67
edits
Important: The GCConnex decommission will not affect GCCollab or GCWiki. Thank you and happy collaborating!
Changes
Jump to navigation
Jump to search
Line 9:
Line 9: + + + + + + + + + + + + + +
→Regulatory Risk Analysis Tool - Monte Carlo Simulation
In situations such as this model inputs may be best defined as probability distributions rather than an individual number. When more than one probability distribution is used in repeated simulations to calculate an objective outcome this is referred to as a Monte-Carlo experiment. This works by repeatedly and randomly selecting outcomes from the distributions of the various inputs and applying them to the model. Each time this is done, a possible outcome of the model is created. The more times this process is repeated the more observations from the total set of possible outcomes are calculated, until eventually enough data is collected to piece together a probability-distribution of the model output.
In situations such as this model inputs may be best defined as probability distributions rather than an individual number. When more than one probability distribution is used in repeated simulations to calculate an objective outcome this is referred to as a Monte-Carlo experiment. This works by repeatedly and randomly selecting outcomes from the distributions of the various inputs and applying them to the model. Each time this is done, a possible outcome of the model is created. The more times this process is repeated the more observations from the total set of possible outcomes are calculated, until eventually enough data is collected to piece together a probability-distribution of the model output.
[[File:Monte.jpg|center|thumb|910x910px|Monte-Carlo Simulation]]
This process can be seen in the above figure where three model parameters exist, and each parameter is represented by a probability distribution. By applying the Monte-Carlo method, model outputs can also be represented by a probability distribution. In the case of a CBA this process would allow us to make statements such as the probability that the Net Present Value (NPV) of a regulatory proposal will be above a certain level (such as zero), the probability that a percentage of times the NPV will fall within a certain range, the standard deviation of the NPV, the average NPV, the most likely outcome NPV, and which parameters are likely to have the greatest impact on the NPV.
Because uncertainty and risk in an outcome can be considered as a cost itself, this information adds important insight towards improving program choice and implementation of a regulation. In many cases costs need to be incurred to decrease uncertainties and thus the regulatory option with the highest net-present value may not always be the preferred option if it is important to limit our exposure to risk.
While Monte-Carlo methods have been around for decades it is only more recently with the increase of computing power that this kind of analysis has become common. Today there are many statistical programs that can perform Monte-Carlo simulations. While these packages are powerful with many features, they may not be accessible to all analysts within the Federal Government. They can be expensive and require special knowledge and training. Furthermore, the human resources required to acquire such packages through government acquisition services can become a barrier to the average analyst. It also becomes difficult to share models created with these software packages with other analysts or departments to confirm reported results or adjust parameters for further analysis as they will need to have the necessary software to do so.
It is for these reasons that I endeavoured to create a simple solution to enable Monte Carlo simulation analysis. The ''Regulatory Risk Analysis Tool (RRAT)'' works within Microsoft Excel using visual basic macros. This permits analysts to work within an environment that they are familiar with and does not require the acquisition of any additional software licenses. It also allows the tool to be incorporated into existing models in excel, increasing their functionality. While it is not as powerful a tool as other commercial software packages, it should fill the needs of most regulatory CBAs, and since it is open-source code, its capacities can be expanded and improved as required by others that may wish to do so.
