Scenario Analysis

From Pardee Wiki
Jump to navigation Jump to search

Scenario Description

A scenario is a story or story outline. Thinking about the future normally involves creating alternative scenarios, or stories, about the possible interactive evolution of variables. Some such scenarios are exploratory and consider the possible unfolding of different futures around key uncertainties, such as the rate of some aspect of technological advance or the fragility of some element in the global environment. Other scenarios are normative and develop stories about preferred futures, such as a global transformation to sustainability.

Scenarios in a computer model typically are built from multiple interventions that collectively help create a coherent story about the future. Often, but somewhat imprecisely, the word scenario is used more loosely to refer to any intervention (such as the change of a fertility rate for a country or an alternative assumption about oil resources).

Scenarios or interventions with respect to what? When IFs or other computer simulations are "run", without making any changes to parameters or initial conditions specified as the default values, they generate a forecast that is typically called the Base Case. The IFs Base Case, always available when a model session is initiated, is itself a scenario. Sometimes the Base Case is incorrectly referred to as a trend extrapolation or a "business as usual" scenario. More accurately, however, the IFs Base Case is a computation that involves the full dynamics of the model and therefore has very nonlinear behavior, often quite different from trends. It is a good starting point for scenario analysis for two reasons. First, it is built from initial conditions of all variables and on parameters that have been given reasonable values from data or other analysis. These initial conditions and parameters make up the package of interventions that constitute the Base Case scenario. Second, the Base Case is periodically analyzed relative to the forecasts of many other projects across the range of issue areas covered by IFs and is to a degree "tuned" for internal coherence and consistency with insights of respected forecasters.

There are two file types involved in IFs scenario creation: scenario files (or Scenario-Load-Files) and run files (or Run-Result-Files). Scenario files, the first type, are saved with an extension of .sce. Very small in size, .sce files contain information that the IFs model uses to create alternative scenarios; i.e., .sce files contain a list of parameter values that diverge from the Base Case. It is important to note that scenario files do not contain any forecasts. Forecasts are generated and saved only in the second type of file, run files with the .run extension. Because they contain forecasts of all IFs variables and parameters, .run files are much larger than scenario files. Although the IFs standalone model software allows users to save both types of files, web users are only able to save .sce files to retrieve their parameters and regenerate their scenarios.

In addition to the Base Case, most versions of IFs will include a number of other previously-run scenarios (see Lesson 0: IFs Vocabulary for additional important terminology), typically those for the Global Environmental Outlook (GEO) of the United Nations Environmental Program (UNEP). If you look, for instance, at the Flexible Displays form, you will see a list of previously-run scenarios in the box at the bottom of the screen. Because those have already been run, based on a set of interventions constituting their foundations, the user can immediately display their results.

== Quick Scenario Analysis with Tree Overview == === Overview === === What is a scenario? === This section of the help menu will guide you through a brief description of a scenario as well as the Base Case used in IFs and will discuss the difference between Scenario-Load-Files (.sce files) and Run-Result-Files (.run files). After you have finished with [[Scenario_Description|Scenario Description]], you should be able to answer/do the following: *What is a scenario? *What is the Base Case of IFs? *What is the difference between a Scenario-Load-File and a Run-Result-File? === Introduction to the Scenario Tree: === The Quick Scenario Analysis with Tree allows you to call up or to mix and match an extensive number of your own interventions and/or a set of stored scenario intervention files. This feature of IFs allows you to change any parameter and selected initial conditions used in the model, thus shaping forecasts. Use the Quick Scenario Analysis with Tree to create Scenario-Load-Files and/or run scenario files through IFs in order to create Run-Result-Files that you can use throughout IFs. Below is the main menu of the Quick Scenario Analysis with Tree: IMAGEhttp://www.du.edu/ifs/help/use-online/scenario/quick/overview.html === Loading Previously-Structured Scenarios === The Quick Scenario Analysis with Tree allows you to load the previously-structured scenarios and user-saved scenarios. From the Scenario Files or the Adding Scenario Component menu options of the Quick Analysis with Tree, you can load a wide range of pre-packaged scenario intervention files and see what interventions were made in the files. After you have finished with this topic http://www.du.edu/ifs/help/use-online/scenario/quick/loading.html , you should be able to do/answer the following: *How does a previously-structured scenario differ from a previously-run scenario? *What previously-structured scenarios came installed in your version of IFs? *How does one search through the scenarios until you find one that deals with environmental change (or some other specific topic of interest)? *What do you have to click on to understand exactly what is being changed by different previously-structured scenarios? === Finding the Intervention You Want === In order to tailor your scenario file to your needs, you must be able to quickly find the parameter you are looking for. After you have finished with this topic http://www.du.edu/ifs/help/use-online/scenario/quick/finding.html , you should be able to do/answer the following: *What is the organizational logic of the Scenario Tree? *What is the difference between the Selected Initial Conditions, Relationship Parameters and the other five main categories used in the Scenario Tree? *How would you search for a specific parameter that might help shape the scenario intervention of interest to you? === Exploring and Changing Parameters === Once you have found the parameter you are looking for, for instance, the total fertility rate multiplier (Households/Individuals, Demographic/Population, tfrm), a number of new options become available. After you have finished with this topic http://www.du.edu/ifs/help/use-online/scenario/quick/exploring.html , you should be able to do/answer the following: *How do you select a parameter to change? What do the Select, Drivers, Explain, View Equations and Define pop-up options (when you click on a parameter name from the tree) all allow you to do? *How do you clear parameter changes from the Scenario Tree? *How do you create a Run-Result-File? *How do you save the results? *How can you display the results of your change in IFs? === Activate Pre-run Scenario(s) for Display === Although not accessed through the scenario tree, it is important to know that hundreds of scenarios (the parameters for which are accessible in the tree) have been pre-run with the full forecasting results stored on the IFs server. It is not necessary for you to load the parameters and re-run these scenarios unless you want to create variations of them with new parameters or new parameter values. Instead you can go to the Display/Activate Pre-Run Scenario for Display http://www.du.edu/ifs/help/use-online/display/activate.html sub-option of the Main Menu, choose the pre-run scenario, and then display results from it using any IFs display option. === Loading Previously-Structured Scenarios === Many packages of scenarios in IFs come pre-built. These previously-structured scenarios (that is, .sce files) are helpful components for building alternative futures in the model. They contain changes made in parameters but not yet in computed variables. They can be run to generate the forecasting results for all IFs variables (but if you are making no changes it will normally be easier simply to activate the results http://www.du.edu/ifs/help/use-online/display/activate.html of running the .sce or scenario file). More often on the web-based version of IFs you will load an .sce file in order to make changes to it or to combine multiple .sce files. There are two different ways of loading .sce files: the Add Scenario Component option (usually the easier of the two options) and the Open/Other sub-option under Scenario Files. Although the format of the two is different, the functionality is the same. From the main menu of Quick Scenario Analysis with Tree (reached with the Scenario Analysis option from the Main Menu of IFs), select Add Scenario Component and you will be presented with a drop-down list with the names of categories in which you can find different previously-structured scenarios, as shown below (the ultimate scenario names are often in sub-categories or even sub-sub categories because there are a large number of previously structured scenarios in IFs for your use). These scenarios can also be reached by clicking on the Scenario Files option (sub-option Open). IMAGE http://www.du.edu/ifs/help/use-online/scenario/quick/loading.html Any of the previously-structured scenarios can be loaded into the Scenario Tree of IFs. You can also create a larger scenario in the tree by adding several smaller scenarios together. Once you have loaded a scenario into the tree (experiment by doing so), if you would like to know more about the parameter changes from the Base Case that make up the scenario, click on Explain Scenario located on the menu. Besides these previously-structured scenarios, hundreds of .sce files have been pre-run with the results saved as .run files (so called previously-run scenarios) on the web. Because previously-run files have already been run (as the name indicates) users can immediately display the results through different display options such as the Flexible Display and the Self-Managed Display. The menu that provides access to previously-run scenarios is the Activate Pre-Run Scenario for Display http://www.du.edu/ifs/help/use-online/display/activate.html sub-option under Display from the Main Menu of IFs. By simply clicking the scenario name that the user would like to load, the activate feature would call it up. After activation, the user can see it in the scenario lists of the Flexible Display, for example. It is possible to activate multiple different scenarios by selecting them individually. === Parameter Types === Parameters are numbers that determine relationships among variables in the equations of IFs. You often set parameters to a single value across time and they therefore do not always "vary" as do "real" variables. Many parameters are "policy handles." More generally, parameters can actually be thought of as a special type of variable, the value of which you set in order to determine the behavior of the model. See the IFs project document called Guide to Scenario Analysis for much more information on parameter types and especially on the important parameters of IFs, organized by the models of the IFs system. Multipliers: They have a normal value of 1, and to increase whatever they multiply (say agricultural yield) by 50 percent you increase the parameter to 1.5. To decrease it by 25 percent you would decrease the multiplier parameter to 0.75. You will almost always spread such changes out over time, keeping the multiplier's value at 1 in the base year and gradually increasing or decreasing it over a period of years. You should almost never change a multiplier in the initial year because the model is set up to provide accurate results for that year and will compensate for and thereby offset your change. For instance, if you set a multiplier on agricultural production equal to 1.5 for the first year and all years thereafter, you might find that the results were no different than in the base case. You must instead gradually introduce your change, preserving the multiplier value of "1" in the initial year. Examples of multipliers include: agdemm, enpm, freedomm, mortm, protecm, qem, rdm, resorm, tfrm, and ylm. Note that multipliers typically end with the letter "m". Additive Factors: Most have a normal value of 0, thereby leaving that to which you add them unchanged. How much you would add to achieve a 50 percent increase might depend on the amount to which you added it to. Some additive parameters are, however, applied multiplicatively to the quantity they modify (that is, 1 plus the parameter is multiplied times the quantity), thereby scaling the parameter. In that case, the base or normal value of the parameter will be zero, but one can achieve a 50 percent increase in the quantity modified with a value of 0.5 and a 50 percent decrease with -0.5. You will very seldom want to change the base year value of additive parameters because it will either incorrectly change model results in the base year or, more likely, will result in model compensation to protect initial model results. An example of an additive parameter is: xshift (export shift as result of promotion of exports). Although early versions of IFs used additive factors and multipliers with comparable frequency, most additive factors have been replaced by multipliers to standardize most parameter changes. Exponents: For instance, many "elasticities" raise something to a power. For these parameters the "normal value" will vary greatly, but they will most often fall between -2 and 2, with many clustering around 0. In most cases it will make sense to change these parameters for all years including the first - generally the model will not use them in the first year and they will affect results only in subsequent years. Elasticities in IFs include: elass and engel. Reactivities: These are factors that relate growth in one process to growth in another. Although many will range between -2 and 2 (with 0 eliminating linkage of the processes), some have very large values. They are very much like elasticities, but the formulations that use them do not have exponential form. Reactivities include: cdmf, cpowdf, cwarf, nwarf, and reac. Growth Rates: It is possible to force some processes to grow at specified rates. More commonly, the specified rates serve as targets and the dynamics of the model often shift actual growth rates somewhat, necessitating experimentation with targets to achieve a desired growth. Examples include: eprodr and tgrld. Allocating Coefficients: Coefficients are often used in multiplicative relationships with other variables, but many such coefficients are not what were earlier called multipliers (with a base value of 1). Instead they can serve an allocating role. For instance, you can use parameters to allocate governmental spending to health, education, and the military. Allocating coefficients frequently have values between 0 and 1. Again, you should generally not change these parameters in the initial year because the model will often compensate for changed values in the first year. Instead, change them by series over time. Allocating coefficients in IFs include: aidlp, carabr, drcpow, drnpow, nmilf, and rfssh. Transforming Coefficients: Some coefficients transform units of variables or link variables in other ways. Examples in IFs are: carfuel1, carfuel2, and carfuel3. Switches: These parameters turn something on or off. They generally take on values of 1 (on) or 0 (off), but can have additional settings. For instance, some switches not only turn on some process, but set a key value within it (like the level of energy exports). Switches are most often on or off for the entire run, but it sometimes makes sense to "throw a switch" in the middle of a run. Switches allow you to fundamentally alter the structure of a model. Switches include: actreaon, agon, ally, enon, enprix, and squeez. Variables: This category should technically not be called parameters at all. They could and would be computed endogenously, if the model included the appropriate theoretical structure. They generally do not determine the interaction of other variables. Such variables include: AQUACUL and EDPRIPTR. Initial Conditions: Again, these are not strictly parameters, but rather first-year values for variables subsequently computed by the model. Although many initial conditions, like the population (POP) of the U.S., are sufficiently well-known that they should not be changed by model users, others, like the ultimate availability of oil and gas resources are only reasonable guesses. Thus users should feel free to change some initial conditions based on new data or even simply to test the implications. This category includes a great many variables, such as: AIDSDTHS and RESER. The focus here is on exogenous parameters only - on those elements of the model that you can change. Many computed variables are used in the computation of other variables in the same way that parameters are, as multipliers, additive factors, coefficients, and so on. You can display those, but unlike true parameters, you cannot change them.