Will Bolivian president Rodrigo Paz Pereira remain in power through the end of 2026?

TL;DR

I assign an 85% chance that Rodrigo Paz Pereira remains continuously president through December 31, 2026. His position is weak: an August urban online poll put approval at 21%, Unidad withdrew its support, and the Cerimedo investigation continues to generate allegations against his circle (Ipsos CIESMORI, ANF, EFE). Yet he survived the much more dangerous May–June removal campaign, protest activity has since fallen, and public records show neither a security-force break nor a mature constitutional process against him (AP, ACLED, Bolivian Constitution).

Context

Bolivia's official presidency still identifies Paz as president and dates his assumption of office to November 8, 2025. He was publicly carrying out presidential duties on September 8, one day before this forecast cutoff (Presidency of Bolivia, La Razón). Roughly 114 days remained before resolution.

The government is no longer facing the nationwide siege seen in May and June, but the underlying fuel, currency and political problems remain. Verified current reporting shows an indefinite fuel-related blockade in Comarapa on September 8, a pro-Morales march in Cochabamba on September 4, and a national Evo Pueblo meeting scheduled for September 26 (Radio Splendid, EFE, La Patria). The state of exception also expires around September 18–20, creating the main near-term test.

Evidence

The historical base rate starts low. From Bolivia's democratic restoration on October 10, 1982 through Paz's inauguration, there were about 43.08 president-years of exposure. Three presidents clearly left early under protest or coercive political pressure: Gonzalo Sánchez de Lozada in 2003, Carlos Mesa in 2005 and Evo Morales in 2019; including Hugo Banzer's illness-related resignation and Hernán Siles Zuazo's shortened term raises the broad count to five (Washington Post on Sánchez de Lozada, Congressional Research Service on Mesa, Human Rights Watch on Morales, VOA on Banzer, Siles Zuazo reference). A constant-hazard calculation gives only a 2.1%–3.5% chance of an unscheduled exit over 113 days. Scanning every eligible 114-day window produces a similar 2.2%–3.7%. This is the calm prior, not the final forecast.

A stress-conditioned reference class is much harsher. In a hand-coded core set of 12 elected Latin American presidents facing sustained national protests that explicitly sought their departure, seven left within 120 days. All seven departures came within a week of the acute peak; among the five presidents who survived beyond that point, only Dilma Rousseff left before day 194, through a developed impeachment process (Ecuador cases, Argentina in 2001, Guatemala in 2015, Rousseff impeachment). Applied mechanically from roughly day 81 after Paz's June crisis peak, this gives about a 20% exit risk through year-end. The sample is small and selected, and Paz lacks a Rousseff-style active impeachment case, so I use 20% as an upper stress anchor rather than my central estimate.

Paz's most valuable evidence is revealed survival. The May–June campaign lasted roughly 53 days, joined labor, campesino, indigenous, mining, transport and Morales-aligned organizations, and explicitly demanded his resignation. Paz negotiated with individual sectors, accepted concessions and personnel losses, obtained emergency powers, and used police and military support to clear roads. Parliament overwhelmingly ratified the emergency decree, and the principal blockades were gone by June 23 (AP, El País, ACLED). That episode showed a high personal threshold for resignation and, at the time of maximum pressure, retained control of the state.

The weekly updated ACLED/HDX aggregation also shows a real de-escalation. It counts demonstration events rather than participants, and the surprisingly low June total means it should be used directionally. Its full available history through the latest complete month is below (ACLED methodology, ACLED Bolivia).

The legal barriers are substantial. Articles 169 and 170 provide for vice-presidential succession after death, resignation, definitive incapacity, final criminal conviction or recall. Article 184 requires a grounded request from the prosecutor general and authorization by two-thirds of legislators present before the Supreme Tribunal can try a sitting president; recall cannot begin until half the term has elapsed, around May 2028 (Bolivian Constitution, electoral authority's recall guidance). The PDC won 49 deputies and 16 senators, 65 of the Assembly's 166 seats, leaving Paz short of a majority even before Unidad withdrew support (Bolivian electoral authority, ANF). Congress formed a supermajority to censure his economy minister in August, but the only publicly reported action against Paz himself is a private filing seeking to compel proceedings. No prosecutor-general request, legislative authorization, trial or final conviction has been reported (DW, ANF).

Paz has little public cushion. Ipsos CIESMORI's August 19–31 survey reported 21% approval and 65% disapproval for Paz, while Vice President Edmand Lara had 10% approval and 75% disapproval. The sample was 400 internet-connected adults in La Paz, El Alto, Cochabamba and Santa Cruz, using self-administered online interviews and a stated reference margin of error of ±4.9 percentage points (Ipsos CIESMORI, Brújula Digital methodology). It is strong evidence of urban weakness, but not a nationally representative measure of the rural organizations most capable of sustaining blockades. Lara's worse rating also makes constitutional succession less attractive to conventional opposition parties and business groups.

The main downside catalyst is the overlap between scandal and economic pain. On September 8, prosecutors added two alleged offenses to the illicit-enrichment investigation of former adviser Fernando Cerimedo, while Lara released another purported recording involving Cerimedo and a minister. The recordings remain disputed and have not established criminal conduct by Paz himself (EFE, Brújula Digital). Fuel and foreign-exchange stress continue: the central bank's official exchange rate effective September 8 was Bs12.60 per US dollar, while its July 31 dashboard put net international reserves including gold-related obligations at US$3.3613 billion (Central Bank exchange rate, Central Bank reserves dashboard). The September 3 inflation release covering August showed 1.10% month over month and 5.02% over twelve months, while the July 29 US$1.9 billion IMF staff-level agreement still required prior actions and Executive Board approval (INE, IMF). These conditions can regenerate unrest even if headline inflation has slowed.

There is still no verified public evidence of police or military defection. False videos alleging a police mutiny and a military demand for Paz's resignation have circulated, but independent checks found them manipulated. Police commander Mirko Sokol resigned on August 31, yet Paz immediately appointed Adrián Álvarez, which signals institutional stress and continued executive command at the same time (ChequeaBolivia, AFP Factual, Erbol). The OAS's September 3 report also found broad agreement among the sectors it consulted that elected mandates must be respected in full and changed only through constitutional mechanisms (OAS). This raises the cost of an extra-constitutional transfer, though it cannot prevent one.

My final model assigns 61% to containable unrest and scandal, with 98% conditional survival; 29% to a renewed national crisis in which opposition and state institutions remain divided, with 81% survival; and 10% to a cross-institutional rupture or direct legal jeopardy, with 20% survival. The weighted calculation is 0.61 × 0.98 + 0.29 × 0.81 + 0.10 × 0.20 = 0.8527. I round this to 85% in prose and report 0.853 as the forecast.

What's non-obvious

The obvious read is that 21% approval, cabinet churn and renewed blockades mean Paz is close to falling. The missing step is succession. Protesters, Lara, Unidad, Libre and the security institutions would need to agree on how to create and manage a permanent vacancy. Lara automatically benefits if Paz leaves, but he is even less popular in the available survey, while conventional opposition leaders gain more from constraining a weak Paz than from installing Lara. This successor problem is a real stabilizer, not a claim that Paz governs effectively.

Some current-looking reporting also overstates September unrest. A widely surfaced item appeared to report transport blockades in four departments on September 8, but the article body is dated May 5 and describes the earlier national crisis (Oxígeno). I did not count it as current evidence. Verified September 8 reporting was narrower, centered on the Comarapa fuel blockade (Radio Splendid). There is a similar vintage trap in reserves: the often-cited 85% gold share is from December 31, 2025, not the July 2026 reserve observation (Central Bank report).

Uncertainties

• The next two weeks matter most. The decree established a 90-day state of exception from June 20, but domestic reporting gives both September 18 and September 20 as the endpoint. An official expiration notice, extension request or post-expiration COB resolution would materially reduce uncertainty (Decree 5636, Radio Fides).
• The Cerimedo recordings lack a conclusive public forensic authentication. A verified recording directly showing Paz ordering a crime or illegal security operation would cut the forecast sharply; a forensic rejection would raise it.
• Security-force loyalty is partly private. Public command continuity and successful deployments favor Paz, but they cannot rule out hidden discontent or a rapid change following casualties.
• The approval survey is small and urban. A national rural-and-urban probability survey, plus current polling of COB, campesino and transport members, would tell us far more about mobilization capacity than the available four-city online sample.
• The historical crisis class is small and definition-sensitive. The 20% survivor-conditioned exit rate rests mainly on one delayed case, Rousseff, and should not be treated as a stable statistical law.