A. Introduction

1. Since 1999, inflation in Colombia has declined to 4½percent, the lowest level in decades. This represents a sharp change from Colombia’s decades-long track record of persistent inflation of about 20–30 percent a year (Uribe, 1994; Dornbusch and Fischer, 1993). The shift in inflation performance can be explained largely by the adoption of a new policy framework in 1999–2000. In September 1999, Colombia decided to abandon the crawling peg exchange rate regime that had been in place since 1967 and to introduce a flexible exchange rate regime. In 2000, the country formally adopted an inflation-targeting framework.

2. This paper looks at the effect of this new policy regime on the persistence of inflation and whether expected inflation now has a larger effect on current inflation. A central bank facing inflation that is largely determined by past inflation will probably have to tighten monetary policy more to achieve an inflation target. Moreover, a successful inflation-targeting framework should enhance the credibility of monetary policy and make current inflation a function largely of expected inflation. This makes monetary policy more effective, requiring less tightening to achieve a given inflation target.

3. Some studies have measured the extent of price stickiness in Colombia. Bejarano (2005), using data between 1982 and 2002, identified that about 70 percent of firms kept their prices fixed for about three quarters. Moreover, Gomez and Julio (2003); Gomez, Vargas and Uribe (2002); and Hamann, Perez and Rodriguez (2006); have recognized the existence of persistence and incorporated in the formulation of the core model of the monetary transmission mechanism used by the central bank. These studies do not present evidence on whether inflation persistence in recent years has changed and by how much.

4. This paper measures persistence using two methodologies. First, it identifies the statistical importance of the autoregressive component of inflation, according to several measures of inflation. Second, it estimates the separate effects of backward and forward looking inflation as well as of marginal costs on current inflation by estimating a hybrid new Keynesian Phillips curve (NKPC). This approach builds on the work of Bejarano (2005) by estimating a reduced form hybrid NKPC and extending the sample until the first quarter of 2006.

5. The findings of this chapter are as follows. First, the importance of the backward-looking component of inflation appears to have declined after the introduction of the inflation targeting framework and the adoption of a flexible exchange rate regime. Second, forward looking inflation expectations have gained importance in explaining observed inflation after Colombia adopted a floating exchange rate regime in 1999. The remainder of this chapter is organized as follows. Section II provides a brief review of the disinflation strategy. Section III presents the empirical results. Section IV concludes.

B. The Disinflation Strategy

6. Since the formal adoption of the inflation targeting regime in 2000, inflation has declined consistently. The average headline inflation rate between 1982 and the second quarter of 1999 was 21 percent and has declined to 7 percent from the third quarter of 1999 through the second quarter of 2006. At the same time, inflation volatility, measured by its standard deviation, has decreased across all measures of inflation, and there have been a number of times that prices—especially the tradable component—have fallen (Table 1, Figure 1).

Table 1.

Descriptive Statistics on Inflation Measures

Sources: Central Bank of Colombia; and Fund staff estimates.

Table 1.

Descriptive Statistics on Inflation Measures

	Inflation Measures
				Non-
			Tradable	tradable	Regulated
Measure	Headline	Food	goods	goods	goods
1983.Q2-1999.Q2
Average	21.4	22.3	20.6	20.5	26.6
Standard Deviation	4.8	6.5	5.8	6.7	7.0
1999.Q3-2006.Q2
Average	6.9	7.3	6.8	4.8	13.0
Standard Deviation	1.6	2.0	3.6	1.0	5.5

Sources: Central Bank of Colombia; and Fund staff estimates.

View Table

Table 1.

Descriptive Statistics on Inflation Measures

	Inflation Measures
				Non-
			Tradable	tradable	Regulated
Measure	Headline	Food	goods	goods	goods
1983.Q2-1999.Q2
Average	21.4	22.3	20.6	20.5	26.6
Standard Deviation	4.8	6.5	5.8	6.7	7.0
1999.Q3-2006.Q2
Average	6.9	7.3	6.8	4.8	13.0
Standard Deviation	1.6	2.0	3.6	1.0	5.5

Sources: Central Bank of Colombia; and Fund staff estimates.

Level and Volatility of Annual Headline Inflation

Citation: IMF Staff Country Reports 2006, 401; 10.5089/9781451808919.002.A002

Sources: Central Bank of Colombia; and Fund staff calculations.

• Download Figure
• Download figure as PowerPoint slide

View Full Size

Level and Volatility of Annual Headline Inflation

Citation: IMF Staff Country Reports 2006, 401; 10.5089/9781451808919.002.A002

Sources: Central Bank of Colombia; and Fund staff calculations.

• Download Figure
• Download figure as PowerPoint slide

Level and Volatility of Annual Headline Inflation

Citation: IMF Staff Country Reports 2006, 401; 10.5089/9781451808919.002.A002

Sources: Central Bank of Colombia; and Fund staff calculations.

• Download Figure
• Download figure as PowerPoint slide

7. This disinflation process has not been free from difficulties. First, the long period of moderate inflation stretching from the 1970s through the 1990s created institutional structures that contributed to inertia.⁶ This inertia, in turn, increased the short-run costs of adopting a credible disinflation strategy, as there was little support for embracing the costs embedded in breaking with the status quo. Second, while the central bank moved to informal inflation targeting in the early 1990’s, other reforms increased fiscal dominance and complicated monetary policy management.⁷ Third, the turbulence in financial capital markets during 1998-99 and in 2002 affected the risk perception toward Colombia and had the potential to interrupt the disinflation strategy.

8. Several reasons help to explain the drastic fall in inflation. Some authors (Clavijo, 2000) characterize this shift as an opportunistic disinflation, as the sustained rise in unemployment following the 1999 crisis virtually eliminated all demand pressures. Nevertheless, inflation has continued to decline even after economic growth recovered to 5 percent a year in 2004–06 and unemployment fell to about 10 percent—close to its estimated natural rate. This probably reflects the strengthening of economic policies since 2002 that improved confidence and the climate for investment and supported faster growth in potential output.

9. The decline in inflation can also be explained by several reforms to strengthen monetary policy implementation. First, the 1999 adoption of a floating exchange rate regime provided enough flexibility to help the economy weather turbulent international capital markets episodes and broke the indexation created by the crawling peg exchange rate regime. Second, the introduction of an explicit inflation target in 2000 as the operational monetary framework enhanced the de facto independence of the central bank, originally granted by the 1991 Constitution. Third, the clear hierarchy given to inflation control over any other monetary policy goal gave the central bank more protection from political pressures than in the past. Fourth, the central bank has had the independence to select its policy instrument since 2000. Fifth, a number of mechanisms have been adopted in recent years to increase the transparency of monetary policy.⁸

10. The Colombian inflation-targeting framework has been one of inflation forecast targeting, with the interest rate as the policy key instrument. Under this scheme, whenever the inflation forecast deviates from the inflation target, the central bank modifies the interest rate accordingly. In operational terms, the central bank sets the minimum and maximum interest rates on auctions for overnight repo and reverse-repo operations, and provides or withdraws liquidity accordingly. These policy rates affect the rate on the overnight interbank money market, on which banks borrow and lend money. The expected evolution of this short-term rate, in turn, affects interest rates at longer horizons, which in the presence of nominal price rigidities, can influence households and firms’ savings and investment decisions. This co-movement of the overnight interbank rate with the rates on three, six and twelve-month certificates of deposit is evident over the period 1999-2006.⁹

Interbank Interest Rate and Policy Rate

(In percent)

Citation: IMF Staff Country Reports 2006, 401; 10.5089/9781451808919.002.A002

Source: Central Bank of Colombia.

• Download Figure
• Download figure as PowerPoint slide

View Full Size

Interbank Interest Rate and Policy Rate

(In percent)

Citation: IMF Staff Country Reports 2006, 401; 10.5089/9781451808919.002.A002

Source: Central Bank of Colombia.

• Download Figure
• Download figure as PowerPoint slide

Interbank Interest Rate and Policy Rate

(In percent)

Citation: IMF Staff Country Reports 2006, 401; 10.5089/9781451808919.002.A002

Source: Central Bank of Colombia.

• Download Figure
• Download figure as PowerPoint slide

Certificates of Deposit and Interbank Interest Rate

(In percent)

Citation: IMF Staff Country Reports 2006, 401; 10.5089/9781451808919.002.A002

Source: Central Bank of Colombia.

• Download Figure
• Download figure as PowerPoint slide

View Full Size

Certificates of Deposit and Interbank Interest Rate

(In percent)

Citation: IMF Staff Country Reports 2006, 401; 10.5089/9781451808919.002.A002

Source: Central Bank of Colombia.

• Download Figure
• Download figure as PowerPoint slide

Certificates of Deposit and Interbank Interest Rate

(In percent)

Citation: IMF Staff Country Reports 2006, 401; 10.5089/9781451808919.002.A002

Source: Central Bank of Colombia.

• Download Figure
• Download figure as PowerPoint slide

C. Measuring Inflation Persistence

11. In this paper, we seek to explain the sources of inflation using the theoretical framework embedded in the hybrid New Keynesian Phillips Curve (NKPC), as presented in Gali and Gertler (1999) (GG). GG assumed staggered price setting framework a la Calvo (1983), where firms have a probability of receiving a signal to reset prices each period. This signal is independent of the time elapsed since the last signal was received and there are firms that maintain their prices unchanged in any given period. GG modified the Calvo pricing scheme by allowing only a fraction of firms to adjust prices according to expected future marginal costs. The remainder of firms set prices by a rule of thumb formula that combines newly adjusted prices and expected inflation that is based on lagged inflation. GG obtain a hybrid NKPC that is equal to

View Expanded

where π_t is the inflation at time t; E_tπ_t+1 is the expected inflation at time t+1, conditional on information at time t; mc_t is the deviation from trend of the marginal cost at time t, and ε is an exogenous marginal cost shock. When all firms set prices according to expected marginal costs, γ_b = 0 and equation (1) becomes the stylized forward-looking NKPC.

12. The hybrid NKPC is also used to identify the sources of inflation persistence. In particular, each of the three right hand-side terms of equation (1), is associated with different definition of persistence, in turn: extrinsic persistence (λmc_t), which arises from changes in real marginal costs; the intrinsic persistence or inflation inertia (γ_bπ_t−1) due to the price-setting mechanism existing in the economy; and the expectations-based persistence (γ_fE_tπ_t+1) that is associated with forward looking inflation expectations. We should keep in mind that the presence of inflation inertia is usually associated with imperfect credibility of monetary policy, and thus with larger output costs of disinflation.

13. This chapter uses two approaches to identify persistence. The first is a statistical approach, and identifies persistence with the autoregressive component of inflation.¹⁰ This is accomplished by estimating the following equation:

View Expanded

The sum of the autoregressive coefficients is used to measure of the dependency of inflation on its own past. Equation (2) is estimated using quarterly inflation for each of the price indexes presented in Table 1, as well as for the GDP deflator. The sample is divided in two periods: one through the second quarter of 1999, just prior to the adoption of a free floating exchange rate regime, and the other starting in the third quarter of 1999 through the first quarter of 2006, which also coincides with the implementation of the inflation targeting framework, and the flexible exchange rate regime.

14. In general, the results show a reduction in inflation inertia (intrinsic persistence) (Table 2). With the exception of tradable goods inflation and inflation of regulated prices, the inflation measures show a reduction in the sum of the autoregressive coefficients for the period after the adoption of the floating exchange rate regime, which can be taken as evidence of lower persistence. The result of particular interest is the big decline in the persistence of non-tradable goods inflation from 0.96 to 0.37.

Table 2.

Univariate Evidence on Inflation Persistence ^1/

^1/Standard errors in parenthesis.

Table 2.

Univariate Evidence on Inflation Persistence ^1/

						Regulated
		Headline	Non-Tradable	Tradable		Prices
Measures of inflation	GDP Deflator	Inflation	Inflation	Inflation	Food Inflation	Inflation
${\pi }_t=\alpha +\underset{i=1}{\overset{5}{\mathrm{\Sigma }}}{\hat{b}}_i^{\pi }t-i$
Sample Period: 1985Q3 - 1999Q2
Intercept	0.01	0.010	0.010	0.010	0.02	0.02
$\begin{array}{c}\mathrm{\sum }_{i=1}^5\end{array}\hat{b}i$	(0.012)	(0.006)	(0.01)	(0.003)	(0.003)	(0.01)
	0.83	0.96	0.96	0.92	0.55	0.70
	(0.22)	(0.13)	(0.057)	(0.072)	(0.268)	(0.057)
R2	0.30	0.63	0.87	0.81	0.34	0.29
SEE	0.02	0.01	0.01	0.01	0.02	0.02
DW	1.95	1.98	2.02	1.98	2.07	1.99
Sample Period: 1999Q3 - 2006Q1
Intercept	0.009	0.0090	0.007	-0.001	0.013	0.001
$\begin{array}{c}\mathrm{\sum }_{i=1}^5\end{array}\hat{b}i$	(0.01)	(0.002)	(0.002)	(0.002)	(0.009)	(0.006)
	0.42	0.93	0.37	0.97	0.28	0.88
	(0.51)	(0.17)	(0.169)	(0.19)	(0.5)	(0.19)
R2	0.11	0.72	0.39	0.85	0.27	0.58
SEE	0.02	0.00	0.00	0.00	0.01	0.01
DW	1.98	1.82	2.03	1.56	2.20	1.87

^1/Standard errors in parenthesis.

View Table

Table 2.

Univariate Evidence on Inflation Persistence ^1/

						Regulated
		Headline	Non-Tradable	Tradable		Prices
Measures of inflation	GDP Deflator	Inflation	Inflation	Inflation	Food Inflation	Inflation
${\pi }_t=\alpha +\underset{i=1}{\overset{5}{\mathrm{\Sigma }}}{\hat{b}}_i^{\pi }t-i$ View Expanded
Sample Period: 1985Q3 - 1999Q2
Intercept	0.01	0.010	0.010	0.010	0.02	0.02
$\begin{array}{c}\mathrm{\sum }_{i=1}^5\end{array}\hat{b}i$ View Expanded	(0.012)	(0.006)	(0.01)	(0.003)	(0.003)	(0.01)
	0.83	0.96	0.96	0.92	0.55	0.70
	(0.22)	(0.13)	(0.057)	(0.072)	(0.268)	(0.057)
R2	0.30	0.63	0.87	0.81	0.34	0.29
SEE	0.02	0.01	0.01	0.01	0.02	0.02
DW	1.95	1.98	2.02	1.98	2.07	1.99
Sample Period: 1999Q3 - 2006Q1
Intercept	0.009	0.0090	0.007	-0.001	0.013	0.001
$\begin{array}{c}\mathrm{\sum }_{i=1}^5\end{array}\hat{b}i$ View Expanded	(0.01)	(0.002)	(0.002)	(0.002)	(0.009)	(0.006)
	0.42	0.93	0.37	0.97	0.28	0.88
	(0.51)	(0.17)	(0.169)	(0.19)	(0.5)	(0.19)
R2	0.11	0.72	0.39	0.85	0.27	0.58
SEE	0.02	0.00	0.00	0.00	0.01	0.01
DW	1.98	1.82	2.03	1.56	2.20	1.87

^1/Standard errors in parenthesis.

15. The second approach assesses the relative importance of economic factors that explain inflation by estimating the hybrid NKPC. In particular, a reduced form of equation (1) is estimated with a GDP deflator as the underlying inflation measure, which follows closely the theoretical foundations of equation (1).¹¹ Following GG, we assume rational expectations and estimate equation (1) using Generalized Method of Moments (GMM) with lags of the variables as instruments and assuming the error term ε is i.i.d. This procedure allows us to distinguish backward-looking from forward-looking behavior. We also estimate the coefficients of the equation with and without the restriction that the parameters of backward and forward looking inflation (λ_f λ_b) add to one, which is the condition for monetary policy to be neutral in the long-run.

Results

17. The estimation of the hybrid NKPC for Colombia shows that forward looking expectations now play a larger role in determining current inflation (Table 3). The coefficients on the marginal cost and the expected inflation are statistically different from zero and positive. A positive coefficient in the marginal cost term (λ> 0) may be associated with higher bottlenecks in the supply side of the economy or with higher demand pressures—a reduction in the output gap—which could create inflationary pressures. The coefficient of expected inflation is significantly different from zero, which can be interpreted as evidence of forward looking behavior in the price setting mechanism. An interesting exercise is to measure whether the importance of this forward looking behavior has changed since Colombia adopted the floating exchange rate regime. This exercise is accomplished by estimating the equation (1) for the restricted sample (before exchange rate floating) and for the whole sample. The increase in λ_f for the whole sample can be taken as an indication that after the adoption of the floating exchange rate regime the importance of the forward looking behavior in inflation dynamics increased.

Table 3.

GMM Estimation of the Hybrid Neo-Keynesian Phillips Curve ^1/2/

^1/Standard errors in parenthesis.

^2/The instruments used were lags of inflation measures, marginal cost, output gap, exchange rate changes and seasonal dummies. Across regressions the test for overidentification can not be rejected.

Table 3.

GMM Estimation of the Hybrid Neo-Keynesian Phillips Curve ^1/2/

	GDP Deflator			Headline Inflation
	γf	γb	λ	γf	γb	λ
Sample Period: 1986Q2 - 1999Q2
Unrestricted	0.801	0.047	0.01	0.57	0.31	0.01
	(0.186)	(0.144)	(0.067)	(0.143)	(0.127)	(0.03)
γf + γb=1	0.926	…	0.119	0.656	…	0.021
	(0.149)		(0.075)	(0.129)		(0.029)
Sample Period: 1986Q2 -2006Q1
Unrestricted	0.9	-0.007	0.136	0.688	0.29	0.042
	(0.182)	(0.130)	(0.081)	(0.028)	(0.127)	(0.028)
γf + γb=1	1.01	…	0.13	0.703	…	0.044
	(0.14)		(0.084)	(0.129)		(0.129)

^1/Standard errors in parenthesis.

^2/The instruments used were lags of inflation measures, marginal cost, output gap, exchange rate changes and seasonal dummies. Across regressions the test for overidentification can not be rejected.

View Table

Table 3.

GMM Estimation of the Hybrid Neo-Keynesian Phillips Curve ^1/2/

	GDP Deflator			Headline Inflation
	γf	γb	λ	γf	γb	λ
Sample Period: 1986Q2 - 1999Q2
Unrestricted	0.801	0.047	0.01	0.57	0.31	0.01
	(0.186)	(0.144)	(0.067)	(0.143)	(0.127)	(0.03)
γf + γb=1	0.926	…	0.119	0.656	…	0.021
	(0.149)		(0.075)	(0.129)		(0.029)
Sample Period: 1986Q2 -2006Q1
Unrestricted	0.9	-0.007	0.136	0.688	0.29	0.042
	(0.182)	(0.130)	(0.081)	(0.028)	(0.127)	(0.028)
γf + γb=1	1.01	…	0.13	0.703	…	0.044
	(0.14)		(0.084)	(0.129)		(0.129)

^1/Standard errors in parenthesis.

^2/The instruments used were lags of inflation measures, marginal cost, output gap, exchange rate changes and seasonal dummies. Across regressions the test for overidentification can not be rejected.

D. Concluding Remarks

18. The reduction in inflation obtained since 1999 has been remarkable. This disinflation appears to be associated with a decline in inflation inertia (intrinsic inflation persistence) and an increase in the forward looking component of inflation. These results, together with the decline in the volatility of inflation, suggest that Colombia’s inflation-targeting framework has enhanced the credibility and effectiveness of monetary policy. However, a statistical test of the link between the new policy regime and these changes in inflation persistence is a topic for another research effort. Going forward, the reduction in the backward looking component of inflation (intrinsic persistence) could also indicate a lower short-term cost of disinflation in terms of output and employment.