Analysts face three major types of uncertainty:
Model Uncertainty – This refers to whether the chosen model is conceptually or structurally appropriate. If the model is flawed, even perfect inputs and estimates won’t yield valid forecasts. This is the most dangerous type because it can lead to fundamentally incorrect conclusions.
Parameter Uncertainty – This arises because the model’s parameters (like coefficients) are estimated using historical data, which always involves some error. Analysts can reduce its impact by paying close attention to estimation risk.
Input Uncertainty – This involves doubts about whether the inputs to the model truly reflect reality. For example, using a proxy for the “market portfolio” in CAPM introduces input risk. Its seriousness depends on the context—it's more problematic when testing theory than when simply seeking useful patterns.
Historical examples highlight model uncertainty’s risks. During the late 1990s, investors believed historical average returns would hold indefinitely, ignoring changing market dynamics—leading to the tech bubble. In the 2007–2009 crisis, flawed assumptions about housing risks and securitization diversification failed, causing widespread financial collapse. Both show how wrong models can give a false sense of security and lead to systemic failure.