Using large language model to analyse national devlelopment plans

Most countries in Africa and South Asia have a national development plan – a vision for how the country can achieve its development goals. Such documents have been a staple of governments strategising since independence. But what guidance can these strategies offer private actors, in particular investors? And what if anything do they have in common, being drawn from a range of countries so diverse in terms of culture, geography and income?

Challenges

British International Investment (BII) is the UK’s development finance institution. It has a dual objective: to support growth and jobs that lift people out of poverty, and to make a financial re-turn, which we invest in more businesses. In this way, it uses its capital over and over again to help create the jobs and economic stability that will enable countries to leave poverty behind.

BII wanted to understand the desires of national governments in the countries where it operates, so that it could align its investment strategy with the priority sectors for economic development.

Expertise

Due to the large number of strategies, a traditional (hand-written) literature review was not feasible. Instead we turned to large language models (LLMs, in this case Anthropic’s Claude model) to read the vast volume of text data and to draft the report.

First, our team of experts manually reviewed a sample of documents and then set about devising prompts that would summarise the strategies, eliciting the right information. After settling on a preferred prompt specification, we rolled it out on the full set of documents via Anthropic’s Application Programming Interface (API), effectively using the LLM to produce a first draft report. This was then extensively reviewed, verified and modified by our expert consultants.

Impact

We found that national development strategies across these regions are remarkably similar in terms of their broad goals and aspirations for their people: everyone wants economic growth, higher living standards for their people, and a diverse and resilient economy.

There was more variety in terms of strategies for how to achieve these goals, but many strategies expressed a broad policy orientation, rather than a specific policy direction beyond laws and actions that were already in force or in preparation.

Very few strategies contained specific guidance for investors, in terms of quantitative investment targets or named investment projects.
 

Banner image: Unsplash