Abstract
Selected Issues
Interest Rates and Inflation in Brazil1
The conventional view among economists is that higher interest rates reduce inflation. However, the prolonged period of low inflation and low interest rates in advanced economies following the global financial crisis appears to be inconsistent with this view. This sparked a debate: Do lower interest rates increase inflation (the conventional view), or do they lead to lower inflation (the so-called Neo-Fisherian view)? This chapter finds strong evidence in favor of the conventional view of monetary policy transmission in Brazil. While lower inflation and lower nominal interest rates can be achieved over the long term by targeting a lower level of inflation, this is likely to come at the cost of lower output (and employment) in the short term—a cost that can be mitigated by enhancing monetary policy transparency and credibility. Monetary policy transmission could be made more efficient by reducing distortions and improving the allocation of resources in the financial sector.
A. Introduction
1. The conventional view among economists is that higher interest rates led to lower inflation. The rationale behind this view is that higher interest rates increase the cost of borrowing and dampen demand across the economy, resulting in excess supply and lower inflation. In this context, there is several channels through which higher interest rates reduce inflation, including the exchange rate channel, the credit channel, and the bank-balance sheet channel (see Mishkin, 1996). Here, a central bank facing the prospect of inflation being higher than its inflation target would raise interest rates enough to increase the real (inflation-adjusted) cost of borrowing, reducing aggregate demand and returning inflation back toward the desired level.
2. There has recently been some debate about whether lower inflation can be achieved by setting lower policy interest rates, the so-called Neo-Fisherian effect. At the heart of the debate, is a well-known equation in economics, the Fisher equation, which relates the nominal interest rate Rt to the real interest rate rt and expected inflation Etπt+1 (all annualized):
Taken at face value, and assuming that the real interest rate is fixed in the long run, the equation implies that a lower long-run inflation rate can be achieved by permanently setting the nominal interest rate to a lower level (see Cochrane, 2016). Indeed, proponents of this view often point to the positive relationship between nominal interest rates and inflation seen across many countries as evidence of ‘Neo-Fisherian’ effects (see Figure 1).2
Brazil: Headline Inflation and Policy Rate
Citation: IMF Staff Country Reports 2017, 216; 10.5089/9781484309919.002.A005
Source: Haver Analytics.Brazil: Headline Inflation and Policy Rate
Citation: IMF Staff Country Reports 2017, 216; 10.5089/9781484309919.002.A005
Source: Haver Analytics.Brazil: Headline Inflation and Policy Rate
Citation: IMF Staff Country Reports 2017, 216; 10.5089/9781484309919.002.A005
Source: Haver Analytics.3. Would a commitment to fix the policy interest rate to a lower level eventually lead to lower inflation? To answer this question, an empirical analysis is conducted to assess the impact of changes in Brazil’s policy rate on inflation in Section B. Section C then evaluates a simple model with long-run Neo-Fisherian effects in the context of several countries’ historical experiences in transitioning to lower levels of inflation. Section D concludes with a summary of key findings and policy conclusions.
B. Empirical Analysis
4. The empirical analysis is based on Vector-Auto-Regressions (VARs). The baseline VAR is estimated using monthly data ranging from 2003 to 2016 and contains six variables: monthly headline inflation, the nominal interest rate (SELIC), the output gap, 12-month-ahead inflation expectations, and monthly percent changes of commodity prices and the real-effective exchange rate. Inflation responses to the interest rate are likely to different across different sectors of the economy, so in addition to the baseline VAR, five additional VARs are estimated as a robustness check, where each VAR includes inflation for a different sector of the economy. Overall, the empirical analysis examines the impact of policy interest rate changes on headline inflation, non-regulated-price inflation, regulated-price inflation, service-price inflation, tradable inflation, and non-tradable inflation.3
5. Cross-correlations show that higher inflation leads to higher interest rates and higher interest rates lead to lower inflation, consistent with the conventional view. The estimated cross-correlation function from the baseline VAR is displayed in Figure 2; the results for all inflation rates can be found in Appendix I, Section B. There is a statistically significant positive relationship between past levels of inflation and the interest rate and a statistically significant negative relationship between past levels of the interest rate and inflation. These results broadly reflect the standard view of the monetary policy transmission to inflation. Since inflation tends to lead the policy interest rate, it appears that the central bank has responded to inflation developments over this sample, partly as the result of unanticipated demand and supply shocks (such as food and regulated-price shocks, and exchange rate shocks). The results also suggest a peak correlation between leads and lags of inflation and leads and lags of the interest rate of around 6 months.
Brazil: Correlation: Headline Inflation and Interest Rate
(Correlation and 20th to 80th percentiles)
Citation: IMF Staff Country Reports 2017, 216; 10.5089/9781484309919.002.A005
Brazil: Correlation: Headline Inflation and Interest Rate
(Correlation and 20th to 80th percentiles)
Citation: IMF Staff Country Reports 2017, 216; 10.5089/9781484309919.002.A005
Brazil: Correlation: Headline Inflation and Interest Rate
(Correlation and 20th to 80th percentiles)
Citation: IMF Staff Country Reports 2017, 216; 10.5089/9781484309919.002.A005
6. Structural VARs also support the conventional view that an unexpected cut in the policy interest rate increases inflation in the short term. Responses of headline inflation to a 100 basis point cut in the policy interest are displayed in Figure 3; the results for all other inflation rates can be found in Appendix I, Section B. Here, the uncertainty around the responses relates both to uncertainty about the parameters of the VARs and to the recursive ordering used to identify the monetary policy shock.4 Examining all possible recursive identification schemes allows the analysis to be agnostic about whether an interest rate shock has a contemporaneous or a lagged impact on inflation. The results show that an unexpected cut in the policy interest rate tends to increase inflation over time, with the magnitude of the impact depending on the sector of the economy and the peak impact generally occurs around 9 months after the shock. The short-term impact of a lower interest rate on inflation is less clear cut, with identification schemes that allow for a cut in the interest rate to immediately impact on inflation (within the same month) sometimes suggesting a positive relationship between inflation and interest rate shocks. Overall, however, the results from the structural VARs strongly support the standard view of monetary policy transmission.
Brazil: Headline Inflation after 100bps Cut in Policy Rate
(Percent, annualized, 20th and 80th percentiles)
Citation: IMF Staff Country Reports 2017, 216; 10.5089/9781484309919.002.A005
Source: Fund staff estimates.Brazil: Headline Inflation after 100bps Cut in Policy Rate
(Percent, annualized, 20th and 80th percentiles)
Citation: IMF Staff Country Reports 2017, 216; 10.5089/9781484309919.002.A005
Source: Fund staff estimates.Brazil: Headline Inflation after 100bps Cut in Policy Rate
(Percent, annualized, 20th and 80th percentiles)
Citation: IMF Staff Country Reports 2017, 216; 10.5089/9781484309919.002.A005
Source: Fund staff estimates.C. How can Lower Inflation Be Achieved over the Long Term?
7. While there is little evidence that a lower interest rate leads to lower inflation in Brazil in the short term, the long-run Fisher equation can still help to inform policy advice. The long run Fisher equation is:
where the steady-state nominal interest rate is equal to the steady state real interest rate plus the inflation target. Assuming the long-run neutrality of money (i.e. nominal variables do not affect real variables in the long run), the inflation target determines the steady state nominal interest rate. This relationship can easily be inserted into a simple (and standard) New Keynesian model (see Appendix I, Section C).
8. Model-based simulations show that a lower long-term inflation and nominal interest rates can be achieved by lowering the inflation target, but this is likely to be costly in the short run when the central bank has limited policy credibility. The results show that a reduction in the inflation target reduces both the nominal interest rate and inflation in the long run (Figure 4). If households and firms in the economy have expectations that are either partially forward-looking or entirely backward looking, the transition to the new inflation target requires lower output in order to move inflation expectations to the new target; the real interest rate must rise to reduce demand in the short term. On the other hand, in a purely forward-looking model the central bank is fully credible, and households and firms fully understand the future implications of monetary policy actions and immediately embed this knowledge in their expectations. In this case, the transition of inflation and nominal interest rates to the new target is instantaneous once the target is announced and output is unaffected.
Brazil: Simulated Responses to a Change in the Inflation Target from 4.5 Percent to 2 Percent
Citation: IMF Staff Country Reports 2017, 216; 10.5089/9781484309919.002.A005
Source: Fund staff estimates.Brazil: Simulated Responses to a Change in the Inflation Target from 4.5 Percent to 2 Percent
Citation: IMF Staff Country Reports 2017, 216; 10.5089/9781484309919.002.A005
Source: Fund staff estimates.Brazil: Simulated Responses to a Change in the Inflation Target from 4.5 Percent to 2 Percent
Citation: IMF Staff Country Reports 2017, 216; 10.5089/9781484309919.002.A005
Source: Fund staff estimates.9. Disinflation episodes across countries show that inflation was slow to adjust to lower levels and the transition to lower inflation was costly in terms of output, reflecting unanchored inflation expectations and limited monetary policy credibility prior to disinflation. Figure 5 shows the behavior of inflation, interest rates, and the output gap in the two years prior to the adoption of inflation targeting in the first five countries that formally adopted the practice, in addition to the Volker disinflation episode in the United States, beginning in 1981.5 The behavior of inflation, interest rates, and the output gap follow broadly similar trends across countries. Nominal interest and inflation rates were positively correlated and tended to decline together once the central bank formally adopted inflation targeting; output gaps generally moved into negative territory. These results are qualitatively (and quantitatively) very similar to the simulation results when households’ and firms’ expectations for inflation and output are not assumed to be entirely forward looking. The large output losses during disinflation across these countries likely reflects a high degree of inflation persistence and limited policy credibility prior to the adoption of inflation targeting. Changing the inflation target would be likely be less costly if the central bank had more policy credibility and more anchored inflation expectations prior to the target change.
Brazil: Disinflation Episodes Across Countries
Citation: IMF Staff Country Reports 2017, 216; 10.5089/9781484309919.002.A005
Source: Fund staff estimates.Brazil: Disinflation Episodes Across Countries
Citation: IMF Staff Country Reports 2017, 216; 10.5089/9781484309919.002.A005
Source: Fund staff estimates.Brazil: Disinflation Episodes Across Countries
Citation: IMF Staff Country Reports 2017, 216; 10.5089/9781484309919.002.A005
Source: Fund staff estimates.D. Summary and Discussion
10. There is strong evidence of the conventional view of monetary policy transmission in Brazil, suggesting that a cut in the policy interest rate leads to higher inflation in the short term. Cross correlations show that higher inflation leads to higher nominal interest rates and higher interest rates result in a reduction in inflation. Since inflation tends to lead the policy interest rate the sample examined, it appears that the central bank has responded to inflation developments, partly as the result of unanticipated demand and supply shocks (such as food and regulated-price shocks, and exchange rate shocks). Structural VARs also suggest that an unexpected cut in the policy interest rate leads to a broad-based rise in inflation across the sectors examined, with the peak impact on inflation occurring around 9 months after a monetary policy shock.
11. Model-based simulations and cross-country evidence suggest that lower inflation and lower nominal interest rates can be achieved over the longer term if the central bank commits to a lower inflation target. If households and firms base their output and inflation expectations on the past (even partially), the transition to the new inflation target comes at the cost of lower output in the short term, with larger output losses and more prolonged transition periods occurring when expectations are more backward looking. An examination of disinflation episodes across several countries broadly supports the key findings from model simulations that assume that expectations were (at least) partially backward looking prior to disinflation.
12. While it appears there is no easy way to permanently lower inflation in Brazil, a lower inflation target could be achieved at less cost with enhanced monetary policy transparency and credibility. Enhanced credibility can better anchor inflation expectations, reduce the persistence of inflation, improve the short run tradeoff between inflation and output, and mitigate the cost associated should a lower inflation target be desired over the medium term. As discussed in Domit and others (2016), there are several dimensions along which Brazil’s inflation targeting framework can be improved to enhance transparency and credibility, including increasing the autonomy of the central bank and changing the inflation target from a range that needs to be met at the end of each year to a longer-term point target. The National Monetary Council made a step in this direction in 2015 by narrowing the target range from 4.5 percent +/- 2 percent to 4.5 percent +/- 1.5 percent from 2017.
13. Monetary policy transmission could also be made more efficient by reducing distortions and improving resource allocation in the financial sector. There is general agreement that the effectiveness of monetary policy in Brazil could be improved by changing various credit policies that involve earmarking and credit subsidies. In particular, as already proposed by the authorities, the gap between the subsidized interest rate on long-term lending (the TJLP) and the policy interest rate (the SELIC) could be reduced over time to improve monetary policy transmission. Linking the TJLP more tightly with the SELIC or another market-determined interest rate (such as long-term yields on government inflation-linked debt) would enhance the transmission of SELIC changes to longer-term interest rates. This could increase the potency of a given change in the SELIC and contribute to lower interest rate volatility over the business cycle. Improving the efficiency of resource allocation in the financial sector could also contribute to a lower long-term real interest rate in Brazil, allowing for lower nominal interest rates for a given inflation target.6
Appendix I. Data and Robustness
A. Data
Data, Sources and Transforms
Data, Sources and Transforms
| Series | Source | Transform* |
|---|---|---|
| Headline IPCA | IBGE | Δlog(x)*1200 |
| Non-tradable IPCA | IBGE | Δlog(x)*1200 |
| Tradable IPCA | IBGE | Δlog(x)*1200 |
| Services IPCA | IBGE | Δlog(x)*1200 |
| Non-regulated IPCA | IBGE | Δlog(x)*1200 |
| Regulated IPCA | IBGE | Δlog(x)*1200 |
| Inflation expectations (12-months ahead) | BCB | x |
| Interest rate (SELIC) | BCB | x-hptrend(x) |
| Activity Index (IBC-BR) | BCB | log(x)*100-hptrend(log(x))*100 |
| Commodity Price Index (IC-BR) | BCB | Δlog(x)*1200 |
| Real Effective Exchange Rate (broad) | JP Morgan | Δlog(x)*1200 |
Data, Sources and Transforms
| Series | Source | Transform* |
|---|---|---|
| Headline IPCA | IBGE | Δlog(x)*1200 |
| Non-tradable IPCA | IBGE | Δlog(x)*1200 |
| Tradable IPCA | IBGE | Δlog(x)*1200 |
| Services IPCA | IBGE | Δlog(x)*1200 |
| Non-regulated IPCA | IBGE | Δlog(x)*1200 |
| Regulated IPCA | IBGE | Δlog(x)*1200 |
| Inflation expectations (12-months ahead) | BCB | x |
| Interest rate (SELIC) | BCB | x-hptrend(x) |
| Activity Index (IBC-BR) | BCB | log(x)*100-hptrend(log(x))*100 |
| Commodity Price Index (IC-BR) | BCB | Δlog(x)*1200 |
| Real Effective Exchange Rate (broad) | JP Morgan | Δlog(x)*1200 |
B. Empirical Results
Brazil: Range of Cross-Correlations Between the Interest and Inflation Rates
(Median and 20th and 80th percentiles)
Citation: IMF Staff Country Reports 2017, 216; 10.5089/9781484309919.002.A005
Source: Fund staff estimates.Brazil: Range of Cross-Correlations Between the Interest and Inflation Rates
(Median and 20th and 80th percentiles)
Citation: IMF Staff Country Reports 2017, 216; 10.5089/9781484309919.002.A005
Source: Fund staff estimates.Brazil: Range of Cross-Correlations Between the Interest and Inflation Rates
(Median and 20th and 80th percentiles)
Citation: IMF Staff Country Reports 2017, 216; 10.5089/9781484309919.002.A005
Source: Fund staff estimates.
Brazil: Range of Inflation Responses to 100bps Increase in Interest Rate
(Median and 20th and 80th percentiles)
Citation: IMF Staff Country Reports 2017, 216; 10.5089/9781484309919.002.A005
Source: Fund staff estimates.Brazil: Range of Inflation Responses to 100bps Increase in Interest Rate
(Median and 20th and 80th percentiles)
Citation: IMF Staff Country Reports 2017, 216; 10.5089/9781484309919.002.A005
Source: Fund staff estimates.Brazil: Range of Inflation Responses to 100bps Increase in Interest Rate
(Median and 20th and 80th percentiles)
Citation: IMF Staff Country Reports 2017, 216; 10.5089/9781484309919.002.A005
Source: Fund staff estimates.C. Simple Model
The IS curve relates the current level of the output gap, yt, to the lagged output gap, expectations of the future output gap and the real interest rate (deviation from steady state):
The Phillips curve relates the current level of inflation to inflation expectations, past inflation, and the output gap (where 0≤ α ≤ 1):
The monetary policy rule relates the nominal interest rate to the steady state nominal interest rate and the expected deviation of inflation from the inflation target:
where μ > 1 to ensure a unique and stable solution, and the long-run Fisher equation is:
The parameters values are σ = 1, γ = 0.05, μ = 1.5, and r* = 6, The parameters in the Phillips and IS curves related to persistence are: δ = α = 1, in the forward-looking model; δ = α = 0.5, in the partially forward-looking model; and δ = α = 0, in the backward-looking model.
References
Cochrane, John, 2016, “Michelson-Morley, Occam and Fisher: The Radical Implications of Stable Inflation at Near-Zero Interest Rates?” Working Paper (Hoover Institution).
International Monetary Fund, 2016, “Upgrading Brazil’s Inflation-Targeting Framework”, in Brazil: Selected Issues, IMF Country Report No. 16/349 (Washington).
García-Schmidt, Mariana, and Michael Woodford, 2015, “Are Low Interest Rates Deflationary? A Paradox of Perfect-Foresight Analysis”, NBER Working Paper No. 21614 (Cambridge, Massachusetts: National Bureau of Economic Research).
Garín J, Lester and Eric Sims, 2016, “Raise Rates to Raise Inflation? Neo-Fisherianism in the New Keynesian Model”, NBER Working Paper No. 22177 (Cambridge, Massachusetts: National Bureau of Economic Research).
Mishkin, Frederic, 1996, “The Channels of Monetary Policy Transmission: Lessons for Monetary Policy”, NBER Working Paper No. 5464 (Cambridge, Massachusetts: National Bureau of Economic Research).
Prepared by Troy Matheson (WHD).
Neo-Fisherian effects exist in standard models used by central banks under the assumption that economic agents have perfect foresight and do not base their decisions on the past observations. See Garcia-Schmidt and Woodford (2015) and Garin and others (2016).
VAR lag lengths are selected using the Swartz-Bayesian Inflation Criteria. Parameter uncertainty is captured in the analysis using bootstrapping methods, where for each VAR is resampled 1000 times.
Each VAR contains six variables so there is 720 different ways to order the variables to identify shocks: 69 of these orderings lead to unique inflation responses to an interest rate shock. 1,000 parameterizations of each reduced-form model are simulated using bootstrapping methods, leading to 69,000 different estimates of the response of inflation to an interest rate for each VAR examined.
For each country, the output gap is defined as the percent deviation of real GDP from a linear trend.
See Minutes of the 205th Meeting of the Monetary Policy Committee (“Copom”) of the Central Bank of Brazil for a discussion.
