Model-based tenge outlook to help time EUR deposits and withdrawals. Last prediction run: 2026-10-21.
| Horizon | Aggregate | Brent ECM model | NBK survey model | IMF WEO | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Predicted | 90% CI | Change vs spot | Predicted | 90% CI | Change vs spot | Predicted | 90% CI | Change vs spot | Predicted | 90% CI | Change vs spot | |
| 1m | 578.8 | 541.4 - 616.2 | +14.9% | 520.8 | 501.7 - 542.0 | +3.4% | 636.8 | 620.2 - 653.3 | +26.4% | - | - | - |
| 2m | 654.1 | 616.8 - 691.5 | +29.8% | 538.5 | 508.2 - 573.8 | +6.9% | 769.7 | 746.4 - 793.1 | +52.8% | - | - | - |
| 3m | 648.8 | 611.4 - 686.2 | +28.8% | 554.4 | 516.1 - 599.6 | +10.1% | 743.2 | 714.6 - 771.8 | +47.5% | - | - | - |
| 4m | 642.2 | 583.9 - 700.6 | +27.5% | 567.8 | 525.9 - 618.9 | +12.7% | 716.7 | 683.6 - 749.7 | +42.3% | - | - | - |
| 5m | 634.5 | 576.1 - 692.8 | +25.9% | 578.8 | 533.3 - 632.1 | +14.9% | 690.1 | 653.2 - 727.1 | +37.0% | - | - | - |
| 6m | 625.7 | 567.3 - 684.0 | +24.2% | 587.8 | 538.9 - 644.2 | +16.7% | 663.6 | 623.1 - 704.0 | +31.7% | - | - | - |
| 7m | 606.9 | 528.6 - 685.3 | +20.5% | 595.1 | 546.3 - 653.5 | +18.1% | 637.0 | 593.3 - 680.7 | +26.5% | - | - | - |
| 8m | 604.1 | 525.8 - 682.5 | +19.9% | 601.1 | 552.9 - 660.0 | +19.3% | 610.5 | 563.8 - 657.2 | +21.2% | - | - | - |
| 9m | 601.0 | 522.7 - 679.4 | +19.3% | 606.2 | 554.2 - 664.0 | +20.3% | 584.0 | 534.4 - 633.5 | +15.9% | - | - | - |
| 10m | 597.7 | 519.3 - 676.0 | +18.6% | 610.6 | 560.4 - 669.5 | +21.2% | 557.4 | 505.2 - 609.7 | +10.7% | - | - | - |
| 11m | 594.2 | 515.8 - 672.5 | +17.9% | 614.6 | 564.2 - 673.7 | +22.0% | 530.9 | 476.1 - 585.7 | +5.4% | - | - | - |
| 12m | 590.6 | 512.2 - 668.9 | +17.2% | 618.1 | 566.8 - 676.9 | +22.7% | 504.3 | 447.1 - 561.6 | +0.1% | - | - | - |
| Horizon | Aggregate | Brent ECM model | NBK survey model | IMF WEO | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Predicted | 90% CI | Change vs spot | Predicted | 90% CI | Change vs spot | Predicted | 90% CI | Change vs spot | Predicted | 90% CI | Change vs spot | |
| 1m | 515.1 | 481.8 - 548.3 | +14.9% | 463.5 | 446.5 - 482.3 | +3.4% | 566.7 | 552.0 - 581.4 | +26.4% | - | - | - |
| 2m | 582.1 | 548.9 - 615.4 | +29.8% | 479.3 | 452.3 - 510.6 | +6.9% | 685.0 | 664.2 - 705.8 | +52.8% | - | - | - |
| 3m | 577.4 | 544.1 - 610.7 | +28.8% | 493.4 | 459.3 - 533.6 | +10.1% | 661.4 | 635.9 - 686.8 | +47.5% | - | - | - |
| 4m | 571.5 | 519.6 - 623.5 | +27.5% | 505.3 | 468.0 - 550.8 | +12.7% | 637.8 | 608.4 - 667.2 | +42.3% | - | - | - |
| 5m | 564.6 | 512.7 - 616.5 | +25.9% | 515.1 | 474.6 - 562.5 | +14.9% | 614.2 | 581.3 - 647.0 | +37.0% | - | - | - |
| 6m | 556.8 | 504.9 - 608.7 | +24.2% | 523.1 | 479.6 - 573.3 | +16.7% | 590.5 | 554.5 - 626.5 | +31.7% | - | - | - |
| 7m | 540.1 | 470.4 - 609.8 | +20.5% | 529.5 | 486.2 - 581.6 | +18.1% | 566.9 | 528.0 - 605.8 | +26.5% | - | - | - |
| 8m | 537.6 | 467.9 - 607.3 | +19.9% | 534.9 | 492.0 - 587.4 | +19.3% | 543.3 | 501.7 - 584.9 | +21.2% | - | - | - |
| 9m | 534.9 | 465.1 - 604.6 | +19.3% | 539.5 | 493.2 - 590.9 | +20.3% | 519.7 | 475.6 - 563.8 | +15.9% | - | - | - |
| 10m | 531.9 | 462.2 - 601.6 | +18.6% | 543.4 | 498.7 - 595.8 | +21.2% | 496.1 | 449.6 - 542.6 | +10.7% | - | - | - |
| 11m | 528.8 | 459.0 - 598.5 | +17.9% | 546.9 | 502.1 - 599.5 | +22.0% | 472.4 | 423.7 - 521.2 | +5.4% | - | - | - |
| 12m | 525.6 | 455.8 - 595.3 | +17.2% | 550.1 | 504.4 - 602.3 | +22.7% | 448.8 | 397.9 - 499.8 | +0.1% | - | - | - |
Every tracked institution's latest published annual-average targets, in native form (these are yearly numbers, not monthly paths). USD/KZT as published; EUR/KZT equivalent at today's EUR/USD in parentheses. NBK survey and IMF WEO additionally power monthly model columns above when fresh; a source joins the monthly tables and the Aggregate once its latest vintage is current and covers both this year and next.
| Institution | Published | Forecast (annual avg USD/KZT) | Status | OOS error 1-3m | 7-12m |
|---|---|---|---|---|---|
| NBK survey | 2026-05-28 | 2026: 499 (~561 EUR/KZT) | 2027: 531 (~597 EUR/KZT) | 2028: 562 (~632 EUR/KZT) | current | 3.9% | 7.8% |
| IMF WEO | 2025-04-15 | 2025: 514 (~578 EUR/KZT) | 2026: 541 (~608 EUR/KZT) | 2027: 574 (~645 EUR/KZT) | 2028: 601 (~675 EUR/KZT) | stale (18 months old) | 5.9% | 7.4% |
| Gov budget | 2025-10-22 | 2026: 540 (~607 EUR/KZT) | 2027: 540 (~607 EUR/KZT) | 2028: 540 (~607 EUR/KZT) | stale (12 months old) | 2.7% | 7.7% |
| EDB | 2026-06-15 | 2026: 490 (~551 EUR/KZT) | current | 2.9% | 5.2% |
| AERC | 2026-01-29 | 2026: 546 (~614 EUR/KZT) | stale (9 months old) | 3.8% | 13.9% |
| Halyk Finance | 2026-04-27 | 2026: 515 (~579 EUR/KZT) | current | 4.3% | 6.8% |
Average percentage miss (how wrong each model has been, out of sample) per forecast horizon, from forward-chaining cross-validation. Green means the model beat the random walk (rate stays at today's spot) at that horizon; the sample count n is shown under each value -- treat small-n cells as indicative only.
Reading across rows: each source is scored only on the months it published, so n differs by row and the rows are NOT a like-for-like ranking -- a sparse source (e.g. EDB) is measured on its own chosen months, not the same set as the Aggregate or ECM. On the months where the two live sources (Brent ECM, NBK survey) overlap, the equal Aggregate is competitive with the better of them (it beats ECM at the 4-6 and 7-12 month horizons). The Aggregate is not more accurate than its best member because the two live sources' errors are ~0.8-0.9 correlated -- averaging highly-correlated forecasts cannot beat the better one by much. A materially better aggregate needs a more accurate, less-correlated live source (e.g. EDB or Halyk once they publish a live monthly forecast), not cleverer weighting of today's two.
| Model | 1m | 2m | 3m | 4m | 5m | 6m | 7m | 8m | 9m | 10m | 11m | 12m |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Brent ECM |
2.1%
n=82 |
3.8%
n=81 |
4.7%
n=80 |
5.2%
n=76 |
5.5%
n=75 |
5.9%
n=74 |
6.1%
n=70 |
6.3%
n=69 |
6.5%
n=68 |
6.5%
n=67 |
6.6%
n=66 |
6.6%
n=65 |
| NBK survey |
2.6%
n=27 |
4.2%
n=27 |
4.8%
n=27 |
5.9%
n=26 |
6.8%
n=26 |
7.0%
n=25 |
7.3%
n=22 |
8.2%
n=22 |
7.7%
n=21 |
7.4%
n=20 |
8.3%
n=20 |
8.1%
n=19 |
| IMF WEO |
3.7%
n=16 |
7.1%
n=16 |
6.7%
n=16 |
6.5%
n=16 |
6.8%
n=16 |
6.7%
n=16 |
6.8%
n=16 |
6.9%
n=16 |
7.5%
n=16 |
7.6%
n=16 |
7.8%
n=16 |
7.5%
n=16 |
| Gov budget |
1.5%
n=11 |
2.8%
n=11 |
4.0%
n=11 |
6.0%
n=10 |
5.5%
n=10 |
5.0%
n=10 |
5.2%
n=11 |
6.9%
n=11 |
7.7%
n=8 |
8.7%
n=8 |
9.8%
n=8 |
9.5%
n=8 |
| EDB |
1.7%
n=12 |
3.0%
n=12 |
4.0%
n=12 |
4.1%
n=12 |
3.3%
n=11 |
3.5%
n=11 |
3.8%
n=10 |
4.8%
n=10 |
5.9%
n=10 |
5.9%
n=10 |
4.9%
n=9 |
6.1%
n=9 |
| AERC |
4.2%
n=13 |
3.0%
n=11 |
4.1%
n=11 |
3.8%
n=11 |
5.2%
n=11 |
7.5%
n=8 |
10.1%
n=6 |
12.8%
n=6 |
16.7%
n=4 |
14.6%
n=3 |
16.4%
n=3 |
18.6%
n=2 |
| Halyk Finance |
2.9%
n=39 |
4.6%
n=39 |
5.3%
n=39 |
5.7%
n=38 |
6.4%
n=37 |
6.2%
n=36 |
6.3%
n=34 |
6.9%
n=33 |
7.0%
n=32 |
6.6%
n=30 |
7.0%
n=30 |
6.9%
n=29 |
| Aggregate |
2.2%
n=50 |
4.1%
n=49 |
5.1%
n=48 |
5.9%
n=47 |
6.6%
n=46 |
7.3%
n=45 |
7.6%
n=44 |
7.4%
n=43 |
7.5%
n=42 |
7.5%
n=41 |
7.5%
n=40 |
7.3%
n=39 |
| Random walk |
1.6%
n=200 |
2.8%
n=197 |
3.6%
n=196 |
4.0%
n=189 |
4.5%
n=186 |
4.7%
n=180 |
5.0%
n=169 |
5.6%
n=167 |
5.8%
n=159 |
5.9%
n=154 |
6.6%
n=152 |
6.9%
n=148 |
Out of sample no model or the Aggregate beats a random walk (rate stays at today's spot) at any horizon -- read every forecast as roughly today's rate with wide bands, and treat short-horizon moves with particular caution. The random walk row is the baseline all models are compared against.
The Aggregate headline blend is per-horizon and equal-weighted (equal weights won a pre-registered out-of-sample comparison). At 1-6 months it averages the two model paths, Brent ECM and NBK survey. At 7-12 months it also adds the annual/semiannual institutions' next-year consensus -- currently the government budget's implied path; the IMF WEO and EDB join when their latest vintage carries a next-year target and is recent enough. A source is included only when its latest vintage can anchor today (a looser 14-month window applies to the annual sources at long horizons); who is in or out is named under the forecast table. Its 90% band comes from the blend's own out-of-sample error spread, not the individual models' internal bands. See the scoreboard above for how the Aggregate and each source have actually done against a random walk.
The Brent ECM model forecasts USD/KZT with an Engle-Granger cointegration model: the tenge has a long-run oil-implied fair value (Brent elasticity plus a structural drift from the KZ-US inflation differential), and short-run deviations correct toward it with a half-life of about 3 months. The forecast iterates that error-correction dynamic forward with Brent held at its latest level; the 90% band comes from Monte Carlo simulation of the model's monthly shocks.
The NBK survey model turns the latest NBK analyst-survey consensus (median annual-average USD/KZT for this year and the next two) into a monthly path: elapsed months use realized USD/KZT, future months interpolate linearly to each year's survey average. Its 90% band scales the survey's own historical 1-year error (sigma(h) = RMSE_1y * sqrt(h/12)). The survey's own track record: same-year nowcasts are near-perfect (about 1% average miss), but its 1-2 year forecasts have historically carried almost no signal (Mincer-Zarnowitz beta 0.11 on 13 observations) -- the scoreboard above shows the same weakness at monthly horizons.
EUR/KZT is derived as forecast USD/KZT times the current EUR/USD rate, which is held constant over the horizon: neither model takes a view on the euro-dollar, so all predicted movement comes from the tenge leg.
Data updates daily from Yahoo Finance (Brent, USD/KZT, EUR/USD). Models: fx_combined_v1, kzt_brent_ecm_v1, nbk_survey_v1.