Forecasting models are simulations that use current economic conditions in order to predict economic activity in the future. Recently, they have been brought into the spotlight following the IMF’s acknowledgement that short term fiscal multipliers used in its forecasts were underestimated (see Update 83). Fiscal multipliers are an estimate of how changes to government spending or taxation impact on overall economic growth. The IMF recognised in its 2012 World Economic Outlook (WEO) that its own forecasts had been systematically over-optimistic in a number of countries and regions. The over-optimistic forecasts for the growth rates of Greece and Portugal led to the IMF altering its policy stance. Portugal was offered a longer time frame to achieve agreed targets for its debt levels relative to GDP while the Greek lending programme initiated in 2010 was replaced amid lenders having to agree to cancel portions of Greece’s outstanding debt and delay repayment, in order to render the programme sustainable. Hence over-optimistic forecasts bring the legitimacy of the IMF’s forecasting models into question.
The IMF’s surveillance and forecasting work relies heavily on forecasting models which underpin the analysis in the WEO, such as the Global Projection Model (GPM). The GPM is actually an amalgamation of several models; principally the Quarterly Projection Model (QPM) and the Dynamic Stochastic General Equilibrium (DSGE) model. The DSGE model – a commonly used macroeconomic model – uses current real GDP and inflation. Additionally it incorporates the assumption that interest rates are set according to the desired inflation target and that an interest rate parity condition for the exchange rate exists, meaning returns on capital are uniform across countries. The formulas that the IMF uses for its models are widely available but the forecasts are then altered through consultations with IMF staff from various area departments, a process which is not made public and has led to accusations of politicisation.
The IMF’s forecasting is accused of being not only politicised but also susceptible to internal wrangling, for example in the case of Argentina where a 2007 study by US institute the Center for Economic and Policy Research suggests the Fund tampered with its predictions to support its policy positions. While the Fund endorsed Argentine economic policy prior to the 2001 crisis it simultaneously over-estimated Argentina’s future growth. Following Argentina’s decision to default and devalue its currency in the teeth of IMF opposition, the Fund subsequently underestimated Argentina’s economic growth. A 2008 academic study found that the Fund regularly engaged in what was termed “defensive forecasting”, whereby over-optimistic predictions are made either to ensure the financial commitments of the Fund’s shareholders are renewed or to uphold its mandate of preventing financial crisis, rather than releasing pessimistic forecasts risk creating self-fulfilling prophecies of crises. They also found that countries which tend to vote in line with the United States in the United Nations General Assembly were rewarded with “better inflation and… growth forecasts”.
Assumptions underpinning forecasting models
The DSGE approach is to conceive of how individuals act at the microeconomic level, and then to assume this is consistent for the whole economy it aims to describe. The DSGE is a so-called ‘representative agent’ model; it forecasts predicted behaviour of many people but assumes that each and every individual’s preferences and attitudes are identical. It also assumes that groups of people behave identically to any of these identical individuals. Each individual is assumed to be fully representative of the behaviour of the entire group not because this is observed amongst real people but because it permits the model to compute interpretable results.
Mainstream models have a number of additional but crucial assumptions about individuals and how they interact. These assumptions are also frequently criticised as unrealistic and producing meaningless results, but are relied upon because they permit the model to give clear results and thus policy recommendations. These include assuming that there are no ‘coordination problems’ amongst individuals, meaning that each individual’s choices and incentives will not conflict and do not require coordination or outside intervention to make the best possible outcome occur. Individuals are assumed to be completely self-interested and hyper-rational, thus they are presumed to have an instant and infinite ability to compute what is in their economic interest, and do so consistently without error. These models also rely on the assumption that all individuals’ expectations of future economic conditions are never systematically wrong and are based on all possibly relevant information to reach these unerring decisions. This assumption is made to ensure mathematical models function, rather than to reflect real behaviour.
Forecasting models used by the IMF based on these assumptions are often derided for being unrealistic. Alternative approaches have attempted to address this for example by modelling empirical observations of behaviour which have revealed people’s economic choices to be non-rational, inconsistent and heterogeneous. However this has led to mathematically complex models that yield multiple or unclear results, which have failed to displace the DSGE approach favoured by the IMF.
Furthermore, implicit in most forecast models is a measure of the output gap – the difference between actual and potential GDP. In 1999 while at the World Bank, William Easterly of New York University questioned the Fund and the Bank’s insistence that the output gap has to be filled by increasing investment. Forecasting models rely upon the assumption that there is a linear or proportional relationship between investment and growth when in actuality, he found that these underlying assumptions were not borne out. As a result, he challenged the policy orthodoxy for promoting growth: “increases in investment are neither necessary nor sufficient for increases in growth over the short-run”.