How can the productive capabilities of each municipality be unleashed taking into consideration the resources available to them? A first pass at this ambitious question begins by understanding the set of outputs a municipality is capable of producing. We answer this by discovering relationships between agricultural inputs and outputs and ask a relatively simpler question: how similar are agricultural outputs in terms of the inputs they use? Answering this question is made difficult by the fact that most UPAs cultivate just one or two crops. This may be a rational response to economies of scale. Given a plot of land and inputs, it may be easier to cultivate one crop on the entire land than plant a number of them with each requiring a different care regimen³. It may be that the inputs available only allow for a few types of crops.

In this paper, we use the rural census data from Colombia to build an agricultural product space capturing the similarities between outputs. We test the predictive power of the product space and use this to answer the question above. In section 2, we discuss the various sources of data and how they are merged, cleaned, and transformed before processing. In section 3, we look at some high level features of the dataset and how inputs, outputs, and land use are related. In section 4, we explore the mechanics of diversification.

We construct similarity and density matrices and show that they do indeed predict what a municipality produces. Finally, in section 5 we use Machine Learning algorithms and the density matrices to predict municipalities that are best suited to produce a given output. Further, we identify "missing" municipalitiesoutput pairs i.e. municipalities that should be producing a given output at high yield but currently are not. Finally, we summarize our findings and suggest areas for further work.

In this report we will be making extensive use of concepts described in more detail in the companion report "How Industry-Related Capabilities Affect Export Possibilities," especially with respect to Machine Learning techniques.

The central question we will explore in this document is: Can we anticipate the opportunities that Colombian cities have to export specific products based on their existing productive capabilities?

In the following pages, we report a collection of results, analyses, and advances in which we assess how industry-related capabilities affect export possibilities. Our final goal will be to create a single measure that synthesizes all the knowledge and existing information about the productive capabilities of each city, both “horizontal” and “vertical”, and that quantifies how competitive a city can be if it aims at exporting a given product it does not yet export.

This document is broken in two main efforts: First, we want to understand the “mechanics” of diversification processes. And second, we want to be able to provide recommendations of products that are not produced in cities, but should be. The first effort requires a multitude of analyses, each trying to describe the characteristics of firms, of cities, and of the mechanisms that expand the export baskets of places. The second effort requires the development of a statistical model that is accurate when predicting the appearances of products in cities. These two efforts, explaining and predicting, are complementary, but different.

Explanations that lack the power of accurately predicting the future are useless in practice; predictions of phenomena for which we lack understanding are dangerous. But together they provide a unified story that can inform policy decisions.

Are regions poor because they have bad institutions or are they poor because they are disconnected from the social channels through which technology diffuses? This paper tests institutional and technological theories of economic convergence by looking at income convergence across Colombian municipalities. We use formal employment and wage data to estimate growth of income per capita at the municipal level. In Colombia, municipalities are organized into 32 departamentos or states. We use cellphone metadata to cluster municipalities into 32 communication clusters, defined as a set of municipalities that are densely connected through phone calls. We show that these two forms of grouping municipalities are very different. We study the effect on municipal income growth of the characteristics of both the state and the communication cluster to which the municipality belongs. We find that belonging to a richer communication cluster accelerates convergence, while belonging to a richer state does not. This result is robust to controlling for state fixed effects when studying the impact of communication clusters and vice versa. The results point to the importance of social interactions rather than formal institutions in the growth process.

Labor informality, associated with low productivity and lack of access to social security services, dogs developing countries around the world. Rates of labor (in)formality, however, vary widely within countries. This paper presents a new stylized fact, namely the systematic positive relationship between the rate of labor formality and the working age population in cities. We hypothesize that this phenomenon occurs through the emergence of complex economic activities: as cities become larger, labor is allocated into increasingly complex industries as firms combine complementary capabilities derived from a more diverse pool of workers. Using data from Colombia, we use a network-based model to show that the technological proximity (derived from worker transitions between industry pairs) of current industries in a city to potential new complex industries governs the growth of the formal sector in the city. The mechanism proposed has robust strong predictive power, and fares better than alternative explanations of (in)formality.

Cities thrive through the diversity of their occupants because the availability of complementary skills enables firms in the formal sector to grow, delivering increasingly sophisticated products and services. The appearance of new industries is path dependent in that new economic activities build on existing strengths, leading cities to both diversify and specialize in distinct areas. Hence, the location of necessary capabilities, and in particular the distance between firms and people with the skills they need, is key to the success of urban agglomerations. Using data for Colombia, this paper assesses the extent to which cities benefit from skills and capabilities available in their surrounding catchment areas. Without assuming a prioria a definition for cities, we sequentially agglomerate the 96 urban municipalities larger than 50,000 people based on commuting time. We show that a level of agglomeration equivalent to between 45 and 75 minutes of commuting time, corresponding to between 62 and 43 cities, maximizes the impact that the availability of skills has on the ability of agglomerations to generate formal employment. Smaller urban municipalities stand to gain more in the process of agglomeration. A range of policy implications are discussed.

Tourism is one of the most important economic activities in the world: for many countries it represents the single largest product in their export basket. However, it is a product difficult to chart: "exporters" of tourism do not ship it abroad, but they welcome importers inside the country. Current research uses social accounting matrices and general equilibrium models, but the standard industry classifications they use make it hard to identify which domestic industries cater to foreign visitors. In this paper, we make use of open source data and of anonymized and aggregated transaction data giving us insights about the spend behavior of foreigners inside two countries, Colombia and the Netherlands, to inform our research. With this data, we are able to describe what constitutes the tourism sector, and to map the most attractive destinations for visitors. In particular, we find that countries might observe different geographical tourists' patterns - concentration versus decentralization -; we show the importance of distance, a country's reported wealth and cultural affinity in informing tourism; and we show the potential of combining open source data and anonymized and aggregated transaction data on foreign spend patterns in gaining insight as to the evolution of tourism from one year to another.

Social relations involve both face-to-face interaction as well as telecommunications. We can observe the geography of phone calls and of the mobility of cell phones in space. These two phenomena can be described as networks of connections between different points in space. We use a dataset that includes billions of phone calls made in Colombia during a six-month period. We draw the two networks and find that the call-based network resembles a higher order aggregation of the mobility network and that both are isomorphic except for a higher spatial decay coefficient of the mobility network relative to the call-based network: when we discount distance effects on the call connections with the same decay observed for mobility connections, the two networks are virtually indistinguishable.