2007 and the fourth quarter of 2019Note: Non-ITR countries are represented on the secondary axis. 

Chart 2 shows the fluctuations in the inflation rate in the sample of countries. Chile presented negative results for this variable between 2009 and 2010. Argentina was the most volatile economy, with an inflation rate of 54% in the last quarter of 2019, which reflects the crisis suffered by this country.

As can be seen in Chart 3, Brazil and Mexico stand out in relation to credit to the private sector. This fact evidences that these economies may have a more developed financial system compared to the other countries, in addition to being the largest ones in the sample. Moreover, according to ^{Nuguer and Garcia (2016)} and ^{Alam et al. (2019)}, Brazil and Mexico have a history of using prudential measures since the 1990s, favoring the resilience of the financial system and making lending more solid.

Source: Own elaboration based on data obtained from ^{IFS (2020)}, ^{CEIC (2020)}, and the ^{central banks of Peru} and ^{Uruguay (2020)}.

Chart 4. Performance of interest rate for the sample of countries in the ITR and non-ITR groups between the first quarter of 2007 and the fourth quarter of 2019 

Chart 4 shows that the ITR countries appear to have followed the main principle of the Taylor rule. During the period from 2007 to 2019, the interest rate and the price levels (Chart 2) were synchronized. This can indicate the functionality of this monetary arrangement in which the interest rate reacts to fluctuations in inflation to control it at levels close to its pre-established goals.

Source: Own elaboration based on the data obtained from ^{Alam et al. (2019)}, ^{Cerutti, Claessens, and Laeven (2017)}, and the documents and databases of the respective central banks of the countries examined.

Chart 5. Implementation of macroprudential policies in the sample of countries between 2007 and 2019 

Finally, Chart 5 presents the dynamics of the implementation and maintenance of macroprudential policies in the Latin American countries studied. As can be seen, there is a concentration of macroprudential measures in the period from 2014 to 2016.

III.1. Results for the ITR group of countries

As explained in part (a) of Figure 1 (Figure 1(a)), there is evidence of sensitivity of real GDP of the sample of countries to an increase in interest rates. The results show that product decreases once contractionary monetary policies are implemented and grows only after the eighth quarter. Thus, real GDP reacts negatively to monetary policies in the long term, which indicates the strong influence of this type of policy.

After all, considering the transmission channels of monetary policies, when interest rates are rising, the reaction observed is the fall of investments in the productive sector, since the higher the interest rates, the greater the borrowing costs. Thus, companies avoid taking credit and, consequently, their levels of investment and consumption decrease (households behave similarly).

Figure 1(b) shows that the effects of macroprudential policies on GDP are analogous those of monetary policies. However, macroprudential measures do not have such an intense impact on real GDP. This is expected because this type of policy does not focus on economic activity, but rather on the conditions of the banking market, as highlighted by ^{Smaghi (2014)} and ^{Yellen (2014)}.

Despite the more subtle result on real GDP, macroprudential policies end up indirectly influencing economic activity. This indicates the importance of considering the effects of this type of policy on the real economy before or together with the implementation of monetary policies, promoting possible adjustments in their management.

Figure 1(c) shows the response of credit to the private sector in relation to the shock in the interest rate. There is a fall until the eighth quarter (-2%) and, after that, this variable starts to have positive results, reaching 6% growth at the end of the third year. The possible explanation for this increase may be related to the fact that, when interest rates rise, controlling inflation (considering that these are countries that follow the ITR assumptions), the expectations of agents improve in relation to the future, which makes them seek more credit in the following periods, thus contributing to the advancement of this variable in the long term.

Source: Own elaboration.

Figure 1. Impulse response of the variables GDP, CTPS, CPI, IR, and PP in the PVAR model for Brazil, Chile, Colombia, Mexico, Peru, and UruguayNote: PP = dummy variable for the policy; CPI = percentage change in the price level; GDP = percentage change in real GDP; CTPS = credit to the private sector; and IR = interest rate. 

As can be seen in the results of Figure 1(d), credit to the private sector presents a sharp fall of 14% in the third quarter due to restrictive prudential measures, and these effects are prolonged in the long term, stabilizing at -13% until the end of the fourth year. In other words, there is strong evidence that macroprudential policies fulfill their role of reducing credit in the ITR group. The impacts of these policies on credit exposed here are comparable to those observed in ^{Zdzienicka et al. (2015)} and ^{Cerutti, Claessens, and Laeven (2017)}.

Inflation, in turn, decreases in response to interest rates, as shown in Figure 1(e), according to the ITR logic. The increase in inflation observed in the short term is explained by the slow absorption process in relation to the interest rate and by inflation targeting regimes being structured by managers focused on future inflation expectations (that is, forward-looking systems).

Macroprudential actions had a similar impact on inflation, as evidenced in Figure 1(f). There is an initial rise in inflation, which loses strength over the quarters. The increase may be because the central bank takes some time to apply both macroprudential and monetary policies, heating the economy and the financial market. This result underlines that macroprudential policies should not aim to reduce inflation, thus reaffirming ^{Smaghi (2014)} and ^{Yellen (2014)}.

These inferences are reinforced in Table 1, which presents the variance decomposition to observe the relative importance of policy tools in explaining the volatility of the variables.

Table 1. Variance decomposition for shocks in macroprudential and monetary policies 

^Note:PP represents the dummy variable for macroprudential policies. IR symbolizes the shock of monetary policies, that is, the interest rate. On the other hand, GDP is real GDP, CTPS represents credit to the private sector, and CPI is the inflation rate.

^Note*:The standard deviation is shown in parentheses.

Source: Own elaboration.

In line with Figure 1, Table 1 evidences that the interest rate explains 5.66% of the behavior of real GDP in the fourth year, while macroprudential policies have lower explanatory percentages. For credit, these policies handle for most of the fluctuations in credit. On the other hand, monetary policies have a more important role in unraveling price variations.

Next, we present the repercussions of macroprudential and monetary policies on private investment and consumption. As they are the components with the greatest weight in real GDP, it is important to test them to see if these policies influence product through consumption or investment.

III.2. Impacts of monetary and macroprudential policies on consumption and investment

In parts (a) and (c) of Figure 2, we can see how consumption and investment, respectively, respond in a similar way to the increase in the interest rate. Both variables show a sharp drop that continues in the long term. Consumption is reduced by 21% at the end of the first year, and investment reaches -12.2%. This is in line with what economic theory predicts: a positive interest rate movement discourages consumption and investment, which negatively impacts the performance of the economy.

Macroprudential measures are also capable of negatively influencing consumption and investment, possibly by lowering the level of credit, as shown in Figure 2 in parts (b) and (d). This type of policy seems to be able to reduce investments by up to 4.8% in the second year.

Therefore, the results evidence that macroprudential policies can be transmitted to economic activity through credit channels. In general, these policies can reduce credit in the economy, which results in banks limiting the supply of loans contributing to the decrease in consumption and investment. According to Karmakar (2016), this strategy provides structural stability to the financial system, as it controls credit levels and alleviates the volatility arising from the procyclical nature of credit markets.

Ultimately, there is evidence that macroprudential and monetary policies have some similar transmission channels among the variables and can cause comparable expectations and behaviors in economic agents. Since higher interest rates also make loan operations more expensive, the demand for credit decreases and, consequently, that of the productive sector (^{Romer, 2018}). These impacts can affect consumption, investment, inflation rate, and real product, as well as macroprudential policies.

Figure 2(e) shows the interest rate response to an increase in macroprudential action. Both types of policies follow the same direction over time, which indicates that they are aligned in time for this group of countries in the period analyzed. Macroprudential and monetary policies seem to have been implemented in an articulated manner during the period examined here, each seeking its specific purpose, but in the same direction. As ^{Svensson (2012)} pointed out, although being independently controlled, these types of policies should be considered together, taking into account the cumulative effects of one type on the other.

As highlighted by ^{Blanchard, Dell’Ariccia, and Mauro (2014)}, regulators should pay attention to the interactions between monetary and macroprudential policies. After all, the former can contribute to the increase in demand for credit, which in turn distances asset prices from their fundamental values, creating bubbles and an environment of growing systemic risk. If there is a shock or a bubble burst, financial institutions start to have problems of capital and liquidity, becoming credit deficient and causing insolvency of economic agents. Highly leveraged financial and non-financial institutions exacerbate the problem, with severe consequences for economic activity. On the other hand, macroprudential policies impact aggregate demand through their effects on credit. Once again, the importance of these two types of policies being coordinated and thought together is highlighted. These recommendations seem to be followed by the countries of the ITR group.

In short, in the ITR group, we can observe that macroprudential and monetary policies have similar effects on the variables studied. Both lead to a decrease in real GDP, inflation rate, private credit, consumption, and private investment. In addition, each type of policy has a greater impact on the economic indicators that are directly under its responsibility.

Next, we present the results obtained with the PVAR model in the non-ITR group of countries in order to prepare a comparative analysis with those that adopt the ITR.

Source: Own elaboration.

Figure 2. Impulse response of the variables _CONSUM, _INVEST, IR, and PP in the PVAR model for Brazil, Chile, Colombia, Mexico, Peru, and UruguayNote: PP= dummy variable for the policy; _CONSUM = percentage of total private consumption; _INVEST= percentage of total private investment; and IR= interest rate.Note*: The bands are wide due to the loss of degrees of freedom with the addition of more variables in the model. 

III.3. Results for the non-ITR group of countries

Part (a) of Figure 3 shows the expected result of real GDP in relation to the increase in the interest rate. It indicates that this variable is negatively affected in view of changes in monetary policies. This impact is in line with economic logic, as the interest rate reduces real GDP in the short term.

In Figure 3(b), real GDP presents effects opposite to what was expected in relation to the more restrictive macroprudential measures since there is an increase in real product from the second to the seventh quarter. These results differ from those found for ITR countries, which suggests that the impacts of macroprudential policies are distorted in non-ITR Latin American countries.

In relation to the other variables, pointing, first, to the effects produced on the inflation rate, we can conclude that this variable has a random behavior considering changes in the interest rate (Figure 3(e)). This is because there is no correct calibration of the real interest rate, as predicted by, for example, the Taylor rule. Monetary policies are not being implemented in a framework in which the rising nominal interest rate implies an increase in the real one (as in economies that adopt the ITR and that have the Taylor rule, or any other calibrated actual interest rate rule).

Regarding credit to the private sector, the interest rate initially causes growth of up to 1% in the second quarter and this increase continues until the end of the third year, as shown in Figure 4(c). Possibly, when new interest rates are announced, agents advance their loans and credit conditions in the present period with the regulations in force (^{Ivashina & Scharfstein, 2010}). In this situation, they know that the credit conditions will have a high cost in the future and, thus, avoid indebtedness at higher rates, which increases credit in the short term using the limits currently available and contracted. In addition, it takes some time for the economy to absorb the impacts of higher interest rates.

After the twelfth quarter, credit showed a sharp drop due to changes in monetary policies. With higher interest rates, these economies are not able to control inflation, as previously explained. Thus, credit is reduced through the channel of expectations, as creditors consider inflation to be a negative attribute since the ability of debtors to generate cash deteriorates and their payment capacity decreases. In addition, higher levels of inflation may indicate greater volatility and, therefore, inflation uncertainty. This increases the chances that inflation will exceed the levels stipulated in the credit agreements and result in the transfer of wealth from creditors to debtors. All this set of factors can encourage banks to reduce their credit supply in the long term. Once again, the impact of monetary policies on these economies is evident, which can also be seen in Table 2.

Source: Own elaboration.

Figure 3. Impulse response of the variables GDP, CTPS, CPI, IR, and PP in the PVAR model for Argentina, Bolivia, Costa Rica, and EcuadorNote: PP = dummy variable for the policy; CPI = percentage change in the price level; GDP = percentage change in real GDP; CTPS = credit to the private sector; and IR = interest rate. 

Table 2. Variance decomposition for shocks in macroprudential and monetary policies 

^Note:PP represents the dummy variable for macroprudential policies of capital requirements. IR symbolizes the shock of monetary policies, that is, the interest rate. On the other hand, GDP is real GDP, CTPS represents credit to the private non-financial sector, and CPI is the inflation rate.

^Note*:The standard deviation is shown in parentheses.

Source: Own elaboration.

Table 2 indicates that macroprudential policies have minimal explanatory factors in the variations in real GDP, credit, and inflation rate. The interest rate, on the other hand, justifies the fluctuations in these variables, including credit in all time horizons. Thus, we highlight, again, the small influence of macroprudential measures on the variables in view of the impacts of monetary policies. In other words, macroprudential policies do not seem able to achieve their main objective of impacting credit in the non-ITR group, while the interest rate causes immense losses in these economies, such as the decrease in credit and the rise in inflation in the long term, in addition to affecting product in the short term.

This is reinforced by what is observed in Figure 4(d). Credit increases in response to restrictive macroprudential measures, reaching an objective opposite to that of these policies. This result can be explained by the economic environment of the countries studied, which shows that they do not use macroprudential tools with the same efficiency as the ITR countries.

For example, Costa Rica presented credit growth after applying a tightening macroprudential measure in 2013 and 2015. According to the ^{Central Bank} of this country (2015), this fact is explained by the increase in the demand for foreign currency loans in 2015, as this period was characterized by the exchange rate stability and low costs. Moreover, since the macroprudential tool applied did not aim to reach foreign currency loans, it allowed the rise in this type of credit based on external resources, which resulted in the country’s total credit growth.

An analogous effect occurred in the Bolivian economy in 2012. According to its ^{Central Bank (2013)}, credit grew even with implementing restrictive macroprudential policies, motivated, mainly, by credit directed to mostly small companies. Moreover, the Central Bank of Bolivia indicated that this was the result of confidence in the country’s business sector and the opening of new companies in that year. In other words, the borrowing agents did not see a constraint of macroprudential policies in an environment of recovery from the financial crisis, inflation volatility, and increase in the interest rate as an indicator of alert and uncertainty about the future, which would make them more cautious aand less demanding of credit in this period.

Similarly, these impacts may be affected by the fact that some of these countries do not have macroprudential policies under the full regulation of the central bank. In Ecuador, for instance, these policies are the responsibility of both its Central Bank and the Financial and Monetary Policy Regulatory Council. In Bolivia, on the other hand, they are coordinated only by the Financial Stability Council. Consequently, when a policy is not monitored by the same agency, it may not be aligned with the monetary policies applied by the central bank, which may result in opposite and overlapping effects on the economy and the cancellation of the impacts of the policies. This is reinforced by Figure 4(d), which shows how monetary policies do not cause reactions to changes in macroprudential ones.

The fact that macroprudential policies do not reduce credit in non-ITR economies also explains the effect on consumption and investment. Figure 4 indicates that tightening prudential measures generate an increase in these variables. These findings are in line with the effect shown by this type of policy on real GDP. They may be occurring because of poorly calibrated or structured macroprudential policies, such as the case of Costa Rica, as mentioned above.

The results of consumption (Figure 4(a)) and investment (Figure 4(c)) in response to monetary policies are in accordance with the effects of real GDP (Figure 3(a)). Both variables are reduced when the interest rate increases.

Source: Own elaboration.

Figure 4. Impulse response of the variables _CONSUM, _INVEST, IR, and PP in the PVAR model for Argentina, Bolivia, Costa Rica, and EcuadorNote: PP= dummy variable for the policy; _CONSUM = percentage of total private consumption; _INVEST= percentage of total private investment; and IR= interest rate.Note*: The bands are wide due to the loss of degrees of freedom with the addition of more variables in the model.