Sanctions on Venezuela: Cause or consequence of the crisis?
By Jose Morales-Arilla (@josemoralesa)
This is a translation of an article published by Prodavinci in Spanish.
Mark Weisbrot and Jeffrey Sachs just published a report on the effects of US 2017 financial sanctions on Venezuela. The authors assert that “sanctions have inflicted, and increasingly inflict, very serious harm to human life and health, including an estimated more than 40,000 deaths from 2017–2018”.
According to the authors, sanctions led to the death of 40,000 Venezuelans by inhibiting oil production in the country. Their argument can be summarized as follows:
1. Sanctions in August 2017 led to the collapse in oil production.
2. This collapse constrained access to foreign currency, which in turn led to the humanitarian crisis in the country.
This influential report has been used by several politicians and pundits to make a case against economic pressures on the Venezuelan regime, and is presented as evidence that the humanitarian crisis in the country is not a consequence of the regime’s policies, but of an “economic war” against the country.
Given the analytic nature of the report and the political relevance of its conclusions, we should evaluate the authors’ methods and assumptions against academic empirical standards to establish causal claims in the social sciences.
Brief notes on causality
The fundamental problem of causal inference is that, by definition, we cannot observe “what would have happened” with one variable if another one had behaved differently. Say that someone eats a burger at 3pm and has a heart attack at 5pm. We cannot establish that the burger caused the heart attack because we do not know her heart would have behaved at 5pm had she not eaten the burger — and it’s perfectly possible that she’d have had the heart attack all the same had she not eaten the burger in the first place. This unobservable “what would have been” is what social scientists usually refer to as the “potential outcome”.
While we cannot establish the effect of a burger on a person’s health, experimental or quasi-experimental studies on many different subjects can indeed assess the average effects of a high-calorie diet on cardiovascular health. The experimental (i.e. random) assignment of treatment in these studies is important because of what’s known as the “selection bias”. If people with health problems worsened their diet due to emotional reasons, the association between diet quality and health outcomes will confound the causal effect of the diet on health outcomes with the emotional channel through which health outcomes affect eating habits.
What this means is that “correlation does not imply causation”. This is why economists often try to “control” for factors that confound causal relationships, as well as “instrument” changes in the “cause” variable so that these are not a response to changes in the “consequence” variable. When concerns on “omitted variable bias” or “reverse causality” cannot be addressed, associations simply cannot be interpreted as causal.
Sanctions as a consequence
An important concern in studying the average economic effect of sanctions on sanctioned countries is that assignment is not random. Sanctions are a consequence of political and economic events in countries receiving and imposing sanctions. For example, if sanctions are often triggered by human rights violations and authoritarian episodes that are themselves triggered by accelerating economic downturns in the sanctioned country, then a simple correlation will confound the economic effects of sanctions with the political mechanism through which economic downturns may lead to sanctions.
From this perspective, it is understandable that several authors question Weisbrot and Sachs’ decision not to highlight pre-sanction trends in Venezuela. These sanctions did not come out of the blue, as they were triggered by the same events that led to the creation of the Lima Group — also in August, 2017. Both were triggered as geopolitical responses to the accelerating economic and social crisis in Venezuela; to the regime’s flagrant violation of human and civil rights of Venezuelans during the 2017 protests; and to the illegitimate creation of the supra-constitutional National Constituent Assembly.
It is perfectly possible that living conditions would have continued to worsen in the post-sanctions period even if sanctions had not occurred. Again, if authors cannot control for this possibility somehow, there is simply no way to assess the causal effect of sanctions, and any estimate presented would be misleading. Not only is it possible that things would continue to get worse, but that was the dynamic that triggered the sanctions. The implication of an accelerating crisis by the time sanctions were enacted (as shown in Hausmann and Muci or Bahar, Bustos, Morales and Santos) is that the continuation of such a dynamic would not be an unexpected outcome in the absence of sanctions.
The many shocks surrounding sanctions
Weisbrot and Sachs try to address an even more complex problem than the average effect of sanctions. Their question is about the effects of a specific sanction (US financial sanctions of 2017) on a specific country (Venezuela). This is a much more difficult question because, as we mentioned, the potential economic outcomes for Venezuela had there not been sanctions is not observable.
Different commentators have suggested that a “difference-in-differences” approach could allow analysts to address some of the concerns outlined above to address this particular question. This is incorrect, as this method requires multiple treatment and control units (sanctioned and non-sanctioned countries) to estimate an average treatment effect (average economic impact of sanctions). In this case, we only have a single treated unit (Venezuela) with an unobservable “potential outcome” (economic results). For this reason, the question can only be addressed as a case study through comparative methods that do not usually yield a statistical estimate of causal effects.
One comparative method which does yield causal statistical estimates is the “synthetic control” method. Nevertheless, the method establishes at least two requirements which are not met in this setting: 1) there needs to be other non-treated units (countries) experiencing similar trends to the treated unit (Venezuela) by the time of the treatment (sanctions), and 2) the only relevant event in the treated unit (Venezuela) at the time of the treatment (sanctions) is the treatment.
These requirements are not met in the context of Venezuela. No country in the world experienced the degree of material impoverishment that Venezuela went through between 2013 and 2017. But more importantly, sanctions were not a treatment that occurred in isolation. Multiple events in and around August 2017 could have accounted for any posterior deterioration in living conditions. For example, the National Constitutent Assembly (NCA), which led to the sanctions and to the Lima Group, could have heightened the judicial uncertainty of all economic agents to the point of leading to further economic deterioration.
Do we know what the effect of the NCA on living standards was? No, we do not. Any estimate of its economic effects would be misleading. It’s exactly the same for sanctions.
Is Colombia a good counterfactual for oil production in Venezuela?
The only piece of analytic evidence in Weisbrot and Sachs’ report is the following figure. Adapted from Francisco Rodriguez’s analysis, the figure shows oil production in Venezuela and Colombia, marking financial sanctions as the only relevant event of 2017.
Their reading of this figure is as follows: Oil production in Venezuela was falling before sanctions, but the government is not at fault because in Colombia -another heavy oil producer in the region- oil production was also falling as a consequence of low international oil prices. Nevertheless, once sanctions hit, Colombian production remains stable but Venezuela’s drops further. This suggests that all post-sanctions falls in oil production are a consequence of the sanctions.
All the points I made above should suggest that presenting accelerations in oil production drops as a causal outcome of sanctions is misleading. Throwing a number as the causal effect of sanctions is irresponsible when circumstances suggest that the effect could perfectly be 0. Nevertheless, it is worth considering Weisbrot and Sachs’ argument, as it is implicitly excusing Chavismo for oil production drops before sanctions.
Weisbrot and Sachs assume that oil production in Colombia is a good “counterfactual” for oil production in Venezuela. A counterfactual is a combination of non-treated units (non-sanctioned countries) used as an approximation to the “potential outcome” of treated units (oil production in Venezuela had sanctions not been enacted). Their main argument is that both Colombia and Venezuela focus on heavy oils which are likely to be affected by low oil prices as the ones experienced in 2016.
However, oil production in Colombia fell due to idiosyncratic reasons. According to the EIA, Canada — Venezuela’s main regional competitor in extra-heavy crude, increased its production between January 2015 and August 2017 by 8.8%. This was not exceptional for high marginal cost producers in the region. The US, with high marginal production costs on shale oil and gas, increased its production in that same period by 5.3%. Brazil, another neighbor with high marginal costs for off-shore production, increased its production by a whopping 29.5% in that same period. OPEC’s production grew by 8%. World aggregate production grew by 3.1%.
In the following figure we see the oil production index for Venezuela, Colombia and Canada along with the oil price series. The Canadian index tracks oil prices closely, but Colombian and Venezuelan oil production drop without apparent association with oil prices. Weisbrot and Sachs mark the moment that oil prices dropped below $30 without showing trends in the price, which hides the fact that oil prices were under $30 only in the first quarter of 2016. During the rest of that year, oil prices grew while Colombian and Venezuelan production fell.
Moreover, Weisbrot and Sachs do not show evidence that oil production drops in Venezuela concentrated in extra-heavy oil production fields, which are commercially vulnerable to price drops. If oil production was falling in Venezuela because of low oil prices, it would be expected for production to fall only in these fields. Medium and light oil fields -with low marginal production costs- would offer a “placebo”: if oil production also falls in these fields, it’s unlikely that the reason behind oil production drops in Venezuela was low oil prices.
Sadly, there is no public data with high periodicity on oil production by type of field in Venezuela. Nevertheless, the consulting firm IPD reported that in the first quarter of 2016 (when oil prices were at their lowest), oil production dropped in extra-heavy oil fields in the Orinoco Oil Basin as well as in medium and light oil fields elsewhere in the country. According to IPD, the main reasons behind oil production drops were “drilling challenges, natural gas compression issues and well maintenance difficulties due to restriction of field services and theft”. That is, when oil prices were at their lowest, oil production drops were not limited to extra-heavy crudes, and the reasons behind these drops were strictly operative and managerial.
The conclusion of this analysis is that oil production in Venezuela should have not fallen in the period before sanctions, and its fall cannot be imputed to low oil prices: High marginal cost producers in the region increased their oil production, and oil production in Venezuela dropped in low-cost fields because of managerial factors. Colombia is not a good counterfactual for Venezuelan production, and hand-picking it as benchmark helps excuse Chavismo as responsible for oil production cuts before and after sanctions.
Pre-sanction vs. post-sanction dynamics
Using worldwide production as a benchmark, we observe a clear pre-sanction negative trend in oil production, which dropped about 20% between January 2015 and August 2018 relative to the world aggregate. Using OPEC countries or an average of high marginal cost producers in the region would yield similar results.
Setting differences between pre-sanction and post-sanction dynamics is useful to identify pre-sanction losses and provide estimates of losses associated with pre-sanciton trends. Using world-wide production and Venezuelan pre-sanction production trends as reference, we can estimate production losses between 2015 and 2018, along with a baseline estimate of the share of such losses that can be imputed to pre-sanction dynamics.
The figure below shows the accounting of monthly oil production losses in Venezuela in comparison with the world-wide reference. As is observed, the bulk of the losses experienced by Venezuela either occurred before the sanctions or can be explained with pre-sanction trends.
Totaling these losses for the whole period suggests that between 2015 and 2018, Venezuela lost at least 635 MM barrels of oil production. 76% of these total losses either occurred before sanctions or can be explained with pre-sanction trends.
Reiterating the points made above, to argue that remaining 24% of losses associated with post-sanctions deteriorations are a consequence of sanctions would be misleading to say the least. On top of all that has been discussed above, a fundamental event for Venezuela’s oil sector around the time of the sanctions was the appointment of Manuel Quevedo as president of PDVSA in November 2017, effectively establishing military control over the Venezuelan oil industry. There is vast anecdotic evidence on how the appointment of Quevedo, and the ensuing military disruption of management and operations in PDVSA and other Joint Ventures led to important hurdles in an already weathered sector.
Given that many relevant shocks occurred in the same period, it is impossible to impute causal effects on any of them. Nevertheless, the figures shown suggest that post-sanction oil production losses did not start to accrue until after November 2017. This finding, adding to the aforementioned anecdotal evidence on the effects of military control over the oil industry, and given that sanctions in August 2017 did not prevent oil trade between the US and Venezuela, lead me to speculate that the most relevant explanation behind the acceleration in oil production drops is the military control over the oil industry.
But beyond any speculation, the points made above suggest that the bulk of oil production losses are due to pre-sanction events and dynamics. If Weisbrot and Sachs really wanted to understand the determinants of the collapse in oil production and behind the economic and humanitarian crisis in Venezuela, they should probably focus in studying explanations that precede sanctions.
Did sanctions cause the death of 40,000 Venezuelans?
The most controversial assertion in Weisbrot and Sachs’ article is that oil sanctions led to the death of at least 40,000 Venezuelans. The authors produce their estimate of 40,000 deaths with a back-of-the-envelope calculation based on non-public general mortality estimates, but do not report them. Consequently, we don’t have the required inputs to replicate their calculations.
Their argument is that sanctions caused the death of 40,000 Venezuelans through their effect on oil production. As we’ve established, Weisbrot and Sachs do not prove that there was an effect of sanctions on oil production. Consequently, in the absence of a mechanism, they also do not prove that sanctions led to 40,000 deaths. But even if sanctions had had some effect on oil production, the authors do not show any evidence attempting to separate the effects of sanctions from the abundant possible alternative explanations that could lead to an increased mortality.
This is a long way of saying that they do absolutely nothing to prove that sanctions led to increased mortality. They just come up with a calculation of 40,000 deaths, and impute it fully to sanctions just because, without any analysis to back it up. To say that any social phenomena led to the death of 40,000 people without any evidence is truly daring — even more for two development economists presenting their results under a halo of academic credibility.
Any study on the impact of sanctions on mortality should consider trends by the time sanctions were enacted. The figure below, for example, shows the evolution of the infant mortality rate by 1 year of age. Venezuela’s infant mortality grew by 76% between 2012 and 2017 (before sanctions), while it dropped in all other countries in the region. This same dynamic of aggressive social deterioration is shown in Hausmann and Muci and in Bahar, Bustos, Morales and Santos with regards to food and medicine imports in Venezuela. The deterioration was clearly accelerating by the time sanctions were enacted.
Weisbrot and Sachs do not prove that the US financial sanctions from 2017 affected oil production, and they do not even try to prove that sanctions killed 40,000 Venezuelans. Their methodologically weak report does not offer any lights on the debate over the causal effect of sanctions on the Venezuelan economy. Their bold and unwarranted claims are especially worrisome given that the authors hold PhD degrees in economics from two of the most prestigious universities in the World.
The analysis presented in this refutation suggests that debates around Venezuela’s catastrophe should focus on identifying its true causes, so that sanctions can be considered under the perspective of whether they address those causes or not. If the authors were willing to contemplate pre-sanction causes, they should not lack plausible explanations.