Foreseeing in the dark

Macroeconomic Forecasting in a Health Crisis

Created: 10 April 2020

By Olivier Pilmis

The current crisis is not only a health crisis but also an economic one: confinement is certainly a barrier to certain consumer practices, but it is also an obstacle for a large portion of productive activity. The situation calls for comparison with the great recessions and depressions of the past: certain indicators bring to mind the 2008 crisis, and others 1929… But such parallels leave us with many unanswered questions: to what depths will the health crisis plunge world economies, and the French economy in particular? Will the anticipated recession be short-lived or long-lasting? The COVID-19 epidemic inaugurates a period of great uncertainty that calls for forecasts that it paradoxically makes more difficult to produce

Forecasting is Governing

Forecasts are, first of all, made necessary by imperatives to act, particularly the imperative of government action. Thus, from the very first days of confinement (19 March, bearing in mind that confinement began at noon on 17 March), the administration presented to Parliament a draft Supplementary Budget Bill (PLFR) for 2020 that would release credits to support economic activity, and businesses in particular. As with all Finance Acts, this bill is based on hypotheses about the future macroeconomic situation. These hypotheses do not constitute a prediction of “what the future will be,” but rather a representation from which it is possible to coordinate and take action—to borrow J. Beckert’s (2013) expression, these are “fictions” that economic actors agree upon.

In the case of the PLFR, there are two hypotheses: first, that confinement measures will be limited to one month (thus adding fifteen days to the period announced), and second, that French and foreign demand will quickly return to their pre-crisis levels after confinement has ended. These hypotheses led the government to revise its growth forecast for 2020 sharply downward from +1.3% (late December 2019) to -1.0% (mid-March 2020). It is on the basis of these hypotheses and this forecast that the budget bill has been drafted. Further proof that the coordination of actors relies on adherence (a decidedly more appropriate term than “belief”) to this fiction: these hypotheses were not called into question at the National Assembly, nor the Senate, Committee meetings, or “public” sessions—yet certain measures of the PLFR were contested, even though its adoption was assured from the outset (it passed the National Assembly unanimously on 19 March, and received unanimous support excepting sixteen abstentions in the Senate).

The implementation of measures designed to curb the devastating yet uncertain effects of the health crisis on the economy therefore requires us first to portray, or imagine, the possible, if not probable, future of the economy. By serving as a guide, forecasting helps to overcome, at least in appearance, an uncertainty born of the crisis that would otherwise prohibit all action.

Lost in recession

But this very uncertainty makes forecasting even more difficult. The most striking sign of this difficulty is undoubtedly INSEE’s (the French National Institute of Statistics and Economic Studies) abstention from publishing their economic outlook note initially scheduled for 24 March 2020. Whereas under normal circumstances, at the end of each quarter institutions like INSEE produce forecasts for the following quarter as well as an estimate of the development of the main macroeconomic variables for the whole of the calendar year, the implementation of confinement when these forecasts and estimates had already been produced ruined the whole process: as the INSEE press release states, “the tightening of health restrictions makes these estimates obsolete in any case.

The COVID-19 epidemic is a reminder of the extent to which forecasting brings different temporalities into play—both a future to be predicted and a past to be described (Pilmis 2018a). Indeed, whatever its technical or statistical specifications, economic forecasting is fundamentally based on the extrapolation of recent trends as well as on reference to past events that serve as points of comparison and constitute a repertoire of “precedents” to draw on. The outbreak of such a crisis is an obvious reminder of how essential it is, as well, to identify turning points, which are supposed to make it possible to anticipate “weak” but pertinent signals that are the forerunners of a change in economic trends. It also underscores the difficulty of including these in economic analysis a fortiori when, as with a health crisis, they concern exogenous shocks, which are by definition kept at a distance from economic reasoning (Pilmis 2018b).

It is thus remarkable that an institution in this unprecedented situation, such as INSEE, would endeavor to regularly publish (on a bi-monthly basis) “short analyses of economic developments” aiming to describe the economic situation at time t—what forecasters often call “nowcasting.” These estimates are meant to enable an estimation of the macroeconomic impact of the health crisis: the figure given for activity—“65% of normal”—was thus largely taken up by the press in the days following its publication (economic update on 26 March 2020).

The Period of Revisions

Crisis situations (health, economic, political, etc.) contribute to broadening the spectrum of possible futures, making it riskier to identify one as “more probable” than others. Will confinement last for fifteen days, as initially announced, or for several months? Once it is brought to an end, what portion of consumption can be recovered immediately, and what portion will be lost for good? Will business bankruptcies lead to a significant and lasting rise in unemployment? Will the timing of the epidemic create temporal discrepancies between countries, leading to a contraction of global demand that is longer-lasting than the epidemic itself in each of the countries that make up the world’s major economies? These are all questions that the macroeconomic framework used by forecasters cannot answer (Fligstein, Brundage & Shultz 2017), yet which are at the heart of forecasting activities.

The two cases discussed here illustrate this uncertainty. Thus, it is not insignificant that in his introduction to the economic outlook note of 26 March 2020, the Director General of INSEE, J.-L. Tavernier, stated his reluctance to publish the note, in particular because a “statistician is loath to provide a result in which he is not sufficiently confident. And it is an understatement to say that what we are presenting today is fragile and subject to revision.” Likewise, the High Council of Public Finance stipulates in its notice of 17 March 2020 that “the government’s scenario presented in this [PLFR] is based on two strong assumptions—that confinement will be limited to one month and that there will be a swift return to normal in both French and foreign demand, which cannot be taken for granted—such that the deterioration of the macroeconomic context could be more significant than is forecast by the PLFR for 2020.” It is therefore unsurprising that at a hearing before the Senate’s Economic Affairs Committee, the Minister of the Economy and Finance, Bruno Le Maire, drastically revised the growth forecast at the heart of the PLFR barely two weeks after its promulgation, stressing that the recession expected in 2020 would not be “-1.0% of GDP” but probably much more than -2.2%,” the level reached in 2009. Eventually, in mid-April 2020, the French government presented a second draft Supplementary Budget Bill (PLFR 2) for 2020 that is based on a strongly revised growth forecast (-8%).

The tenuousness of forecasts, acknowledged by the very people who produce them, underscores the degree to which, in this peculiar social space, the question is not one of veracity, of “saying something that is meant to be or to become true,” but to build an instrument that enables coordination between actors. But above all, it forcefully underscores the paradox of an economic forecasting made even more necessary in the most unfavorable of contexts.

© Shutterstock/
© Shutterstock/Viacheslav Lopatin

Article written on 7 April 2020, updated on 17 April 2020

Return to the file


Olivier PILMIS