How Hayek can help South Africa
Nowcasting can help entrepreneurs and policymakers realise the benefits of local, timely knowledge
In September 1945, the economist Friedrich Hayek published what would become a seminal paper in economics. ‘The Use of Knowledge in Society’ challenged the then-prevailing idea of central economic planning and argued that individuals, given their own preferences, experiences and opportunities, could make more effective decisions than a central authority. Knowledge, he argued, was personal and subjective – each of us knows best what we desire. And because no entity receives this information in its totality, a centralised approach to economic planning could never fully grasp all this dispersed knowledge. Decentralised economic decision-making was, therefore, key to a flourishing society.
But if no one dictates what will be produced, how do firms know what to produce and how much of it? In a free market, Hayek argued, prices serve as a signal that communicates this dispersed information, helping firms to coordinate their plans with current conditions. When a company has accurate information about the prices of its inputs and potential outputs, it can make informed decisions about what and how much to produce.
Hayek’s paper was one of the first to emphasise the need for firms to gather, interpret, and respond to information continuously. In today’s data-driven world, this includes not only prices but also other market indicators, consumer behaviours, and trends. By paying attention to these signals, firms can adapt to changing circumstances and make decisions that enhance their competitiveness and profitability.
It is for this reason that firms globally are dedicating sizeable budgets to something called nowcasting. Nowcasting, a term derived from ‘now’ and ‘forecasting’, is the practice of predicting the present, the very near future, and the very recent past. It requires real-time collection and interpretation of dispersed data, offering a near-instant understanding of current circumstances and imminent trends. In essence, it is a modern adaptation of Hayek’s conceptualisation of the free market – a dynamic system that aggregates and interprets disparate data through the mechanism of prices.
The digital revolution and artificial intelligence have made nowcasting increasingly complex – and important. Firms (and, increasingly, governments) have access to a multitude of data sources, each feeding into the holistic understanding of economic conditions and industry-specific trends. Take Google, for example. Hal Varian, Google’s Chief Economist, recently presented the Okun Lecture at Yale University’s Department of Economics. His talk, ‘Nowcasting with Google Trends’, explored how Google search queries can be used to predict economic variables in real-time.
Google Trends delivers daily or weekly reports detailing search volumes across various topics. In his talk, Varian illustrated how this search data often parallel economic activities in various industries and provides early signals for government data, usually published at less frequent intervals. For example, Varian underscored how certain searches often culminate in immediate actions. The query ‘coffee near me’ is a strong indicator of an impending coffee purchase. Other economic activities, however, take time to realise. Searches related to car rentals, vacation plans, or house purchases can act as crystal balls for aggregate economic activity months or even years later.
These data points can combine to predict broad economic trends. While official GDP data in the US are released quarterly, academics have devised methods to utilise real-time data for short-term economic forecasting. A case in point is the OECD’s weekly GDP growth tracker, which employs Google Trends data. This tracker gathers search behaviour information from 46 countries, encompassing consumption, labour markets, housing, and industrial activity. This data then steers the OECD's GDP predictions and shapes economic policy decisions.
South Africa is included in the list of 46 countries. The figure shows the remarkably strong correlation between the OECD’s weekly tracker and the official South African GDP statistics.
Locally, nowcasting is gaining momentum, with (free) tools becoming rapidly available. The Bureau for Economic Research at Stellenbosch University uses a mix of quarterly, monthly and daily data – from national accounts, business surveys, and financial sector data – to nowcast things like GDP. But models keep on improving. In a 2021 working paper, five South African Reserve Bank economists built nowcasting models to predict South Africa’s GDP, showing that the volatility of GDP has increased and its mean has fallen markedly over the last five years. A consequence was that analyst and Reserve Bank forecasts overestimated GDP growth. The economists concluded: ‘We show that several of the statistical nowcasting models we present in this paper provide more accurate nowcasts than the official Reserve Bank and market analysts’ nowcasts.’
All five authors have since left the SARB, with two setting up their own company. Codera Analytics now hosts the South Africa Nowcast, which provides a weekly automated update of forecasts for GDP and unemployment and an assessment of how recent data publications contribute to the outlook for the economy. These nowcasts leverage Codera's EconData platform that makes publicly available data easy to use and enables automation of analytics and decision-making that draws on economic and financial data for South Africa. This is one of only a handful of similar free publicly available systems worldwide.
We should expect more targeted applications as data becomes publicly available and model quality improves. Take unemployment, for example, probably South Africa’s number one economic problem. In his talk, Google’s Hal Varian demonstrated how the actions of the recently unemployed could offer early indications of labour market outcomes. In the immediate wake of a layoff, many turn to Google for guidance on unemployment benefits. Search queries such as ‘Where is the unemployment office?’ or ‘What do I have to bring to the unemployment office?’ tend to spike on the Monday following a Friday dismissal. These searches usually reflect the actual unemployment benefits issued days or weeks later, serving as a real-time diagnostic tool for labour market health. One could imagine a similar system for South Africa, spatially disaggregated to show real-time labour market frictions, allowing policymakers to target policy interventions.
Friedrich Hayek’s prescient insights into the dispersed nature of knowledge have found an exciting application in nowcasting – real-time economic forecasting using big data. Hayek propounded that decisions made closest to the point of information collection tend to be the most effective, as they harness ‘local’ knowledge, often inaccessible to central planners. Nowcasting, through its capacity to instantaneously interpret economic indicators, democratises this ‘local’ knowledge. For policymakers, this means an enhanced ability to tailor immediate responses to economic indicators, reducing lag time and potentially mitigating downturns. Entrepreneurs, meanwhile, can leverage real-time insights to adjust strategies, making sharper, more nimble pivots in response to market dynamics. In essence, the world is witnessing Hayek’s wisdom encoded into algorithms: local, timely knowledge is indeed powerful, and in the hands of many, can lead to more efficient and effective decision-making.
An edited version of this article was first published on News24. Image created with Midjourney v5.1.
Can a centralised AI play the role of planner? This would capsize Hayek’s thesis regarding central vs distributed planning: unlike the USSR, Google is capable or handling superhuman amounts of data.