If we make a model of a set of data, the model necessarily represents past or near contemporary states of the economy. By the very nature of data we can’t have data about future states unless they are similar to past states. Since we are looking for new economic thinking we probably need to think beyond models that represent data. This probably means greater reliance on narratives of what’s possible than summaries through data of what already is.

Restricting research to existing data, especially big data sets, is a support for conservative thinking (the future should be like the past). Querying big data sets cannot introduce new categories since anything new will not be represented in the data.

We cannot query large data sets about things which have not yet happened and lie outside the boundaries of the preset and past.

For example, the impact of some form of universal income raises questions about what people would do and what social trends would emerge. No amount of existing big data can answer – because the relevant data does not exist.

Then the model is unhelpful because it can only represent varying states of the currently assumed parameters.

Big data cannot tell us about future states except in so far as they are like previous ones.

This means that imagination becomes more important and analysis less important.