## _wp15258 — Sections 1–2

## Source details

**Canonical URL:** [_wp15258 — Sections 1–2](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2015/_wp15258.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2015/_wp15258.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2015/_wp15258.pdf.json)

---

### Introduction: research question and context
- Research question: "is it the case that by targeting money, these countries have ended up implementing a Taylor Rule that respects the Taylor Principle?"
- Focus countries: Kenya, Tanzania and Uganda.
- Sample periods used for estimation: 2000–2011 and 2000–2013.
- Motivation:
  - East African central banks have formally targeted money aggregates rather than interest rates.
  - Money targets are often missed and readjusted, raising the possibility that central banks are de facto following a Taylor Rule.
- Related event examined: central banks from Kenya, Rwanda, Uganda and Tanzania met in October 2011 and agreed to tighten monetary conditions; the paper investigates whether there was a structural change in policy after 2011.

### Data and measurement
- Data sources: IFS and country central bank statistics.
- Observables used:
  - Inflation: quarterly variation in the CPI index.
  - Output gap: Hodrick-Prescott filter applied to the log of real quarterly GDP.
  - Interest rates: three alternative measures used due to limited availability of official Monetary Policy Rate (CBR) — "lending rate", "3-month T-bill rate" and "interbank rate".
- Sample coverage: 2000–2011 and 2000–2013.
  - Note: for Tanzania Q4 2013 GDP was imputed using WEO (April, 2015) annual growth rate applied to Q4 2012 level.
- Demeaning concern and robustness:
  - Inclusion of 1990s data shows declining trends in interest-rate measures; demeaning by the average could bias series.
  - Robustness check: jointly estimate a neutral rate using a Kalman Filter instead of simple demeaning.

### Small-scale model (structure and interpretation)
- Model structure: three equations — IS curve, Phillips curve, and a Taylor Rule; three stochastic shock processes.
- IS curve (equation (1)):
  - h_t = β · E_t[h_{t+1}] + ζ · (i_t − E_t[π_{t+1}]) + ε^h_t
  - Output gap depends on expected future output gap, ex-ante real interest rate, and transitory shock ε^h_t.
  - Steady state: zero output gap consistent with actual interest rates equal to neutral value.
- Phillips curve (equation (2)):
  - π_t = ρ · π_{t−1} + (1−ρ) · E_t[π_{t+1}] + κ · h_t + ε^p_t
  - Forward- and backward-looking terms with indexation parameter ρ; inflation depends on output gap h_t and shock ε^p_t.
- Taylor Rule (equation (3), implicit form used in estimation):
  - i_t = φ · i_{t−1} + (1−φ) · (γ_π · π_t + γ_h · h_t) + γ_{Δh} · (h_t − h_{t−1}) + ε^r_t
  - Long-run response of interest rate to inflation is γ_π; if γ_π > 1 the Taylor Principle holds.
  - Rule allows for persistence via φ and responses to the output gap and its change.

### Priors, shocks and estimation approach
- Estimation method: Bayesian estimation — posterior modes found, then MCMC to generate posterior distributions.
- Priors:
  - κ and ζ: Normal(mean 1, sd 0.5).
  - Autoregressive terms (including φ): Beta distributions with mean 0.75 and standard deviation 0.1.
  - Taylor Rule prior weights: inflation weight prior 1.5; output gap prior 0.25; standard deviations for these priors set to 0.5.
- Shock processes: ε^i_t = ρ_i · ε^i_{t−1} + η_i, with η_i ~ N(0, Σ_i).

### Estimation results — summary and key numeric findings
- Overall finding: Estimates for Kenya, Uganda and Tanzania suggest central banks (formal money targeters) behave on average in a way that respects the Taylor Principle: estimated γ_π is greater than 1 in sample results.
- Kenya (2000–2011; three models using interbank, lending, and T-bill rates)
  - φ posterior modes: 0.96, 0.94, 0.96.
  - γ_π posterior modes: 1.46, 1.49, 1.44.
  - ζ posterior modes: 0.42, 0.53, 0.42.
  - κ posterior modes: 0.18, 0.06, 0.17.
  - σ_{π} posterior modes: 2.74, 2.77, 2.76.
  - σ_{h} posterior modes (across models as listed): 0.25, 0.31, 0.58, 0.32.
- Kenya (2000–2013)
  - Inclusion of 2012 and 2013 yields small increases in γ_π: posterior modes reported as 1.56, 1.52, 1.48.
- Uganda (2000–2011; three models)
  - φ posterior modes: 0.94, 0.97, 0.95.
  - γ_π posterior modes: 1.45, 1.35, 1.49.
  - ζ posterior modes: 0.51, 0.22, 0.49.
  - κ posterior modes: 0.2, 0.21, 0.22.
  - σ_{π} posterior modes: 2.66, 1.97, 2.72.
- Uganda (2000–2013)
  - Inclusion of 2012 and 2013 leads to small increases in γ_π: posterior modes 1.51, 1.48, 1.56.
  - Weights on output gap and its variation change substantially across interest-rate measures.
- Interpretation of cross-country estimates:
  - For Kenya and Uganda, γ_π is estimated well above 1 on average, implying adherence to the Taylor Principle despite formal money targeting.
  - The Phillips curve parameter κ is often discussed as being close to 1.5 in the text (noting high correlation between HP-based output gap and inflation in discussion).
  - Interest-rate persistence φ is high (posterior modes around 0.94–0.98 in many models).
  - Including 2012–2013 observations does not produce large shifts in estimated reaction-function parameters; this provides limited empirical support for a decisive structural break in 2011, but only eight observations were added.
  - Estimations for Uganda and Kenya are on average similar; differences emerge in the size of deviations from the Taylor Rule.

### Tanzania: model fit, parameter changes and issues
- Model fit:
  - "Things are different for Tanzania: the overall fit of the model is much poorer, reflecting the widely-held view that country authorities are less flexible in the pursuit of its money aggregate objectives."
  - Evidence of improvement after 2011: "the coefficient on inflation increases significantly with the inclusion of eight additional observations, but also the effectiveness of monetary policy captured by the term ... is three times higher."
- Selected posterior values (Tanzania 2000–2011, Table 5 — model summaries for interbank, lending, Tbill rates):
  - α: posterior modes 0.99, 0.99, 0.96.
  - β: posterior modes 0.01, 0.01, 0.31.
  - ρ: posterior modes 0.67, 0.67, 0.28.
  - θ: posterior modes 0.04, 0.02, 0.55.
  - λ: posterior modes 0.8, 0.86, 0.58.
  - σ_{π}: posterior modes 1.6, 1.72, 1.41.
  - σ_{h}: posterior modes 1.86, 1.87, -0.18.
  - ψ_{p}: posterior modes 0.56, 0.53, 0.13.
  - ψ_{r}: posterior modes 0.99, 0.99, 0.91.
- Selected posterior values (Tanzania 2000–2013, Table 6):
  - α: posterior modes 0.99, 0.99, 0.99.
  - β: posterior modes 0.58, 0.47, 0.01.
  - ρ: posterior modes 0.19, 0.11, 0.68.
  - θ: posterior modes 0.12, 0.04, 0.03.
  - λ: posterior modes 0.68, 0.77, 0.78.
  - σ_{π}: posterior modes 2.57, 2.43, 1.51.
  - σ_{h}: posterior modes 0.07, -0.15, 1.94.
  - ψ_{p}: posterior modes 0.76, 0.77, 0.57.
  - ψ_{r}: posterior modes 0.99, 0.99, 0.99.
- Text interpretation: Tanzania shows large deviations from the Taylor Rule concentrated in early years, with improved stability after 2005. After 2011 the linkage of interest rates to inflation increases substantially in some specifications.

### Estimating the neutral rate via Kalman Filter (section 5.2)
- Specification change: Taylor Rule uses a jointly estimated neutral rate r_n_t with state equation:
  - r_n_t = r_n_{t-1} + e_n_t
  - e_n_t ~ N(0;1)
- Summary statement: "As the table below indicates, not much changes when the Kalman Filter is used to tease out the neutral rate of interest."
- Notable Tanzania-specific result in Kalman-filtered estimates: "After including 2012 and 2013 in the sample, the effectiveness of monetary policy increases dramatically — goes from close to zero to 0.72 — and so too the weight attributed to the inflation term in the Taylor Rule, which increases by a whopping 65%."
- Selected posterior modes from Table 7 (neutral-rate specification):
  - Uganda (neutral rates), Posterior 2011 vs Posterior 2013:
    - α: 0.96 | 0.95
    - β: 1.43 | 1.47
    - ρ: 0.44 | 0.43
    - θ: 0.18 | 0.18
    - λ: 0.75 | 0.75
    - σ_{π}: 2.8 | 2.83
    - σ_{h}: 0.43 | 0.41
    - σ_{Δh}: 0.98 | 0.95
    - ψ_{p}: 0.52 | 0.51
    - ψ_{r}: 0.85 | 0.86
  - Tanzania (neutral rates), Posterior 2011 vs Posterior 2013:
    - α: 0.92 | 0.91
    - β: 1.48 | 1.49
    - ρ: 0.51 | 0.46
    - θ: 0.23 | 0.22
    - λ: 0.73 | 0.74
    - σ_{π}: 2.67 | 2.78
    - σ_{h}: 0.37 | 0.4
    - σ_{Δh}: 0.59 | 0.6
    - ψ_{p}: 0.45 | 0.46
    - ψ_{r}: 0.78 | 0.75
  - Kenya (neutral rates), Posterior 2011 vs Posterior 2013:
    - α: 0.99 | 0.96
    - β: 0.03 | 0.28
    - ρ: 0.64 | 0.29
    - θ: 0.05 | 0.72
    - λ: 0.77 | 0.75
    - σ_{π}: 1.59 | 2.41
    - σ_{h}: 1.85 | -0.13
    - σ_{Δh}: 0.0 | 0.22
    - ψ_{p}: 0.62 | 0.3
    - ψ_{r}: 0.99 | 0.91

### Taylor Principle vs adherence to the Taylor Rule and deviations (section 5.3)
- Conceptual distinction: abiding by the Taylor Principle on average (γ_π > 1) is different from closely following the Taylor Rule through time (small ε^r_t).
- Comparative strategy: compare deviations ε^r_t for Kenya, Uganda and Tanzania to deviations in formal inflation targeters Brazil and South Africa (interbank rate focus).
- Normalized measure of non-adherence: standard deviation of ε^r_t divided by the interest rate i_t.
- Cross-country empirical findings:
  - Uganda: deviations much smaller than those for Kenya and Tanzania, and nearly identical to Brazil's. "Taylor Rule deviations in Uganda — a nouveauTargeter — are somewhat smaller than those found for South Africa."
  - Tanzania: despite ψ_r > 1 in some specifications, "the huge deviations from the Taylor Rule suggest the country has not consistently resorted to interest rates as an instrument to achieve its inflation goals." Deviations concentrated in the early years; stability improved after 2005.
  - Kenya: intermediate case; relatively high coefficient of variation driven by three isolated episodes: a contractionary movement in 2003, lax policy in 2004, and a significant loosening in 2011.

### Final comments and policy-relevant implications (section 6)
- Main synthesis:
  - "Estimation of small-scale models for Kenya, Uganda and Tanzania reveals that monetary policy in these countries respects the Taylor Principle, that is, on average nominal interest rates have been pushed higher by more than one-to-one with inflation by these countries' central banks."
  - Uganda and Kenya: "the estimated weights attributed to inflation deviations in authorities' reaction functions do not seem to have changed much after the 2011 CBs meeting. In these two countries the story seems to be that monetary conditions got tighter after 2011 because inflation went up."
  - Tanzania: "the parameter linking interest rates to inflation jumps considerably when one includes the after-2011 period in the sample."
- On structural change after October 2011:
  - Two interpretations considered: (i) central banks changed reaction functions after their meeting in October 2011; (ii) no parameter change, but policy appears stronger because inflation had risen.
  - Evidence from including 2012–2013 observations: small increases in γ_π provide limited support for a structural shift; caveat that only eight observations were added.
- On deviations from the Taylor Rule:
  - Kenya and Tanzania: "deviations from the Taylor Rule ... are far from negligible, reminiscent of money target strategies."
  - Uganda: "the normalized error term is much smaller, and in fact, smaller than those found for South Africa, the older Inflation Targeter in that continent."

*Source: _wp15258 — Sections 1–2*

### Section 1

### _wp15258 - Section 1

### Introduction: research question and context
- Research question: "is it the case that by targeting money, these countries have ended up implementing a Taylor Rule that respects the Taylor Principle?"
- Context:
  - Focus countries: Kenya, Tanzania and Uganda.
  - Time spans used for estimation: samples beginning in 2000 and ending in 2011, and ending in 2013.
  - Motivation:
    - Many low-income countries do not use interest rates as their main monetary policy instrument; East African central banks have formally targeted money aggregates.
    - Money targets are often missed and readjusted, raising the possibility central banks are de facto following a Taylor Rule.
  - Related event: central banks from Kenya, Rwanda, Uganda and Tanzania met in October 2011 and agreed to tighten monetary conditions; the paper investigates whether there was a structural change in policy after 2011.

### Data and measurement
- Data sources: IFS and country central bank statistics.
- Observables used in estimation:
  - Inflation: quarterly variation in the CPI index.
  - Output gap: Hodrick-Prescott filter applied to the log of real quarterly GDP.
  - Interest rates: due to limited availability of official Monetary Policy Rate (CBR), three alternative measures used: "lending rate", "3-month T-bill rate" and "interbank rate".
- Sample coverage: 2000–2011 and 2000–2013 (note: for Tanzania Q4 2013 GDP was imputed using WEO (April, 2015) annual growth rate applied to Q4 2012 level).
- Demeaning concern: inclusion of 1990s data shows declining trends in interest-rate measures; because model variables are deviations from steady state (proxied by the average), using the 1990s would bias demeaned series. Robustness check: jointly estimate a neutral rate using Kalman Filter instead of simple demeaning.

### Small-scale model (structure and interpretation)
- Model comprises three equations: IS curve, Phillips curve, and a Taylor Rule; three stochastic shock processes.
- IS curve (equation (1)):
  - h_t = β · E_t[h_{t+1}] + ζ · (i_t − E_t[π_{t+1}]) + ε^h_t
  - Interpretation: output gap depends on expected future output gap, ex-ante real interest rate, and a transitory shock ε^h_t.
  - Steady state: zero output gap consistent with actual interest rates equal to neutral value.
- Phillips curve (equation (2)):
  - π_t = ρ · π_{t−1} + (1−ρ) · E_t[π_{t+1}] + κ · h_t + ε^p_t
  - Features: forward-looking and backward-looking terms; captures indexation with parameter ρ; inflation depends on output gap h_t and shock ε^p_t.
  - In economies with explicit inflation targets, π (steady-state inflation) would equal the target.
- (Implicit) Taylor Rule (equation (3)):
  - i_t = φ · i_{t−1} + (1−φ) · (γ_π · π_t + γ_h · h_t) + γ_{Δh} · (h_t − h_{t−1}) + ε^r_t
  - Long-run response of interest rate to inflation is γ_π; if γ_π > 1 the Taylor Principle holds (nominal rates increase more than proportionally to inflation), stabilizing inflation.
  - Rule allows for persistence via φ and for responses to the output gap and its change.

### Priors, shocks and estimation approach
- Estimation method: Bayesian estimation; posterior modes found, then MCMC to generate posterior distributions.
- Priors:
  - Elasticity parameters κ and ζ: Normal distribution with mean 1 and standard deviation 0.5.
  - Autoregressive terms (including φ): Beta distributions with mean 0.75 and standard deviation 0.1.
  - Taylor Rule prior weights: inflation weight prior 1.5; output gap prior 0.25. Standard deviations for these priors set to 0.5 (larger than Smets-Wouters’s 0.25 and 0.125).
- Shock processes: ε^i_t = ρ_i · ε^i_{t−1} + η_i, with η_i ~ N(0, Σ_i).

### Estimation results — summary and key numeric findings
- Overall finding: Estimations for Kenya, Uganda and Tanzania suggest central banks (formal money targeters) behave on average in a way that respects the Taylor Principle: estimated γ_π is greater than 1 in sample results.
- Kenya (2000–2011 and 2000–2013):
  - 2000–2011 (Table 1, three models using interbank, lending, and T-bill rates):
    - φ (prior mean 0.75) posterior modes: 0.96, 0.94, 0.96.
    - γ_π posterior modes: 1.46, 1.49, 1.44.
    - ζ (IS interest-rate elasticity) posterior modes: 0.42, 0.53, 0.42.
    - κ (Phillips-curve output gap coefficient) posterior modes: 0.18, 0.06, 0.17.
    - σ_{π} posterior modes: 2.74, 2.77, 2.76.
    - σ_{h} posterior modes: 0.25, 0.31, 0.58, 0.32 (across models as listed).
  - 2000–2013 (Table 2): inclusion of 2012 and 2013 yields small increases in γ_π (e.g., γ_π posterior modes: 1.56, 1.52, 1.48) — the rise is described as tiny given only eight added observations.
- Uganda (2000–2011 and 2000–2013):
  - 2000–2011 (Table 3, three models):
    - φ posterior modes: 0.94, 0.97, 0.95.
    - γ_π posterior modes: 1.45, 1.35, 1.49.
    - ζ posterior modes: 0.51, 0.22, 0.49.
    - κ posterior modes: 0.2, 0.21, 0.22.
    - σ_{π} posterior modes: 2.66, 1.97, 2.72.
  - 2000–2013 (Table 4): inclusion of 2012 and 2013 leads to small increases in γ_π (posterior modes: 1.51, 1.48, 1.56). Weights on output gap and its variation change substantially across interest-rate measures.
- Interpretation of results:
  - For Kenya and Uganda, γ_π is estimated well above 1 on average, implying adherence to the Taylor Principle despite formal money targeting.
  - The Phillips curve parameter κ is often close to 1.5 in reported discussion (suggesting the simple HP-based output gap is highly correlated with inflation).
  - Interest-rate persistence φ is high (posterior modes near the prior mean 0.75 and estimated around 0.94–0.98 in many models).
  - Including 2012–2013 observations does not produce large shifts in estimated reaction-function parameters, which provides limited empirical support for a decisive structural break in 2011; however, the sample extension is only eight observations.
  - Estimations for Uganda and Kenya are on average similar; differences emerge when examining the size of deviations from the Taylor Rule (not detailed in Section 1).

### Policy-relevant implications and empirical questions addressed
- Main implication: Even when central banks formally target money aggregates, their implicit policy paths for interest rates may effectively implement a Taylor Rule that respects the Taylor Principle (γ_π > 1), implying nominal interest rate responses that help stabilize inflation.
- On structural change after October 2011:
  - Two interpretations considered: (i) central banks changed reaction functions after their meeting in October 2011; (ii) no parameter change, but policy appears stronger because inflation had risen.
  - Evidence from inclusion of 2012–2013 observations: small differences in posterior modes (γ_π increases slightly), offering limited support for a structural shift; caveat that only eight observations were added.
- On deviations from the Taylor Rule:
  - The size of deviations between actual interest rates and Taylor-Rule-implied rates matters: large deviations would indicate de facto monetary targeting; small deviations would suggest a gradual transition toward interest-rate instruments. Section 1 notes that Uganda’s deviations are smaller (similar to Brazil) while Kenya and Tanzania experienced steeper deviations (details provided in later sections).

*WP/15/258 — Taylor Visits Africa (Section 1)*

### Section 2

### _wp15258 - Section 2

### Tanzania: model fit and parameter changes (2000-11 and 2000-13)
- General finding: "Things are di§erent for Tanzania: the overall Öt of the model is much poorer, reáecting the widely-held view that countries authorities are less áexible in the pursuit of its money aggregate objectives. On the other hand, in Tanzaniaís case there seems to be more of an improvement after 2011. Not only the coe¢ cient on ináation increases signiÖcantly with the inclusion of eight additional observations, but also the e§ectiveness of monetary policy captured by the termis three times higher."
- Table 5: Tanzania 2000-11 (Parameters | Priors | model 1: interbank rates | model 2: lending rates | model 3: Tbill rates)
  - α: 0.75 | 0.99 | 0.99 | 0.96
  - β: 1.0 | 0.01 | 0.01 | 0.31
  - ρ: 0.75 | 0.67 | 0.67 | 0.28
  - θ: 1.0 | 0.04 | 0.02 | 0.55
  - λ: 0.75 | 0.8 | 0.86 | 0.58
  - σ_{π}: 1.5 | 1.6 | 1.72 | 1.41
  - σ_{h}: 0.25 | 1.86 | 1.87 | -0.18
  - σ_{Δh}: 0.5 | 0.01 | 0.49
  - ψ_{p}: 0.75 | 0.56 | 0.53 | 0.13
  - ψ_{r}: 0.75 | 0.99 | 0.99 | 0.91
- Table 6: Tanzania 2000-13 (Parameters | Priors | model 1: interbank rates | model 2: lending rates | model 3: Tbill rates)
  - α: 0.75 | 0.99 | 0.99 | 0.99
  - β: 1.0 | 0.58 | 0.47 | 0.01
  - ρ: 0.75 | 0.19 | 0.11 | 0.68
  - θ: 1.0 | 0.12 | 0.04 | 0.03
  - λ: 0.75 | 0.68 | 0.77 | 0.78
  - σ_{π}: 1.5 | 2.57 | 2.43 | 1.51
  - σ_{h}: 0.25 | 0.07 | -0.15 | 1.94
  - σ_{Δh}: 0.5 | 0.03 | 0.0 | 0.01
  - ψ_{p}: 0.75 | 0.76 | 0.77 | 0.57
  - ψ_{r}: 0.75 | 0.99 | 0.99 | 0.99

### Estimating the neutral rate via Kalman Filter (section 5.2)
- Specification change: the Taylor Rule now uses a jointly estimated neutral rate r_n instead of demeaning by the average. State equation added:
  - r_n_t = r_n_{t-1} + e_n_t
  - e_n_t ~ N(0;1)
- Summary statement: "As the table below indicates, not much changes when the Kalman Filter is used to tease out the neutral rate of interest."
- Notable Tanzania-specific result: "After including 2012 and 2013 in the sample, the e§ectiveness of monetary policy increases dramatically ñgoes from close to zero to 0.72 ñ and so too the weight attributed to the ináation term in the Taylor Rule, which increases by a whopping 65%."
- Table 7: estimating neutral rates instead of demeaning (Parameters | Priors | Posterior 2011 | Posterior 2013) — presented separately for Uganda (neutral rates), Tanzania (neutral rates), Kenya (neutral rates)
  - Uganda (neutral rates)
    - α: 0.75 | 0.96 | 0.95
    - β: 1 | 1.43 | 1.47
    - ρ: 0.75 | 0.44 | 0.43
    - θ: 1 | 0.18 | 0.18
    - λ: 0.75 | 0.75 | 0.75
    - σ_{π}: 1.5 | 2.8 | 2.83
    - σ_{h}: 0.25 | 0.43 | 0.41
    - σ_{Δh}: 0.5 | 0.98 | 0.95
    - ψ_{p}: 0.75 | 0.52 | 0.51
    - ψ_{r}: 0.75 | 0.85 | 0.86
  - Tanzania (neutral rates)
    - α: 0.75 | 0.92 | 0.91
    - β: 1 | 1.48 | 1.49
    - ρ: 0.75 | 0.51 | 0.46
    - θ: 1 | 0.23 | 0.22
    - λ: 0.75 | 0.73 | 0.74
    - σ_{π}: 1.5 | 2.67 | 2.78
    - σ_{h}: 0.25 | 0.37 | 0.4
    - σ_{Δh}: 0.5 | 0.59 | 0.6
    - ψ_{p}: 0.75 | 0.45 | 0.46
    - ψ_{r}: 0.75 | 0.78 | 0.75
  - Kenya (neutral rates)
    - α: 0.75 | 0.99 | 0.96
    - β: 1 | 0.03 | 0.28
    - ρ: 0.75 | 0.64 | 0.29
    - θ: 1 | 0.05 | 0.72
    - λ: 0.75 | 0.77 | 0.75
    - σ_{π}: 1.5 | 1.59 | 2.41
    - σ_{h}: 0.25 | 1.85 | -0.13
    - σ_{Δh}: 0.5 | 0.0 | 0.22
    - ψ_{p}: 0.75 | 0.62 | 0.3
    - ψ_{r}: 0.75 | 0.99 | 0.91

### Taylor Principle vs Taylor Rule adherence and deviations (section 5.3)
- Conceptual distinction: "Now, it is one thing to abide by the Taylor Principle on average, quite another to closely follow the Taylor Rule through time."
- Comparative strategy: deviations (ε^r_t) for the three countries are compared to deviations observed in two formal Inflation Targeters — Brazil and South Africa. Results presented for the interbank rate.
- Table 8: Brazil and South Africa (Parameters | Priors | Posterior 2013)
  - South Africa
    - α: 0.75 | 0.94 | 0.75 | 0.87
    - β: 1 | 0.56 | 1 | 1.62
    - ρ: 0.75 | 0.82 | 0.75 | 0.81
    - θ: 1 | 1.39 | 1 | 1.28
    - λ: 0.75 | 0.43 | 0.75 | 0.39
    - σ_{π}: 1.5 | 3.41 | 1.5 | 3.33
    - σ_{h}: 0.25 | 0.31 | 0.25 | 0.6
    - σ_{Δh}: 0.5 | 0.25 | 0.5 | 0.79
    - ψ_{p}: 0.75 | 0.57 | 0.75 | 0.41
    - ψ_{r}: 0.75 | 0.87 | 0.75 | 0.69
  - Brazil (same columns reflected)
- Normalized measure of non-adherence: standard-deviation of ε^r_t divided by the interest rate i_t.
- Cross-country empirical findings:
  - Uganda: deviations much smaller than those for Kenya and Tanzania, and nearly identical to Brazil's. "Taylor Rule deviations in Uganda - anouveauTargeter ñ are somewhat smaller than those found for South Africa."
  - Tanzania: despite ψ_r > 1, "the huge deviations from the Taylor Rule suggest the country has not consistently resorted to interest rates as an instrument to achieve its ináation goals." These are concentrated in the early years of the sample, with stability improving after 2005.
  - Kenya: intermediate and "somewhat odd" case — relatively high coefficient of variation driven by three isolated episodes: a contractionary movement in 2003, lax policy in 2004, and a significant loosening in 2011.

### Final comments (section 6)
- Main synthesis:
  - "Estimation of small-scale models for Kenya, Uganda and Tanzania reveals that monetary policy in these countries respects the Taylor Principle, that is, on average nominal interest rates have been pushed higher by more than one-to-one with ináation by these countriesí central banks."
  - Uganda and Kenya: "the estimated weights attributed to ináation deviations in authoritiesí reaction functions do not seem to have changed much after the 2011 CBs meeting. In these two countries the story seems to be that monetary conditions got tighter after 2011 because ináation went up."
  - Tanzania: "the parameter linking interest rates to ináation jumps considerably when one includes the after-2011 period in the sample."
- Deviations from Taylor Rule:
  - Kenya and Tanzania: "deviations from the Taylor Rule ... are far from negligible, reminiscent of money target strategies."
  - Uganda: "the normalized error term is much smaller, and in fact, smaller then those found for South Africa, the older Ináation Targeter in that continent."

*Source: _wp15258 - Section 2*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2015/_wp15258.pdf_