## cr17109

## Source details

**Canonical URL:** [cr17109](https://www.imf.org/-/media/files/publications/cr/2017/cr17109.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/cr/2017/cr17109.pdf.md)
- [Structured JSON version](/-/media/files/publications/cr/2017/cr17109.pdf.json)

---

### A. Fiscal context and recent developments
- Oil accounts for more than 50 percent of exports.
- Government oil-related revenues averaged 10-12 percent of GDP over the last decade—more than 40 percent of total revenues.
- The overall fiscal position reversed from an average surplus of 4½ percent of GDP in 2011-13 to a deficit of 5 percent of GDP 2015-16.
- Oil revenues fell by more than 70 percent in dollar terms from 2013 to 2016; they declined by just half as much in tenge terms due to the authorities’ decision to allow the tenge to depreciate.
- NFRK assets stood at $60 billion at end-2016, equivalent to 43 percent of GDP.
- Public debt stood at 21 percent of GDP and total external debt at 122 percent of GDP.

### B. National Fund of the Republic of Kazakhstan (NFRK) framework and recent changes
- NFRK established in 2000; receives all oil-related revenues including: (i) the government’s share of profits in oil projects; (ii) receipts from the export duty tax; and (iii) various mineral extraction taxes and fees.
- Government’s share of oil revenues historically averaged 30-35 percent.
- Under the current framework (pre-2018), guaranteed transfers are set at $8 billion (±15% depending on the cyclical position of the economy and the financing needs); targeted transfers are discretionary.
- New NFRK framework (effective from 2018):
  - Guaranteed transfers will decline gradually to about KZT 2 trillion (about $6 billion at the current exchange rate) by 2020.
  - NFRK assets cannot fall below 30 percent of GDP.
  - The budget will target a specific and pre-determined path for the non-oil deficit in the medium-term.
  - The new framework aims to reduce the non-oil deficit from 8.3 percent of GDP in 2016 to 7 percent of GDP in 2020 and to 6 percent of GDP in 2025.

### C. Risks under an unchanged policy (sustainability concerns)
- Under a passive policy scenario with a broadly unchanged non-oil deficit going forward:
  - Government’s net financial assets (NFRK assets minus government debt) would fall close to zero by 2021 (Figures 4-5 referenced).
  - Keeping the debt-to-GDP ratio constant while covering government financing needs largely from the NFRK would exhaust NFRK assets by the mid-2020s.
- Table 1 presents the year when net financial assets would turn negative under various combinations of the long-term oil price and the (constant) non-oil primary deficit (table referenced; values in table not reproduced here).

### D. Long-term benchmarks and methodology
- Benchmarks derive from estimating Kazakhstan’s net wealth = initial stock of net financial assets + present value to the government of oil assets in the ground.
- Key uncertainties in the oil-wealth valuation: reserves and extraction rates, future oil prices, exchange rate, government’s share of oil revenues.
- Two implementations of the Permanent Income Hypothesis (PIH) used:
  - Perpetuity: transfer to the budget equals revenues from interest receipts on the stock of NFRK assets.
  - Annuity: assumes a lifespan of the NFRK of forty years.
- Four scenarios considered:
  - (i) a perpetuity-non-oil primary deficit (NOPD) constant in percent of non-oil GDP starting from 2017;
  - (ii) delaying the adjustment to option (i) by 5 years (constant perpetuity-NOPD starting from 2022);
  - (iii) a constant real perpetuity-NOPD;
  - (iv) a 40-year annuity-NOPD constant in percent of non-oil GDP.
- Figures 6a-b present NOPD paths and government net financial assets under the various cases and are compared to staff’s medium-term baseline projections.

### E. PIH results and implications
- PIH main findings:
  - A constant perpetuity-NOPD as a share of GDP of 1.1 percent would maximize Kazakhstan’s net wealth at end of the forecasting horizon.
  - Targeting a constant NOPD as a share of GDP of 6.2 percent over 40 years would be consistent with net wealth as of end-2016 under the annuity approach; however, under this annuity approach (40-year), net wealth would be fully exhausted by the end of the 40-year annuity period.
  - If the NOPD is constant in real terms over the 40-year horizon, the 40-year NOPD annuity would average 10 percent of GDP in the first decade before declining steadily (not shown in figures).
- Considerations for choosing among approaches:
  - The perpetuity approach (1.1 percent NOPD) could be considered too tight for an emerging market economy with social and investment needs.
  - Delaying adjustment by 5 years yields a marginally lower perpetuity-NOPD but much slower accumulation of assets.
  - The results are sensitive to long-term parameter assumptions (growth, oil prices, etc.); Tables 2a and 2b show NOPD and projected net financial assets in 2056 under different combinations of long-run non-oil GDP growth and the oil price for the constant perpetuity NOPD option.

### F. Box 1 — Baseline long-term assumptions (explicit parameters)
- Oil reserves: Kazakhstan’s proven oil reserves were assumed at 30 billion barrels (Source: BP Statistical Review of World Energy, June 2015).
- Oil extraction: Daily production is expected to reach 2.12 million barrels by 2021, up from 1.64 million barrels in 2016. Thereafter, oil output is projected flat until the reserves are exhausted.
- Oil price: Oil price forecast through 2021 comes from the latest IMF World Economic Outlook. Thereafter, the oil price remains unchanged in constant U.S. dollar terms, implying 2 percent annual growth in nominal terms in line with long-term U.S. inflation.
- Government net financial assets: defined as the difference between NFRK assets and the stock of government debt; they stood at 22 percent of GDP at end-2016.
- Long-term macroeconomic parameters:
  - Long-term non-oil GDP growth is assumed at 4 percent.
  - Domestic inflation is set at the NBK’s long-term target of 4 percent.
  - Long-term real interest rate is assumed at 4.5 percent (consistent with the historic average).
  - Long-term nominal interest rate assumed to be the sum of the real interest rate and inflation.
  - The same interest rate is assumed on assets and liabilities for simplicity.
  - The tenge is projected to appreciate by 0.25 percent in real terms annually against the U.S. dollar (Balassa-Samuelson effect).
  - Historical trend real appreciation of the tenge against the dollar was 2 percent annually during 1995-2016 (period includes recovery from the collapse of the U.S.S.R. and the oil boom of 2000s).

### G. Price-based structural balance rules (policy instrument)
- Price-based rules target a structural primary balance calculated at a predetermined benchmark oil price.
- Rationale:
  - Help delink primary spending from the price-cyclical component of oil revenue.
  - Help ensure oil price volatility does not contribute to cyclical swings of fiscal policy.
- Benchmark price construction:
  - Usually based on long-term historical and future prices or their combination.
- Contrast with PIH:
  - Price-based rules have a short- to medium-term focus and can provide a near-term anchor for fiscal policy.
  - The PIH model is oriented to long-term fiscal sustainability and intergenerational considerations.

### H. Policy conclusions and recommended approach
- Fiscal policy should:
  - Promote macro-fiscal stability by cushioning the economy from short-term volatility in oil revenues.
  - Foster long-term fiscal sustainability consistent with saving for future generations, given exhaustibility of oil resources.
- Assessment of the new NFRK non-oil deficit path:
  - The non-oil deficit path introduced in the new NFRK concept is broadly consistent with the proposed approach.
  - Timely and decisive implementation of revenue and expenditure measures is needed to follow the new path.
- Overall recommendation:
  - Price-based fiscal rules—with structural fiscal balance targets guided by long-term benchmarks—can help anchor short- and medium-term fiscal policy.
  - Medium-term consolidation is needed to reduce the risk of a forced and disruptive fiscal adjustment given the sustained oil-price shock.

### Oil price benchmarks and rule mechanics (detailed)
- Benchmark 1: “5.1.5” rule — benchmark oil price = 11-year average that includes the past 5 years, the current year, and 5 years of the future price.
- Benchmark 2: “RUS rule” — benchmark oil price = the lower of a 10-year historical average and a 3-year historical average.
  - The 10-year average aims to smooth the benchmark price, and therefore government spending.
  - The 3-year average helps avoid excessive deficits in the event of sustained oil price drop in preceding years.
  - The minimum function (lower of the two averages) forces slower adjustment to price increases and faster adjustment to price decreases, building in a bias toward savings.
- Note: Price-based rules may not delink primary spending from volatility in oil revenue arising from changes in oil extraction; a structural balance rule would need to be based on predetermined benchmark oil price and output if oil production is not smooth.

### Trade-offs between expenditure volatility and financial savings
- Using a long moving average of historical oil prices:
  - Achieves the greatest degree of expenditure smoothing.
  - May lead to insufficient savings or dis-saving if the oil price drops sharply.
- Using a combination of past and future prices (e.g., “5.1.5”) or a short moving average of past prices:
  - Generates higher spending volatility.
  - Generates higher savings due to faster adjustment of the benchmark to price drops.
- Numeric illustrations from policy simulations:
  - Under the constant perpetuity benchmark (non-oil deficit in percent of non-oil GDP), net financial assets of the government would be about 200 percent of GDP by the end of the projection period.
  - A more gradual adjustment to this rule would yield financial assets of about 150 percent of GDP by 2056 (Figure 6b).
  - A balanced structural balance rule (targeting zero structural balance) would roughly maintain the current level of net financial assets, failing to save for future generations, and providing less accumulation than the relaxed PIH option.
  - A structural balance rule targeting a surplus of 2½-3 percent of non-oil GDP (under either price benchmark) would generate savings in the 150-170 percent of GDP range in the long term (Figure 8b).

### Medium-term adjustment implications
- Price-based structural balance rules entail a large adjustment in the medium term:
  - In 2017, the NOPD projected under the staff’s baseline is only slightly higher (excluding support to the banking sector) than the NOPD resulting from price-based rules targeting 2-3 percent of non-oil GDP structural surplus.
  - Going forward, baseline NOPD remains broadly constant while rules targeting a structural surplus of 2-3 percent of non-oil GDP require a considerable reduction in NOPD as the benchmark price adjusts to the oil price drop since 2014.
  - Over time, this divergence produces a striking difference in net financial assets, underlining the case for commitment to sizeable medium-term fiscal consolidation.
- The medium-term adjustment implied by price-based rules would entail a reduction of the NOPD of about 4-5 percent of non-oil GDP over the medium term.

### Evaluation of the new NFRK concept
- The non-oil deficit path in the new NFRK concept adopted in late 2016:
  - Is broadly in line with the path suggested by the oil price-based rules.
  - Follows relatively closely the path implied by the Russian rule with zero balance (Figure 9).
  - However, more consolidation would be needed to accumulate more sizable NFRK assets.
- Quantitative comparison:
  - The Russian rule with zero balance would result in net financial assets of about 46 percent of GDP by the end of the projection period — well below the 150-170 percent of GDP range based on perpetuity PIH benchmarks.
- Implementation risk:
  - Lack of revenue and expenditure measures and deviations from the NFRK concept path would result in a decline in net financial assets and could impair long-term fiscal sustainability.
- Policy implication:
  - Stronger consolidation efforts (as in the Russian experience) would allow for higher accumulation of net financial assets and preserve a greater share of oil wealth for future generations.

### Oil revenue shocks: scenarios and sensitivity
- Shocks analyzed: one- and two-standard deviation shocks in oil prices; one standard deviation corresponds to about $16.
- PIH and RUS rule sensitivity:
  - PIH-perpetuity is relatively inelastic to changes in oil prices.
  - PIH-annuity is highly sensitive because the annuity assumes a finite life-span of the NFRK which is directly affected by oil revenues.
  - The non-oil fiscal balance path implied by the RUS rule with zero balance is directly affected by oil price shocks.
- Selected quantified outcomes (non-oil balance implications):
  - Negative 2σ Shock: Perpetuity -0.52; Annuity -2.97
  - Negative 1σ Shock: Perpetuity -0.77; Annuity -4.4
  - Baseline: Perpetuity -1.02; Annuity -5.83
  - Positive 1σ Shock: Perpetuity -1.27; Annuity -7.25
  - Positive 2σ Shock: Perpetuity -1.51; Annuity -8.68
- Policy implication from shocks:
  - In the most pessimistic scenario—an oil price shock of 2 standard deviations—the rule suggests a non-oil deficit of about 3 percent of non-oil GDP, consistent with the non-oil balance suggested by the PIH-Annuity.
  - In the case of a positive shock, the rule permits a higher non-oil deficit in the medium- and long-term.

### Conclusions and policy recommendations (fiscal)
- The sharp and sustained drop in oil prices since 2014 calls for a medium-term fiscal consolidation.
  - With a broadly unchanged non-oil deficit going forward, the government’s net financial assets would fall close to zero by 2021 (see Table 1), risking market access and fiscal sustainability.
- Recommended fiscal objectives:
  - Authorities should reduce the non-oil deficit closer to levels consistent with long-term fiscal benchmarks.
  - The medium-term adjustment could be anchored by price-based fiscal rules targeting a structural surplus of 2-3 percent of non-oil GDP.
  - Adopting these rules would entail a reduction of the NOPD of about 4-5 percent of non-oil GDP over the medium term.
- On the NFRK concept:
  - The new NFRK concept introduces targets that preserve long-term fiscal sustainability; implementation of the implied adjustment is paramount.
  - The non-oil fiscal deficit targeted by the new concept is in line with the Russian rule with zero balance and results in positive accumulation of net financial assets in the medium-long term.
  - Deviation from this path will result in higher deficits and could jeopardize long-term fiscal sustainability.

### Monetary policy: equilibrium real rate estimation and recommendations
- Observability and estimation:
  - Neither the natural rate (equilibrium real interest rate), nor potential output are directly observable; estimation approaches include time series techniques and micro-founded models.
- HP filter results for Kazakhstan (Q1 1996- Q2 2016):
  - Real interest rate = policy rate − expected inflation.
  - Policy rate = NBK refinancing rate until August 2015 and base rate (overnight repo rate) thereafter.
  - Expected inflation approximated by a four-quarter moving average of past inflation.
  - HP filter estimates suggest the equilibrium real interest rate has declined significantly since the late 1990s to around 4 percent on average in the last year.
- State-space (Laubach-Williams style) estimates:
  - Equilibrium real rate represented as r_t^* = g_t + z_t with potential output and trend components modeled as random walks.
  - Practical estimation adjustment: estimated ζ_z for Kazakhstan fixed at 0.03 in final estimation.
  - Key finding: current natural rate of interest on the order of 4 percent; steady decline since 2007.
- Robustness and measurement uncertainty:
  - Estimated output gap and equilibrium real interest rate are subject to considerable uncertainty.
  - Taylor (2016) example: perceptions about changes in the equilibrium rate can have consequences as large as 2 percentage points (U.S. example).
- Policy recommendations for NBK:
  - Consider robust rules and evaluate outcomes of several alternative rules given uncertainty.
  - Strengthen analytical capacity to conduct formal assessments as more information accumulates.
  - Use the NBK survey of inflation expectations to evaluate forward-looking rules.
- Strengthening interest rate transmission channel:
  - Documented weak link between policy rate, market interest rates and inflation; further develop the interest channel.
  - NBK has started issuing securities at various maturities to help build the yield curve; Ministry of Finance should cooperate to develop the domestic bond market.
  - Consider faster unwinding of long-term FX swaps undertaken with local banks in 2014-15.
  - Phase out direct instruments of monetary control (e.g., caps on annual growth of consumer credit, subsidized lending); if needed, these should be fiscal policy instruments executed by the MoF.
  - Pursue macro and prudential policies aimed at de-dollarizing the economy.

### Exchange rate pass-through (ERPT) and inflation dynamics
- Key empirical findings (January 2000 - June 2016 monthly sample):
  - Pass-through to consumer prices estimated in the range of 20- 30 percent.
  - Contemporaneous correlations between changes in log CPI and log NEER (contemporaneous and first three lags): 0.27, 0.46, 0.37 and 0.15.
  - Single-equation estimates:
    - Short-term elasticity of inflation to exchange rate depreciation about 2 percent (point estimate) and not statistically significant.
    - Sum of all coefficients on the exchange rate about 0.155 in Model A and 0.141 in Model B.
    - Coefficients on the first and second lag of the exchange rate are highly significant, implying at least a month lag to affect consumer prices.
  - VAR-in-differences (preferred tool) results:
    - Contemporaneous response to exchange rate changes to CPI about 2 percent.
    - Accumulated effect about 25 percent after 3 months.
    - Accumulated effect about 30 percent after 6 months; stabilizes around 27 percent in the long run.
    - For import prices: ERPT after 3 months is 30 percent; ERPT after 6 months is 55 percent (also long-term effect).
    - A 10 percent depreciation of the tenge expected to trigger an increase in consumer prices on the order of 3 percent.
- Robustness checks:
  - Replacing NEER with the KZT/USD exchange rate and re-estimating yields an ERPT of 28 percent in the long run.
  - Applying baseline specification to import prices yields contemporaneous coefficient 0.065 and sum of coefficients 0.247.
  - Results broadly unchanged when varying lag structure, output measures, and seasonal adjustments.
- Policy implications:
  - Timing and magnitude of ERPT matter for inflation-targeting decisions.
  - Credible monetary policy frameworks anchor expectations and tend to be associated with lower pass-through.
  - Reduce dollarization and increase competition in domestic product markets to lower ERPT.
  - Incorporate ERPT information into NBK policy decisions and communicate anticipated effects clearly.

### Potential output and growth accounting
- Definition and objective:
  - Potential output: level of output attainable when all factors of production fully utilized.
  - Objective: quantify potential output using univariate filters, a multivariate filter, and a production function.
- Methods and data:
  - Univariate filters: HP, Baxter-King (BK), Christiano-Fitzgerald (CF); quarterly data back to 1999 with staff projections to 2022.
  - Multivariate filter: Blagrave et al. (2015) method using GDP, inflation, and unemployment; annual data 1996-2016 used.
  - Production function: Cobb-Douglas with country-specific elasticities estimated from firm-level Orbis data (sample 993 firms ~14 percent of gross value added).
    - Estimated coefficients approximately 0.56 for labor and 0.44 for capital.
    - Human capital returns r set at .107; depreciation δ assumed 0.07.
- Results — trend growth:
  - All approaches indicate trend growth has fallen to around 3 percent in recent years.
  - Univariate filters: trend growth peaked ~10 percent in 2004; fell to around 2.5 percent in 2016; projected to rise to 3.5-4 percent by 2022.
  - Multivariate filter: estimated potential growth in 2016 around 3.3 percent.
  - Production function: estimated potential growth rate for 2016 is 3.0 percent.
- Growth accounting contributions:
  - Labor and human capital combined contributed about 2.2 percentage points to growth in the last five years.
  - Capital accumulation contributed over 6 percentage points on average (though lower than pre-global financial crisis).
  - TFP declined gradually and was sharply negative in years of negative oil price shocks.
- Sectoral and policy implications:
  - Potential growth is closely linked to the oil sector; dependence on oil causes high volatility.
  - Structural reforms and a rebound in global oil prices could raise TFP and capital contributions.
  - Policy recommendation: implement structural reforms to reduce growth volatility and make 4 percent growth over the medium term attainable, conditional on reforms and external conditions.

*Prepared by: Matteo Ghilardi, Jonah Rosenthal, Rossen Rozenov, Azim Sadikov (MCD); Republic of Kazakhstan — IMF staff analysis (content unit cr17109).*

### REFERENCES_____________________________________________________________________________17

### REFERENCES

### A. Fiscal context and recent developments
- Oil accounts for more than 50 percent of exports.
- Government oil-related revenues averaged 10-12 percent of GDP over the last decade—more than 40 percent of total revenues.
- The overall fiscal position reversed from an average surplus of 4½ percent of GDP in 2011-13 to a deficit of 5 percent of GDP 2015-16.
- Oil revenues fell by more than 70 percent in dollar terms from 2013 to 2016; they declined by just half as much in tenge terms due to the authorities’ decision to allow the tenge to depreciate.
- NFRK assets stood at $60 billion at end-2016, equivalent to 43 percent of GDP.
- Public debt stood at 21 percent of GDP and total external debt at 122 percent of GDP.

### B. National Fund of the Republic of Kazakhstan (NFRK) framework and recent changes
- NFRK established in 2000; receives all oil-related revenues including: (i) the government’s share of profits in oil projects; (ii) receipts from the export duty tax; and (iii) various mineral extraction taxes and fees.
- Government’s share of oil revenues historically averaged 30-35 percent.
- Under the current framework (pre-2018), guaranteed transfers are set at $8 billion (±15% depending on the cyclical position of the economy and the financing needs); targeted transfers are discretionary.
- New NFRK framework (effective from 2018):
  - Guaranteed transfers will decline gradually to about KZT 2 trillion (about $6 billion at the current exchange rate) by 2020.
  - NFRK assets cannot fall below 30 percent of GDP.
  - The budget will target a specific and pre-determined path for the non-oil deficit in the medium-term.
  - The new framework aims to reduce the non-oil deficit from 8.3 percent of GDP in 2016 to 7 percent of GDP in 2020 and to 6 percent of GDP in 2025.

### C. Risks under an unchanged policy (sustainability concerns)
- Under a passive policy scenario with a broadly unchanged non-oil deficit going forward:
  - Government’s net financial assets (NFRK assets minus government debt) would fall close to zero by 2021 (Figures 4-5 referenced).
  - Keeping the debt-to-GDP ratio constant while covering government financing needs largely from the NFRK would exhaust NFRK assets by the mid-2020s.
- Table 1 presents the year when net financial assets would turn negative under various combinations of the long-term oil price and the (constant) non-oil primary deficit (table referenced; values in table not reproduced here).

### D. Long-term benchmarks and methodology
- Benchmarks derive from estimating Kazakhstan’s net wealth = initial stock of net financial assets + present value to the government of oil assets in the ground.
- Key uncertainties in the oil-wealth valuation: reserves and extraction rates, future oil prices, exchange rate, government’s share of oil revenues.
- Two implementations of the Permanent Income Hypothesis (PIH) used:
  - Perpetuity: transfer to the budget equals revenues from interest receipts on the stock of NFRK assets.
  - Annuity: assumes a lifespan of the NFRK of forty years.
- Four scenarios considered:
  - (i) a perpetuity-non-oil primary deficit (NOPD) constant in percent of non-oil GDP starting from 2017;
  - (ii) delaying the adjustment to option (i) by 5 years (constant perpetuity-NOPD starting from 2022);
  - (iii) a constant real perpetuity-NOPD;
  - (iv) a 40-year annuity-NOPD constant in percent of non-oil GDP.
- Figures 6a-b present NOPD paths and government net financial assets under the various cases and are compared to staff’s medium-term baseline projections.

### E. PIH results and implications
- PIH main findings:
  - A constant perpetuity-NOPD as a share of GDP of 1.1 percent would maximize Kazakhstan’s net wealth at end of the forecasting horizon.
  - Targeting a constant NOPD as a share of GDP of 6.2 percent over 40 years would be consistent with net wealth as of end-2016 under the annuity approach; however, under this annuity approach (40-year), net wealth would be fully exhausted by the end of the 40-year annuity period.
  - If the NOPD is constant in real terms over the 40-year horizon, the 40-year NOPD annuity would average 10 percent of GDP in the first decade before declining steadily (not shown in figures).
- Considerations for choosing among approaches:
  - The perpetuity approach (1.1 percent NOPD) could be considered too tight for an emerging market economy with social and investment needs.
  - Delaying adjustment by 5 years yields a marginally lower perpetuity-NOPD but much slower accumulation of assets.
  - The results are sensitive to long-term parameter assumptions (growth, oil prices, etc.); Tables 2a and 2b show NOPD and projected net financial assets in 2056 under different combinations of long-run non-oil GDP growth and the oil price for the constant perpetuity NOPD option.

### F. Box 1 — Baseline long-term assumptions (explicit parameters)
- Oil reserves: Kazakhstan’s proven oil reserves were assumed at 30 billion barrels (Source: BP Statistical Review of World Energy, June 2015).
- Oil extraction: Daily production is expected to reach 2.12 million barrels by 2021, up from 1.64 million barrels in 2016. Thereafter, oil output is projected flat until the reserves are exhausted.
- Oil price: Oil price forecast through 2021 comes from the latest IMF World Economic Outlook. Thereafter, the oil price remains unchanged in constant U.S. dollar terms, implying 2 percent annual growth in nominal terms in line with long-term U.S. inflation.
- Government net financial assets: defined as the difference between NFRK assets and the stock of government debt; they stood at 22 percent of GDP at end-2016.
- Long-term macroeconomic parameters:
  - Long-term non-oil GDP growth is assumed at 4 percent.
  - Domestic inflation is set at the NBK’s long-term target of 4 percent.
  - Long-term real interest rate is assumed at 4.5 percent (consistent with the historic average).
  - Long-term nominal interest rate assumed to be the sum of the real interest rate and inflation.
  - The same interest rate is assumed on assets and liabilities for simplicity.
  - The tenge is projected to appreciate by 0.25 percent in real terms annually against the U.S. dollar (Balassa-Samuelson effect).
  - Historical trend real appreciation of the tenge against the dollar was 2 percent annually during 1995-2016 (period includes recovery from the collapse of the U.S.S.R. and the oil boom of 2000s).

### G. Price-based structural balance rules (policy instrument)
- Price-based rules target a structural primary balance calculated at a predetermined benchmark oil price.
- Rationale:
  - Help delink primary spending from the price-cyclical component of oil revenue.
  - Help ensure oil price volatility does not contribute to cyclical swings of fiscal policy.
- Benchmark price construction:
  - Usually based on long-term historical and future prices or their combination.
- Contrast with PIH:
  - Price-based rules have a short- to medium-term focus and can provide a near-term anchor for fiscal policy.
  - The PIH model is oriented to long-term fiscal sustainability and intergenerational considerations.

### H. Policy conclusions and recommended approach
- Fiscal policy should:
  - Promote macro-fiscal stability by cushioning the economy from short-term volatility in oil revenues.
  - Foster long-term fiscal sustainability consistent with saving for future generations, given exhaustibility of oil resources.
- Assessment of the new NFRK non-oil deficit path:
  - The non-oil deficit path introduced in the new NFRK concept is broadly consistent with the proposed approach.
  - Timely and decisive implementation of revenue and expenditure measures is needed to follow the new path.
- Overall recommendation:
  - Price-based fiscal rules—with structural fiscal balance targets guided by long-term benchmarks—can help anchor short- and medium-term fiscal policy.
  - Medium-term consolidation is needed to reduce the risk of a forced and disruptive fiscal adjustment given the sustained oil-price shock.

*Prepared by: Matteo Ghilardi, Jonah Rosenthal, Rossen Rozenov, Azim Sadikov (MCD); Republic of Kazakhstan — IMF staff analysis (content unit cr17109).*

### 15. Two different oil price benchmarks are considered. In the first rule, the benchmark oil

### 15. Two different oil price benchmarks are considered. In the first rule, the benchmark oil price is set to an 11-year average of the oil price that includes the past 5 years, the current year, and 5 years of the future price (“5.1.5” rule in the figures). In the second rule, the benchmark price is set to the lower of a 10-year historical average and a 3-year historical average of the oil price. This is the formula followed by Russia (“RUS rule” in the figures).

### Oil price benchmarks and rule mechanics
- Benchmark 1: “5.1.5” rule — benchmark oil price = 11-year average that includes the past 5 years, the current year, and 5 years of the future price.
- Benchmark 2: “RUS rule” — benchmark oil price = the lower of a 10-year historical average and a 3-year historical average.
  - The 10-year average aims to smooth the benchmark price, and therefore government spending.
  - The 3-year average helps avoid excessive deficits in the event of sustained oil price drop in preceding years.
  - The minimum function (lower of the two averages) forces slower adjustment to price increases and faster adjustment to price decreases, building in a bias toward savings.
- Price-based rules may not delink primary spending from volatility in oil revenue arising from changes in oil extraction; a structural balance rule would need to be based on predetermined benchmark oil price and output if oil production is not smooth.

### Trade-offs between expenditure volatility and financial savings
- Using a long moving average of historical oil prices:
  - Achieves the greatest degree of expenditure smoothing.
  - May lead to insufficient savings or dis-saving if the oil price drops sharply.
- Using a combination of past and future prices (e.g., “5.1.5”) or a short moving average of past prices:
  - Generates higher spending volatility.
  - Generates higher savings due to faster adjustment of the benchmark to price drops.
- Numeric illustrations from policy simulations:
  - Under the constant perpetuity benchmark (non-oil deficit in percent of non-oil GDP), net financial assets of the government would be about 200 percent of GDP by the end of the projection period.
  - A more gradual adjustment to this rule would yield financial assets of about 150 percent of GDP by 2056 (Figure 6b).
  - A balanced structural balance rule (targeting zero structural balance) would roughly maintain the current level of net financial assets, failing to save for future generations, and providing less accumulation than the relaxed PIH option.
  - A structural balance rule targeting a surplus of 2½-3 percent of non-oil GDP (under either price benchmark) would generate savings in the 150-170 percent of GDP range in the long term (Figure 8b).

### Medium-term adjustment implications
- Price-based structural balance rules entail a large adjustment in the medium term:
  - In 2017, the NOPD projected under the staff’s baseline is only slightly higher (excluding support to the banking sector) than the NOPD resulting from price-based rules targeting 2-3 percent of non-oil GDP structural surplus.
  - Going forward, baseline NOPD remains broadly constant while rules targeting a structural surplus of 2-3 percent of non-oil GDP require a considerable reduction in NOPD as the benchmark price adjusts to the oil price drop since 2014.
  - Over time, this divergence produces a striking difference in net financial assets, underlining the case for commitment to sizeable medium-term fiscal consolidation.
- The medium-term adjustment implied by price-based rules would entail a reduction of the NOPD of about 4-5 percent of non-oil GDP over the medium term.

### Evaluation of the new NFRK concept
- The non-oil deficit path in the new NFRK concept adopted in late 2016:
  - Is broadly in line with the path suggested by the oil price-based rules.
  - Follows relatively closely the path implied by the Russian rule with zero balance (Figure 9).
  - However, more consolidation would be needed to accumulate more sizable NFRK assets.
- Quantitative comparison:
  - The Russian rule with zero balance would result in net financial assets of about 46 percent of GDP by the end of the projection period — well below the 150-170 percent of GDP range based on perpetuity PIH benchmarks.
- Implementation risk:
  - Lack of revenue and expenditure measures and deviations from the NFRK concept path would result in a decline in net financial assets and could impair long-term fiscal sustainability.

### Oil revenue shocks: scenarios and sensitivity
- Shocks analyzed: one- and two-standard deviation shocks in oil prices; one standard deviation corresponds to about $16.
- Approach: compute the PIH and the RUS rule with zero balance under positive and negative shocks to create confidence bands to guide fiscal policy.
- Sensitivity:
  - PIH-perpetuity is relatively inelastic to changes in oil prices.
  - PIH-annuity is highly sensitive because the annuity assumes a finite life-span of the NFRK which is directly affected by oil revenues.
  - The non-oil fiscal balance path implied by the RUS rule with zero balance is directly affected by oil price shocks.
- Selected quantified outcomes (non-oil balance implications, values presented in the source):
  - Negative 2σ Shock: Perpetuity -0.52; Annuity -2.97
  - Negative 1σ Shock: Perpetuity -0.77; Annuity -4.4
  - Baseline: Perpetuity -1.02; Annuity -5.83
  - Positive 1σ Shock: Perpetuity -1.27; Annuity -7.25
  - Positive 2σ Shock: Perpetuity -1.51; Annuity -8.68
- Policy implication from shocks:
  - In the most pessimistic scenario—an oil price shock of 2 standard deviations—the rule suggests a non-oil deficit of about 3 percent of non-oil GDP, consistent with the non-oil balance suggested by the PIH-Annuity.
  - In the case of a positive shock, the rule permits a higher non-oil deficit in the medium- and long-term.

### Conclusions and policy recommendations
- The sharp and sustained drop in oil prices since 2014 calls for a medium-term fiscal consolidation.
  - With a broadly unchanged non-oil deficit going forward, the government’s net financial assets would fall close to zero by 2021 (see Table 1), risking market access and fiscal sustainability.
- Recommended fiscal objectives:
  - Authorities should reduce the non-oil deficit closer to levels consistent with long-term fiscal benchmarks.
  - The medium-term adjustment could be anchored by price-based fiscal rules targeting a structural surplus of 2-3 percent of non-oil GDP.
  - Adopting these rules would entail a reduction of the NOPD of about 4-5 percent of non-oil GDP over the medium term.
- On the NFRK concept:
  - The new NFRK concept introduces targets that preserve long-term fiscal sustainability; implementation of the implied adjustment is paramount.
  - The non-oil fiscal deficit targeted by the new concept is in line with the Russian rule with zero balance and results in positive accumulation of net financial assets in the medium-long term.
  - Deviation from this path will result in higher deficits and could jeopardize long-term fiscal sustainability.
  - Stronger consolidation efforts (as in the Russian experience) would allow for higher accumulation of net financial assets and preserve a greater share of oil wealth for future generations.

*IMF staff analysis as presented in the source content.*

### 6. The above rules assume knowledge of the equilibrium real interest rate (or the

### cr17109 - 6. The above rules assume knowledge of the equilibrium real interest rate (or the

### Observability and estimation approaches
- Neither the natural rate (equilibrium real interest rate), nor potential output are directly observable; their estimation is fraught with difficulties.
- Two general approaches used in the literature to estimate the equilibrium real interest rate:
  - Time series techniques (e.g., Hamilton and others, 2016).
  - Micro-founded models incorporating optimizing behavior of households and firms (Justiniano and Primiceri, 2010).
- Example univariate time series method: Hodrick-Prescott (HP) filter – decomposes the actual time series into a trend component and a cyclical component; the trend component can be thought of as a measure of the equilibrium rate.

### HP filter results for Kazakhstan
- Data and construction:
  - Applied to a calculated quarterly real interest rate for Kazakhstan in the period Q1 1996- Q2 2016.
  - Real interest rate = policy rate − expected inflation.
  - Policy rate = NBK refinancing rate until August 2015 and base rate (overnight repo rate) thereafter.
  - Expected inflation approximated by a four-quarter moving average of past inflation.
- Key findings:
  - The HP filter estimates suggest that the equilibrium real interest rate has declined significantly since the late 1990s to around 4 percent on average in the last year.
  - The decline likely reflects abundant liquidity in the economy.
  - Caveat (Taylor, 2016): the observed downward trend in the equilibrium rate may be due to a trend in other variables that affect the economy.
- Empirical context noted in literature:
  - Average global ten-year real rate declined from 6 percent to close to zero.
  - For some economies, such as the core euro area countries, real rates have become negative (IMF, 2014a).

### State-space methods and Laubach-Williams style joint estimation
- Rationale:
  - Natural rate and potential output can be examined jointly using state space methods (Kalman filter) to model dynamics of unobservable variables.
- Model structure (as presented):
  - Equilibrium real rate represented as a sum of two components: r_t^* = g_t + z_t, where g_t is the trend growth rate of potential output and z_t captures other determinants of r_t^*.
  - Potential output (log) modeled as a random walk with stochastic drift: y_t^* = y_{t-1}^* + g_{t-1} + ε_t^{*}.
  - Trend g_t is a random walk: g_t = g_{t-1} + η_{g,t}.
  - z_t follows: z_t = z_{t-1} + η_{z,t}.
  - Reduced-form IS and inflation equations included (see original model equations for full specification).
  - Output gap: x_t = 100(y_t − y_t^*).
  - Real interest rate r_t computed as in HP section.
- Estimation details for Kazakhstan:
  - State space model estimated for Kazakhstan.
  - Real GDP in 2005 prices available only since 1999; real growth rates based on an older series (1994 prices) used to extend data backwards.
  - Estimation done in three stages following Holston and others (2016); ratios ζ_g = σ_g/σ_{ε^*} and ζ_z = σ_z/σ_{ε^*} obtained using the median unbiased estimator of Stock and Watson (1999) and then imposed when estimating remaining parameters.
  - Practical adjustment: estimated ζ_z for Kazakhstan turned out very large (about ten times higher than typical values for advanced economies) and prevented calculations in the final stage, so it was fixed at 0.03 (a value in line with Holston and others, (2016)).
- Key findings from state-space estimates:
  - Results for Kazakhstan suggest a current natural rate of interest on the order of 4 percent.
  - Since 2007 there has been a steady decline in the natural rate; the most recent level broadly corresponds to the estimate based on the HP filter.
  - The sharp decline in 1998 is associated with the large negative output gap at the time.

### Considerations for robustness and measurement uncertainty
- The estimated output gap and equilibrium real interest rate are subject to considerable uncertainty; this is a general problem not unique to Kazakhstan.
- Measurement errors can materially affect monetary policy outcomes:
  - Orphanides and others (2000) find that mismeasurements of the output gap can result in a substantial deterioration in economic outcomes.
  - Taylor (2016) shows that perceptions about how the equilibrium rate has changed recently can have large consequences for interest rate setting; in the case of US, the difference can be as large as 2 percentage points.
  - Taylor argues the natural rate in the policy rule should not be adjusted in a discretionary way and should be considered with shifts in related variables (e.g., potential GDP) to ensure consistency.
  - Hamilton and others (2016) recommend using policy inertia in interest rate rules (assigning a positive value to parameter ρ in equation (2)) as a remedy.
- Robust policy rule design:
  - Robustness evaluated historically by comparing rule performance across a set of models (Levin and others, 1999; Taylor, 1999) and more recently using decision-theoretic approaches (e.g., Onatski and Williams, 2003).
  - General findings:
    - Placing higher weights on inflation makes a rule more robust to inflation shocks; similarly for output.
    - Including a lag in the interest rate does not necessarily improve performance and can lead to instability in some models.
    - Rules that are robust to one type of uncertainty may fail under a different form of uncertainty.
    - Cogley and others (2011) find an optimal simple Taylor rule that accounts for parameter uncertainty.
- Alternative rule specifications to mitigate measurement error:
  - Difference rules (setting ρ = 1 in (2)) link change in the interest rate to deviations of inflation from target and actual output from potential; Levin and others (1999) argue difference rules perform better than level-based rules in their models and are more robust to uncertainty.
  - Replace the output gap with the growth rate of output (McCallum, 1998) to reduce measurement error sensitivity.
  - In unemployment-based rules, include change in unemployment rate instead of the unemployment gap (Orphanides and Williams, 2002) recognizing imperfect substitution but appropriate directionality.

### Conclusions and policy recommendations for Kazakhstan (NBK)
- Role and benefits of interest rate rules:
  - Interest rate rules provide a quantitative benchmark for the policy variable consistent with stabilizing the economy.
  - They structure thinking about monetary policy stance and enhance transparency and independence.
- Role of discretion:
  - Policies prescribed by rules should not be followed rigidly. Discretion remains important because:
    - The nature of some shocks may make the rule response inadequate.
    - Output gap and natural rate estimates are hard to estimate and highly uncertain.
- Specific findings and recommendations for NBK:
  - Joint estimation suggests a current natural rate of interest of around 4 percent; it has been on a declining trend over the last ten years.
  - Given uncertainties, NBK should consider robust rules and evaluate outcomes of several alternative rules, analyzing sources of differences to inform rule choice.
  - Strengthen analytical capacity to conduct formal assessments as more information accumulates (NBK experience with inflation targeting is still in initial phase).
  - Use the NBK survey of inflation expectations to evaluate forward-looking rules.
- Strengthening the interest rate transmission channel (policy actions suggested):
  - No rule can achieve price stability if changes in the policy rate have no impact on inflation; Kazakhstan has documented weak link between policy rate, market interest rates and inflation.
  - Further develop the interest channel; improvements noted since introduction of IT but more needed.
  - NBK has started issuing securities at various maturities to help build the yield curve; government (Ministry of Finance) should play a strong role in developing the domestic bond market—requires cooperation between NBK and MoF.
  - Consider faster unwinding of the long-term FX swaps undertaken with local banks in 2014-15.
  - Phase out use of direct instruments of monetary control (e.g., caps on annual growth of consumer credit, subsidized lending); if needed, these should be fiscal policy executed by the MoF through the budget as they are not consistent with the IT framework.
  - Pursue macro and prudential policies aimed at de-dollarizing the economy in line with international best practices given the relatively high degree of dollarization.
  - Address issues in the financial sector to maintain macroeconomic stability.

- Key statistics and data references preserved from the source:
  - Period analyzed (HP filter and state-space): Q1 1996- Q2 2016.
  - Average global ten-year real rate declined from 6 percent to close to zero.
  - For Kazakhstan: equilibrium (natural) rate on the order of 4 percent (current estimate); HP filter indicates around 4 percent on average in the last year.
  - Practical estimation adjustment: fixed ζ_z at 0.03 in state-space estimation for Kazakhstan.
  - Taylor (2016) — difference in perceptions for US can be as large as 2 percentage points.

*Source: IMF staff; Republic of Kazakhstan chapter (cr17109).*

### 6.  Chile

### 6.  Chile

### Key findings on exchange rate pass-through (ERPT) and inflation dynamics
- The pass-through to consumer prices in Kazakhstan is estimated in the range of 20- 30 percent.  
- The National Bank of Kazakhstan (NBK) moved toward inflation targeting (IT); on August 20, 2015, the NBK announced a transition to IT and greater exchange rate flexibility.  
- At the close of the morning session of the Kazakhstan Stock Exchange (KASE) on August 20, the tenge traded at 255 per US dollar, up from KZT 188 per USD a day earlier. The exchange rate later stabilized at around KZT 340 per USD.  
- Consumer price inflation reached 13.6 percent at the end of 2015 (compared to 4.4 percent in September 2015) and peaked at 17.7 in July 2016.  
- Monthly data sample used: January 2000 - June 2016.  
- Correlations between changes in log CPI and log NEER (contemporaneous and first three lags) for the whole sample are 0.27, 0.46, 0.37 and 0.15, respectively.  
- Single-equation estimates:
  - Short-term elasticity of inflation to exchange rate depreciation is about 2 percent (point estimate) and is not statistically significant.
  - Sum of all coefficients on the exchange rate is about 0.155 in Model A and 0.141 in Model B.
  - Coefficients on the first and second lag of the exchange rate are highly significant, implying it takes at least a month for exchange rate changes to start affecting consumer prices.
- Tests for asymmetry and non-linearities (including an “IT” binary variable from August 2015, a “Depreciation” binary variable, and a “Large depreciation” interaction for monthly depreciations above thresholds) show that none of these additional variables has a statistically significant impact on inflation in the models reported.

### Analytical context and channels
- ERPT matters for monetary policy because exchange rate movements are an essential ingredient in inflation forecasts under IT, especially in small open economies.
- ERPT affects the effectiveness of monetary policy via the nominal exchange rate’s role as a shock absorber and through expenditure-switching effects; differences in pass-through to tradables versus non-tradables can alter policy effectiveness.
- Transmission chain described in two stages:
  - Stage 1: exchange rate fluctuations reflected in import prices.
  - Stage 2: import price changes transmitted to consumer prices directly (import component of CPI) or indirectly (via producer prices).
- Pricing strategies influence ERPT:
  - Producer currency pricing (PCP) tends to produce more complete ERPT to import prices.
  - Local currency pricing (LCP) often yields incomplete ERPT to domestic prices.
- Other factors likely to affect ERPT include degree of competition/market segmentation, substitutability between domestic and imported goods, and the level of dollarization.

### Empirical approach and data properties
- Two complementary methodological approaches used: single-equation specifications and vector-based methods (VARs/VECMs); sensitivity of estimates checked across methods.
- Unit root testing:
  - Level series for CPI, PPI, import prices, industrial output, NEER, and foreign prices fail to reject unit root.
  - First differences strongly reject unit root, indicating the series are likely I(1).
- Choice of lag structure:
  - For monthly data, number of lags selection trades off capturing delayed effects and degrees of freedom; exchange rate lags fixed at six in the single-equation family of models, while lags of industrial output and foreign prices varied between six and one.
  - Akaike Information Criterion suggests k=6 for output and foreign prices; Bayesian Information Criterion selects k=1.

### Policy implications and recommendations
- Putting a credible monetary policy framework in place—supported by effective communications—may anchor expectations and be associated with lower pass-through.  
- Reducing dollarization could help reduce ERPT.  
- Introducing more competition in domestic product markets could help reduce ERPT.  
- For policy making, estimate ERPT using an array of techniques (single-equation and system models) and compare results, keeping methodological biases in mind.

*Source: IMF staff calculations and analysis as presented in the IMF chapter on exchange rate pass-through (Kazakhstan), January 2000 - June 2016 sample.*

### 18. Robustness checks indicate that estimates are not very sensitive to alternative

### 18. Robustness checks indicate that estimates are not very sensitive to alternative

### Robustness checks and alternative specifications
- Replacing industrial output with an output gap proxy (calculated using the HP filter), as in Mihaljek and Klau (2008), has a marginal impact on the results.
- Applying the baseline specification (Model A) to import prices instead of the CPI yields higher elasticities:
  - The estimated contemporaneous coefficient on the exchange rate is 0.065.
  - The sum of all coefficients on this variable is 0.247, implying a faster and larger pass-through to import prices.
- Results remain broadly unchanged when:
  - Reordering the industrial production variable.
  - Estimating the model with different numbers of lags.
  - Using an output gap variable, or a seasonally unadjusted index of industrial production (see Figure A1 in the Annex).
- Replacing the nominal effective exchange rate with the KZT/USD exchange rate and re-estimating the model with 2 lags produces an ERPT of 28 percent in the long run – very close to the baseline estimate (see Figure A2 in the Annex).
- For import prices specifically:
  - ERPT after 3 months is 30 percent.
  - ERPT after 6 months is 55 percent, which is also the long-term effect.
  - Confidence bands are much wider around these estimates.

### D. System methods — rationale and implementation
- Main advantage: system methods treat variables as endogenous, accounting for potential endogeneity of the exchange rate and price variables.
- Preferred frameworks: cointegrated VAR and VAR in differences.
- Example: Winkelried (2014) documents a decline in ERPT in Peru after adopting a fully-fledged IT regime using a 6-variable SVAR following the price chain; identification based on Cholesky decomposition; VAR in differences yields ERPT from cumulative impulse responses.
- For Kazakhstan, a stationary VAR with recursive identification considers the vector (p*, e, pm, y, ppi, cpi) where variables denote log differences of foreign prices, exchange rate, import prices, industrial output, producer and consumer prices in that order.
  - Recursive identification assumes error term in each equation is uncorrelated with error terms in previous equations.
  - Assumption used: foreign supply, exchange rate and import price shocks have contemporaneous effect on output but demand shocks do not affect the first three variables instantaneously.
  - Results are not sensitive to the position of output (second or third); impulse responses of price variables to an exchange rate shock look very similar.

### VAR lag selection and estimation
- Formal lag selection:
  - Akaike, Hannan-Quinn and Schwarz information criteria suggest only one lag.
  - LR criterion points to five lags.
- Estimating VAR with one lag shows significant residual autocorrelation (violating independence); autocorrelation is removed when five lags are selected.
- Therefore, a VAR(5) model is estimated.

### VAR-in-differences results: ERPT to CPI and import prices
- Pass-through of exchange rate changes to CPI:
  - Contemporaneous response is very small — about 2 percent.
  - Accumulated effect increases to about 25 percent after 3 months.
  - Accumulated effect increases to 30 percent after 6 months.
  - Effect diminishes somewhat after the first six months and stabilizes at around 27 percent in the long run.
  - Summary statement: results suggest that the pass-through of exchange rate changes to CPI is around 30 percent.
- Impulse response timing:
  - Response is significant and peaks 1-2 months after the depreciation.
- Import prices respond more strongly:
  - ERPT after 3 months is 30 percent.
  - ERPT after 6 months is 55 percent (also the long-term effect).
  - Wider confidence bands around import-price estimates.
- Interpretation: higher pass-through to import prices is expected since part of depreciation effect is dampened in the distribution phase.

### Cointegrated VAR: theoretical and empirical challenges
- Cointegration tests and interpretation:
  - Testing for cointegration amounts to testing for a linear combination of I(1) variables that is stationary.
  - If more than one cointegrating vector exists, interpreting coefficients as long-run elasticities is less straightforward; indeterminacy arises because any linear combination of cointegrating vectors is also a cointegrating vector.
  - Beirne and Bijsterbosch (2009) propose taking the first cointegrating vector (highest eigenvalue), but Masten (2004) argues correct identification of equilibrium ERPT often requires additional restrictions.
  - Masten Proposition 1: equilibrium ERPT is identified if and only if the cointegrating rank of the system is equal to 1 plus the number of variables with non-zero coefficients in the vector of long-run responses (based on Johansen (2002)).
- Kazakhstan-specific cointegration findings:
  - Johansen test suggests two cointegrating relationships for the baseline specification (variables as in stationary VAR, 5 lags, linear trend in level data, cointegrating equations stationary around a nonzero mean).
  - First cointegration equation reported as:
    - CPI = 1.277 p_m - 0.515 y + 0.745 NEER + 1.527 p*
      - (text presents: ݅݌ܿ ൌ1.277݉݌െ0.515ݕ൅0.745݁൅1.527݌∗)
  - All coefficients in the first cointegration equation, except the one on the output variable, are significant at the 5 percent level.
  - Estimated coefficient on the nominal effective exchange rate in the cointegrating vector is 0.75, but this does not necessarily measure pass-through without further restrictions.
- Empirical issues with cointegrated VAR:
  - Adjustment coefficients are in most cases close to zero and statistically insignificant, indicating weak exogeneity with respect to cointegrating parameters.
  - Data do not support substantial adjustment of domestic prices to deviations from equilibrium; the only variable that adjusts significantly and consistently is the exchange rate.
  - Results are sensitive to model specification:
    - Replacing seasonally adjusted output with unadjusted (with 5 lags) yields a coefficient of 0.54 on the exchange rate in the first cointegration equation, not statistically significant.
    - Using the USD exchange rate (and 3 lags) produces an estimate of 0.34 on the exchange rate, highly statistically significant — more in line with VAR-in-differences results.

### E. Conclusions and policy implications
- Empirical summary for Kazakhstan:
  - ERPT has a moderate impact on domestic inflation.
  - Contemporaneous response of prices is small; effect increases significantly one and two months after an exchange rate shock.
  - VAR in differences (preferred tool) implies: a 10 percent depreciation of the tenge would be expected to trigger an increase in consumer prices on the order of 3 percent.
  - This finding is close to results for other developing and emerging market economies and consistent with Kazakhstan’s recent experience.
- Monetary policy implications:
  - Timing and magnitude of ERPT matter for inflation-targeting decisions on the policy rate.
  - Degree of ERPT affects speed of external adjustment under a flexible exchange rate regime:
    - If pass-through to import and producer prices is high while pass-through to consumer prices is low, nominal depreciation implies real depreciation and improved competitiveness.
  - Credible monetary policy frameworks anchor expectations and tend to be associated with lower pass-through.
  - Recommendation: NBK should incorporate ERPT information into monetary policy decisions and clearly communicate anticipated effects on prices following an exchange rate shock.
- Structural and financial policy recommendations:
  - Reducing dollarization and introducing more competition in domestic product markets could help reduce ERPT.
  - Financial dollarization amplifies exchange rate shocks and may be associated with higher pass-through to consumer prices.
  - Efforts to de-dollarize (deepening financial markets, encouraging development of hedging instruments) are key.
  - Greater competition (variable markups) allows firms to adjust markups downwards after depreciation to remain competitive.
  - Kazakhstan can benefit from faster implementation of the government’s privatization program and reforms to reduce state monopolies and promote private entrepreneurship.

### Data sources (summary)
- Domestic CPI, PPI, and NEER series: National Bank of Kazakhstan (inverse of original NEER used so that a positive change implies depreciation).
- Import prices and industrial production index: Committee of Statistics.
- External prices index: weighted average of producer/wholesale price indices of Kazakhstan’s main trading partners using NEER weights; primary source IFS, Haver Analytics used to gap-fill missing observations.

*Source: IMF staff calculations and analysis in “Estimating exchange rate pass-through in Kazakhstan” (section 18–31).*

### 1. Potential output is the level of output attainable given the underlying capacity of the

### cr17109 - 1. Potential output is the level of output attainable given the underlying capacity of the

### Definition and context
- Potential output: the level of output attainable given the underlying capacity of the economy, when all factors of production are fully utilized.
- The term “potential growth” is used interchangeably with the trend growth of output, abstracting from business cycle and seasonal variation.
- Recent literature shows trend growth in emerging economies is subject to substantial volatility (see Aguiar and Gopinath, 2007), making it difficult to disentangle trend shocks from cyclical fluctuations.
- Kazakhstan’s experience over the past 20 years: rapid expansions followed by rapid contractions, with significant volatility.

### Objective of the paper
- Quantify the level of potential output for Kazakhstan using three approaches:
  - Univariate filters that extract the trend from a single output series.
  - A multivariate filter that incorporates additional macroeconomic information.
  - A production function (growth accounting) decomposing output growth into inputs and total factor productivity (TFP).
- All three approaches find that potential growth has declined in recent years.
- Future prospects may improve with an expected increase in oil production and effects of structural reforms in the non-oil sector.

### Univariate filters (methods and caveats)
- Filters used:
  - Hodrick-Prescott (HP) filter: identifies a smooth trend path; choice of smoothing parameter determines trend behavior.
  - Band pass filters: Baxter-King (BK) and Christiano-Fitzgerald (CF); assume cyclical variations occur at frequencies corresponding to 6 to 32 quarters.
- Limitations and implementation details:
  - Univariate filters are purely statistical and not informed by other macroeconomic variables; they may not properly measure latent potential.
  - May fail to capture capacity utilization effects and impacts of inflation and employment movements.
  - HP filter susceptible to the “endpoint problem”; to mitigate, the sample was extended to include five years of forecasts.
  - Data: quarterly frequency, seasonally adjusted using X13; historical quarterly data on real GDP back to 1999 and IMF staff projections included to 2022 (Appendix I).
  - Baxter-King returns values from 2002 to 2019 due to lead/lag requirement; CF requires stationarity—series transformed after Augmented Dickey Fuller testing.

### Multivariate filter (method and advantages)
- Method: multivariate filter proposed by Blagrave et al. (2015).
- Advantages:
  - Incorporates economic theory.
  - Relatively modest data requirements: GDP, inflation, and unemployment.
  - Produces narrower confidence bands compared to standard HP filter.
  - Allows incorporation of expectations to identify shocks and improve end-of-sample accuracy.
- Implementation notes:
  - Consensus forecasts not available for Kazakhstan; annual vintages of five-year Fall IMF WEO projections used instead.
  - Allows judgmental inputs (potential growth, output gap, NAIRU) and high-frequency indicators.
  - Still susceptible to the endpoint problem.

### Production function / growth accounting (method and implementation)
- Production function: Cobb-Douglas form Y_t = A_t * K_t^α * L_t^(1-α) (as specified in the text with country-specific notation).
- Rationale: use country-specific elasticities rather than standard 0.65 labor / 0.35 capital used for advanced economies.
- Steps:
  - Estimation of country-specific parameters using firm-level data from Orbis and OLS regression:
    - Regression specification: ln(y_i) = a + α ln(k_i) + β ln(l_i) + ε_i.
    - Sample: 993 firms, representing about 14 percent of gross value added.
    - Estimated coefficients approximately 0.56 for labor and 0.44 for capital.
  - Extensions for input quality:
    - Human capital incorporated via years of schooling (H_t) and marginal returns to schooling r set at .107.
    - Capital stock initial year derived as K_0 = I_0 * (1 / (g + δ)), with δ = 0.07 assumed for depreciation.
    - Data on educational attainment from the Wittgenstein Centre at five-year intervals; gaps filled by interpolation.
  - Estimation of potential output:
    - TFP obtained as a residual.
    - HP filter applied to factors of production with smoothing parameter of 100 to extract trends.
    - Potential output computed by plugging trend factors into the production function; potential growth rate computed as year-on-year change of the level.
- Caveat: TFP calculations may reflect measurement errors and capacity utilization; capacity utilization in Kazakhstan heavily influenced by global oil prices and external shocks.

### Results — levels and trajectories
- General finding: all approaches indicate Kazakhstan’s trend growth has fallen in recent years to around 3 percent.
- Univariate filters:
  - Show trend growth peaked at around 10 percent in 2004.
  - Fell to around 2.5 percent in 2016.
  - Projected to rise to 3.5-4 percent by 2022.
- Multivariate filter (annual data 1996-2016):
  - Suggests potential growth has declined since mid-2000s.
  - Estimated value in 2016 of around 3.3 percent (statistical estimate without subjective judgment).
- Production function approach:
  - Estimated potential growth rate for 2016 is 3.0 percent.
- Sectoral note:
  - Assessment reflects overall GDP including the oil sector; methodologies could not be applied to the non-oil sector due to data limitations.
  - Kazakhstan’s potential growth historically closely linked to global oil prices; correlation between potential growth and oil prices is positive and relatively high over the sample period.
  - Rising oil prices led to greater dependence on oil exports; drops in oil prices had immediate negative impacts on net exports and growth.

### Results — contributions to growth (growth accounting)
- Over the sample period:
  - Labor and human capital combined contributed about 2.2 percentage points to growth in the last five years.
  - Capital accumulation contributed over 6 percentage points on average, though lower than pre-global financial crisis period.
  - TFP declined gradually and was sharply negative in years of negative oil price shocks (reflecting lower capacity utilization and financial sector weaknesses).
  - Early 2000s: TFP contributed positively and was the main driver of growth, likely reflecting greater openness and reforms.
- Projections and conditionalities:
  - Staff baseline shows relatively modest future contributions from capital accumulation.
  - Credible implementation of structural reforms and a rebound in global oil prices could raise the contribution from TFP and capital accumulation.
- Figures referenced:
  - Figure 4 shows growth rates of capital and output and contributions to growth (percentage points) across 1996–2016.
  - Trend components (HPTREND series) plotted for capital, employees, TFP, years of schooling, and trend output growth.

### Conclusions and policy implications
- All techniques converge on the finding that potential growth in Kazakhstan slowed substantially to around 3 percent in recent years.
- Main driver of volatility: oil sector dependence, which causes large year-to-year swings due to oil price shocks and low capacity utilization in downturns.
- Structural characteristics:
  - Physical capital inputs: consistently high contribution but falling.
  - Labor (number of employees and human capital): consistently low contribution.
  - TFP: highly uneven contribution, with negative contributions during oil price shocks.
- Policy recommendation:
  - Implementation of structural reforms would help reduce growth volatility and make 4 percent growth over the medium term an attainable goal, conditional on reforms and external conditions.

*Prepared by Jonah Rosenthal; sources as cited within the chapter.*

---


_Source: https://www.imf.org/-/media/files/publications/cr/2017/cr17109.pdf_