Forecasting Case Study: Ghana’s Oil Revenues

Posted on Posted in Superforecasting

by Samuel Bekoe and David Mihalyi

In order to help Ghanaian MPs and the general public understand the potential impact of volatile petroleum prices on the implementation of the 2015 budget, we built an oil revenue forecasting model during the budget debates in 2014. We found that some of the techniques familiar to superforecasters improved our forecasting process, including the importance of regular belief updating, finding the right base rate, and conducting a rigorous post-mortem to assess lessons learned. We offer our experience as a case study in forecasting to help bring public scrutiny to budget forecasting in developing countries.

The tool was built exclusively using publicly available data and information, was provided in Microsoft Excel, and released under an open license; it can be used and edited by anyone. Because no model is perfect, we believe opening it up allows for a more transparent and constructive debate around both the data and assumptions that make up the model and any interpretations that might be drawn from it.


As 2015 drew to a close, we reviewed how petroleum revenues had been performing and explored the implications for this year’s 2016 budget. We found that our revenue forecasts in November 2014 were far higher than actual revenues as they accrued during 2015. (They were more accurate than official projections.) Unpacking our forecasts’ shortcomings helped us shed light on which of our assumptions were wrong and why. For example, in our main scenario we did not anticipate the continued slide in global oil prices and how the government might change the rules for petroleum revenue allocation during the year in response. We also found a larger-than-expected drop in corporate tax payments. This raised further questions.

Based on preliminary information presented by the Ministry of Finance in the 2015 petroleum report (covering oil revenues between January and September), we estimated, using a simple extrapolation, that Ghana’s petroleum revenues for 2015 were likely to be $456 million in total. This was $780 million (63 percent) lower than what was budgeted by the Ministry of Finance and $500 million (52 percent) lower than we predicted with our own model at the end of 2014. This is a major difference relative to what the ministry had anticipated and what was possible to model with the information then available in the public domain.


The shortfall of $500 million compared to our own estimates (at $70 per barrel) can be attributed to two main factors:

  • First, the larger-than-expected drop in prices. Brent crude had averaged $55 per barrel over the first nine months of 2015—not the $70 we used in the model. The greater price change explains $216 million of the model’s shortfall.
  • Second, corporate income tax revenues were much lower than expected, accounting for $274 million of the shortfall after adjusting for the price drop. Our model assumed that companies would have recovered the initial development costs of the Jubilee field (Ghana’s main producing oil project, located offshore) and that their taxable income would be lessened only by current operating costs. The very limited payment of any corporate tax by oil companies in the first nine months of the year indicates this is likely not the case. Last year, when we published the model, we flagged that possible deductions of development costs from neighboring projects were and are a major unknown in our corporate tax payment projections. We suspect that given the existing ring-fencing provisions in publicly available contracts, the cost from neighboring TEN-field projects might have been offset against corporate taxes for Jubilee, therefore reducing the taxes due during 2015.

The additional $10 million in shortfall relates to a variety of factors with opposing effects. The slow progress in setting up the gas infrastructure has led to delays in this revenue stream. On the other hand, other revenue streams, such as royalties, the production share of oil going to government and surface rentals are performing better than our model predicted and offset some of the shortfall.

The publication of the model and related discussions with experts taught us a few important lessons.

1.  Belief Updating. Predicting the oil price is hard—forecasts therefore require constant revision. When the analysis was done in response to a sudden drop in price in the fall of 2015, very few would have predicted that prices would stay that low for that long.

As the figure above illustrates, models based on oil price forecasts need constant updating. We were not the only one to update our model—the government also submitted a revised budget midyear partly due to the oil price shock. Ghana’s government now projects oil revenues to be 1.5 percent of GDP in 2017, compared to the 3.1 percent of GDP it projected a year earlier. What does this new (and gloomy) outlook rest on? How do these take into account any uncertainties regarding which new oil projects may materialize? When analyses are done behind closed doors, it is hard to check their validity or relevance in changing contexts. Opening up models and assumptions is key.

2. Base Rate. Our models assume fiscal rules are followed. This is not always the case. Learning from the experiences of countries such as Chile and Botswana, Ghana adopted a set of fiscal rules protect its budget from commodity price volatility as part of the Petroleum Revenue Management Act of 2011. The act also prescribed the allocation of petroleum revenues between the national oil company, a spending account for development priorities and two sovereign wealth funds.

However, the implementation of this rule was criticized as being undermined by overoptimistic oil revenue forecasts and debatable interpretations of certain provisions. Faced by dropping oil prices in 2015, the parliament amended the rules and then changed the allocation of oil revenues. While our model can be adjusted for oil price variations, it was not designed to capture such within-year policy changes. More importantly, these ad hoc changes can also make planning and implementation difficult for government. The old rules were meant to help smooth revenues. They were meant to mitigate volatility risks as well as the potential for political interference in forecasting. While the exceptional circumstances resulting from the oil price drop may warrant adjustments to estimation procedure, this process should be transparent and well-defined. It should lack risks of discretion and ad hoc changes.

3. Post-Mortem for Lessons Learned. More transparency is needed to oversee petroleum revenues. This open modelling—with all its imperfections—was only made possible due to the advanced state of oil sector disclosures in Ghana: some oil contracts are published, the country is compliant with the EITI Transparency Standard, national laws require regular reporting, and international oil companies such as Tullow and Kosmos disclose additional key information.

However, there is not enough transparency on oil project costs, most importantly on how cost deductions from new oil projects are affecting taxes on profits from the Jubilee oil field. An ongoing audit by the Ghana Revenue Authority will hopefully shed light on the amount affected and when these delayed revenues may materialize. Transparency is crucial for oversight actors (such as the Public Interest and Accountability Committee) to effectively enforce rules on petroleum revenue management.

Building an open model and convening discussions around our analysis provided some valuable lessons, even when incorrect forecasting occurred. We were able to highlight our concerns and potential reasons for the shortfall as part of conversations with officials at the Ministry of Finance, Ghana EITI and the IMF. The model enabled discussions with various civil society organizations on the impact of the oil price fall on Ghana’s petroleum revenues. It also informed our recommendations on amending the income tax and the petroleum revenue management laws. In terms of the way forward, we identified the key pieces of information we need to improve the robustness of our model. With the increasing importance of the TEN and Sankofa fields in next years, contract transparency and more clarity on cost deductions will be key to better monitoring revenues.

This is a modified version of a post that originally appeared on NRGI’s blog on 17th December 2015. NRGI has been working on using open data to advance accountability and better decisionmaking in the petroleum sector in Ghana and globally. We also encourage other agencies in the extractive sector to open up their forecasts, following the footsteps of the IMF (which released its own fiscal analysis model for public use).

Samuel Bekoe is an Anglophone Africa associate and David Mihalyi is an economic analyst with NRGI.

Share on FacebookTweet about this on TwitterShare on LinkedInEmail this to someone

Leave a Reply

Your email address will not be published. Required fields are marked *