Skip to content
W. Marañon
  1. Home
  2. Data & tools
  3. Bolivia FX monitor
LIVE

Bolivia FX monitor

Daily tracking of the official exchange rate, the USDT/BOB rate in the P2P market and the exchange-rate regime, calm or stress, with an improved version of the model in the paper on exchange-rate regimes.

Live application of a research paper

Persistence, duration and transition dynamics between stability and exchange-rate stress regimes

Current exchange-rate regime

Calm

at the close of Tue 3 Nov · since 1 Oct 2026 (34 days)

USDT/BOB is moving within its usual range, with a typical daily change of ±0.43%.

Probability of calm at the close: 92%

Value at Risk for Wed 4 Nov (95%) -0.99%

With 95% confidence, USDT/BOB would not fall more than 0.99% by the close of Wed 4 Nov.

Probability of stress

Wed 4 Nov 14%

0%alert 52%100%

Wed 4 Nov to Sun 8 Nov (5 days) 39%

0%alert 62.5%100%

No alerts: they turn on above 52% (one day) and 62.5% (5 days)

Thresholds calibrated with the paper’s own method. See how

Last update: 3 Nov 2026 (close at 23:20; all times are La Paz time)

Source: own MS-t model

Official exchange rate

12.06

Bs per USD · −0.03 (−0.2%) vs. previous official rate

Median of USD 33.0 million in 2,472 bank purchases on Tue 3 Nov; half between 11.98 and 12.09.

Last update: 4 Nov 2026, 00:00

In force on Wed 4 Nov · Source: BCB

USDT/BOB

12.08

Bs per USDT · +0.04 (+0.3%) vs. previous close

Last update: 3 Nov 2026, 23:20

Close of Tue 3 Nov · Source: Binance P2P

Gap vs. official rate

-0.1%

−0.01 Bs per dollar

Last update: 3 Nov 2026, 23:20

Close and official rate of Tue 3 Nov · own calculation

The regime

Calm and stress, explained

The model classifies each day into one of two regimes that differ in the size of USDT/BOB moves, not in their direction, and estimates every day the probability of switching from one to the other given recent volatility.

Calm

Current

±0.43%

typical daily change

Small, gradual moves in either direction; it is the market’s usual state and, with current volatility, it is expected to last about 12 days.

Stress

±2.35%

typical daily change

Large daily jumps, typical of confidence shocks; episodes are intense but short-lived, about 4 days under current conditions, and usually give way to normalization.

Regime history31 stress episodes between Jul 2024 and Nov 2026; the most recent from 7 Sep 2026 to 30 Sep 2026 (24 days).
Stress: 23 Jul 2024 to 18 Aug 2024 (27 days)Stress: 22 Aug 2024 to 23 Aug 2024 (2 days)Stress: 26 Aug 2024 to 27 Aug 2024 (2 days)Stress: 5 Sep 2024 to 16 Sep 2024 (12 days)Stress: 18 Nov 2024 to 27 Nov 2024 (10 days)Stress: 19 Jan 2025 to 20 Jan 2025 (2 days)Stress: 10 Mar 2025 to 19 Mar 2025 (10 days)Stress: 25 Mar 2025 to 26 Mar 2025 (2 days)Stress: 2 Apr 2025 to 13 Apr 2025 (12 days)Stress: 23 Apr 2025 to 2 May 2025 (10 days)Stress: 12 May 2025 to 30 May 2025 (19 days)Stress: 21 Jun 2025 to 22 Jun 2025 (2 days)Stress: 27 Jun 2025 to 3 Jul 2025 (7 days)Stress: 11 Jul 2025 to 14 Jul 2025 (4 days)Stress: 18 Jul 2025 to 12 Aug 2025 (26 days)Stress: 15 Aug 2025 to 18 Aug 2025 (4 days)Stress: 27 Aug 2025 to 12 Sep 2025 (17 days)Stress: 6 Oct 2025 to 1 Nov 2025 (27 days)Stress: 3 Nov 2025 to 20 Nov 2025 (18 days)Stress: 26 Nov 2025 to 12 Dec 2025 (17 days)Stress: 16 Dec 2025 to 25 Dec 2025 (10 days)Stress: 2 Jan 2026 to 2 Jan 2026 (1 days)Stress: 26 Jan 2026 to 30 Jan 2026 (5 days)Stress: 4 Mar 2026 to 10 Mar 2026 (7 days)Stress: 14 Mar 2026 to 14 Mar 2026 (1 days)Stress: 22 Apr 2026 to 22 Apr 2026 (1 days)Stress: 27 Apr 2026 to 4 May 2026 (8 days)Stress: 7 Jul 2026 to 14 Jul 2026 (8 days)Stress: 20 Jul 2026 to 10 Aug 2026 (22 days)Stress: 24 Aug 2026 to 2 Sep 2026 (10 days)Stress: 7 Sep 2026 to 30 Sep 2026 (24 days)
  • Calm
  • Stress

Official exchange rate and USDT/BOB

Daily, in bolivianos per dollar, with stress periods shaded; hovering over the chart shows each day’s detail.

81012Jan 26AprJulOctFlexible official rate
81012Jan 26AprJulOct
  • Official (BCB)
  • USDT/BOB (Binance P2P)
  • Stress periods

Early warning

Stress alerts, one and five days ahead

Probability computed at each day’s close for the next day, last 12 months; above the 52% line the alert is issued, the shaded areas mark the stress periods and the dashed end is the current forecast, for Wed 4 Nov.

0%20%40%60%80%100%Jan 26AprJulOctAlert 52%
0%20%40%60%80%100%Jan 26AprJulOctAlert 52%
How it works
Each close at 11:25 p.m. yields the probability that the next day will be one of stress and the probability of at least one stress day in the next five, each with its own alert, above 52% and 62.5%.
Where the 52% and the 62.5% come from
They are the thresholds that best balance anticipated stress and false alarms (maximum F1) in the paper’s training period, July 2024 to October 2025, with the every-day series; with the same method the paper set 20% for its model.
Horizons
The paper validates the one-day horizon; the five-day horizon was assessed with the same protocol and tells stress apart better than repeating today’s state (AUC 0.90 vs 0.78).

Verification

Forecast versus outcome

A probability is not checked with a single day, since a 10% forecast is not wrong if stress occurs; it is checked over many days, by seeing whether more forecast risk comes with more actual stress. A day counts as one of actual stress when the volatility of the last 5 days exceeds 1.21%, the paper’s definition.

Was the latest forecast right?

Forecast for Tue 3 Nov

made at the close of Mon 2 Nov

Probability of stress
16%
Value at Risk (95%)
-1.06%

Observed on Tue 3 Nov

close at 23:20

USDT/BOB change
+0.33%
5-day volatility
0.42%

Match: no alert and the day was not one of stress; the move stayed within the Value at Risk.

Forecast as published, saved before the outcome was known.

Out-of-sample record

364 days, 1 Nov 2025 to 30 Oct 2026, with the parameters estimated up to October 2025 (the paper’s training period).

Forecast probabilityDaysStress daysAverage move
Below 10%460%0.33%
10% – 33%1481%0.45%
33% – 60%5620%0.71%
60% or more11456%1.14%
  • Next-day alert (above 52%): 129 issued and 68 confirmed; it anticipated 68 of the 76 stress days.
  • 5-day alert (above 62.5%): 163 issued and 101 confirmed; it anticipated 101 of the 113 cases.
  • Value at Risk at 95%: 9 larger falls against 18 expected.

The order holds, since more forecast risk came with more stress and wider moves; at high risk the model overstates the level (it forecast 80% on average and stress occurred on 56% of those days) and projects more volatility than observed, which makes it prudent, as its Value at Risk shows.

Projected versus actual volatility

Daily USDT/BOB volatility over 5 days, in percent, last 12 months: the one the model projected a day ahead and the one observed.

0.0%2.0%4.0%6.0%Jan 26AprJulOct
0.0%2.0%4.0%6.0%Jan 26AprJulOct
  • Projected a day ahead
  • Actual
  • Stress periods

Both move together and the projected one stayed above the actual one on 91% of the days, except in the sharpest jumps; the chart uses the current model and the information available each day.

Every forecast is saved before the outcome is known; the live record will appear here once it covers 20 days.

Uses

What it is for

Research

The research behind the monitor

Conference paper · 2026

Persistence, duration and transition dynamics between stability and exchange-rate stress regimes in Bolivia's parallel dollar market (2024–2026): evidence from an autoregressive Markov-switching model

Walter Manuel Marañon Quiñones

The paper characterizes the persistence, duration and transitions between the calm and stress regimes of the parallel dollar market with an MS-AR(10) estimated on 462 daily USDT/BOB returns, validates its forecasts out of sample and concludes that exchange-rate pressure episodes are intense but transitory.

Key findings

68×
higher variance in the speculative regime than in the calm regime
4.77
mean duration of a stress episode (trading days)
11.25
mean duration of a calm spell (trading days)
0.86
out-of-sample ROC-AUC for anticipating the stress state one day ahead

From the paper to the monitor

The paper points to models with heavy tails and time-varying transition probabilities as the next step; the monitor adds both in an MS-t with 8 parameters instead of 16 which, on the paper’s series, forecasts better out of sample (log-score −1.27 vs −1.37) and discriminates stress better (AUC 0.89 vs 0.86). The monitor applies it to every calendar day, with the 11:25 p.m. close.

Methodology

How it is built

A day in the monitor (La Paz time)

  1. 00:05

    The monitor captures the official rate the BCB published the night before and publishes the day’s page: latest close, official rate in force and the forecast for the day.

  2. Every houron the hour

    The USDT/BOB price is recorded and the page shows it live, with its change since the close and the gap; from 3:00 p.m. it adds a provisional reading of the day’s regime.

  3. 23:10 · 23:20

    Last two records before the close.

  4. 23:25

    USDT/BOB close: the last price before this time, every day.

  5. ~23:30

    The BCB publishes the official rate computed with that day’s bank purchases; it applies from the next business day (weekends and holidays keep Friday’s).

  6. 23:35

    The model estimates the regime and the forecast for the next day.

  1. 01

    Data and daily cut

    USDT/BOB is the last Binance P2P price before 11:25 p.m. La Paz time, every day, just before the BCB publishes the official rate. The official rate (TCO) is the amount-weighted median of the banks’ dollar purchases from their clients between midnight and 5:00 p.m. on each business day, which the BCB publishes around 11:30 p.m. and which applies from the next business day. Every close goes through an automatic quality control.

  2. 02

    Model

    Every day the MS-t is re-estimated with the whole history since July 2024, every calendar day, and updates the probability of each regime, the probability of stress for the next day and over the next five, the expected durations and the Value at Risk.

  3. 03

    Validation

    The alert threshold and the performance were assessed with the paper’s protocol: training up to October 2025 and an out-of-sample test from November 2025, without using future information.

Reference values for information only; they do not constitute investment advice.