Modern economics uses “scientific” methodology, under the assumption that economic laws are invariant across time, space, and society. In previous posts, we saw how this leads to loss of precious insights about money gained from historical experiences (Monetary Economies: A Historical Perspective, Lessons from Monetary History: The Quality-Quantity Pendulum). In this post, we will discuss the modeling strategy we will use to derive lessons from history which extend beyond the particular historical context from which they are derived.
Models are simplified representations of reality. When considering monetary history, the factors driving changes are notably intricate. Over the 20th century, the monetary system underwent significant transformations. World War I marked the breakdown of the gold standard, followed by unsuccessful attempts at restoration. World War II further hindered restoration efforts, leading to the Bretton Woods agreement and the adoption of a gold-backed dollar standard. Nixon’s actions in 1971 severed the link between the dollar and gold, ushering in an era of floating currencies detached from commodities. Distilling broader lessons from the complexity of historical specifics necessitates a methodological approach centered on models. Models abstract from specific historical details to illuminate structures which may be widely applicable across various historical and temporal contexts.
In this text, we will be using realist models – these differ greatly from the nominalist models used in conventional textbooks of economics. The difference can be explained as follows. Our world comprises observable phenomena as well as underlying structures that produce these observations. Nominalism holds that models should focus solely on explaining observables, disregarding whether they accurately reflect hidden reality. This notion, though counterintuitive, emerged due to the belief that hidden reality is unknowable, making the pursuit of matching it futile. Instead, nominalism advocates for assessing a model’s success based on its ability to explain observed phenomena. Conversely, realist models strive to mirror the hidden reality behind observations.
Friedman’s essay on “The Methodology of Positive Economics” strongly advocates the use of nominal models. This methodological principle has been widely accepted by economists. Friedman illustrates nominal models with the example of a skilled pool player. He suggests that even if the player lacks any understanding of physics, assuming knowledge of the laws of physics can lead to accurate predictions of their shots. In essence, the player behaves as if they comprehend physics, making successful shots based on calculations, despite their ignorance of the underlying principles of physics. This is known as the “as-if” methodology, and it is the dominant approach to models in modern economics.
In contrast, realist approaches reject such assumptions. For the pool player, a realist model might study his past experiences, and his skills at different types of shots. Realism aims to understand the internal workings of hidden reality, while nominalism accepts models that predict outcomes, without concern about matching hidden reality. Friedman developed his as-if theory in response to empirical surveys which showed the most firms do not maximize profits. He argued that the assumption of profit maximization, even if it did not match the motivations of the managers of the firms, should be assessed on the basis of its ability to predict decisions about hiring and production. However, by now, this methodology has been in use for several decades, and it has led to repeated failures. A good fit to observations for a particular finite set of data is not a guarantee of the validity of a model. It can, and often does, happen by chance. For a more detailed discussion of the superiority of real models to nominal models in the context of econometrics, see “A Realist Approach to Econometrics” (bit.ly/azrae)
We will use a recently introduced modeling strategy — Agent-Based Models (ABMs) – which has not made its way into mainstream methodology. ABM models have multiple agents – laborer, producer, shopkeeper, government, etc. – each of which has their own economic decisions to make, and behavioral patterns. This strategy has become feasible because of the vastly increased computational power now available, which permits us to run simulations and compute outcomes. The foundations of modern economic methodology, established around the mid-20th century, relied on simplifying assumptions to facilitate manual computations. For example. Leading macroeconomic models have only one agent, who has perfect foresight. Why? Because computations by hand would be impossible with two or more agents. At a Congressional inquiry into the failure of economists to predict the Global Financial Crisis of 2007, Solow testified that the GFC was caused by large scale deception and fraud. Macroeconomists could not predict it because these are impossible in models with only one agent. In contrast, models with heterogenous behavior are much better at capturing the complex internal structures of modern economies, in accordance with the principle of realist models.
Using ABMs, we can capture three Keynesian insights, all of which are essential for the understanding of money, and all of which are missing from conventional textbooks:
Complexity: This technical term refers to a situation where the group behaves very differently from the individuals within the group. For example, that even though laborers and the firms which hire them may seek to lower the real wage, they can only negotiate on the nominal wage. The real wage involves the price level of the economy which is out of their control. Keynes argued that lowering nominal wages at the micro level throughout the group may end up increasing the real wage – a perfect example of complexity. This phenomenon is beyond the reach of conventional economic theory because the textbook models are oversimplified to prevent the occurrence of complexity.
Radical Uncertainty: In a model with heterogeneity, each agent has access to a limited amount of information. The economic outcomes depend on the actions of all agents, which can never be known to the agents. As a result, the agent operates in an environment where the outcomes of the decisions he takes are not predictable. Standard textbook models use intertemporal optimization, where the agent knows his future incomes, potential consumption bundles and prices. This is simply impossible in our agent-based models. Similarly, profit maximization is impossible for firms because they incur production costs in current period, but will produce and sell goods in the next period. The price at which they can sell will depend on decisions others make, and cannot be predicted. So profits are subject to radical uncertainty, and cannot be maximized.
Non-Neutrality of Money: Once we take into account heterogeneity and uncertainty, new insights into the role of money emerge, not available in conventional textbooks. Workers save money, and firms acquire money profits, but, due to radical uncertainty, no one knows what the value of money will be in the next period. A stable value of money allows for some degree of planning, but the QQ-pendulum shows that this stability cannot be relied upon. The assumptions of full information made in conventional textbooks make money merely an accounting unit, which does not play an essential role in the economy. However, with radical uncertainty, and differential information and behavior of different agents, money plays an essential role in the economy. Workers save money as insurance against adverse outcomes in the job market, and firms save money to guard against future losses. These different motivations for holding money, and the psychological aspects which relate to public trust in the future value of money, will come to the fore in our ABM models.
To wrap up, we have discussed two types of models – nominal and real. Nominal models dominate mainstream economics, and are judged for their ability to match observations. In contrast, Realist models are judged on whether or not they match the hidden structures of reality which produce the observations. In the next section, we will build some simple realist monetary models, and show that these produce results and yield insights outside the range of orthodox monetary models.
Links to Related Materials
- Bit.ly/ME01 Monetary Economies: A Historical Perspective
- Bit.ly/MONE02 Lessons from Monetary History: The Quantity- Quality Pendulum
- Bit.ly/MONE03 Nominal and Real (Monetary) Models
- Bit.ly/WEAmar Models and Reality
An Islamic WorldView 