**PAD520_W5: ****Policy Analysis and Program Evaluation – Forecasting Expected Policy Outcomes**

Slide #
| Topics |
Narration |
---|---|---|

Slide 1 | Intro | Welcome to Policy Analysis and Program Evaluation.In this lesson, we will discuss Forecasting Expected Policy Outcomes.Next slide. |

Slide 2 | Topics | The following topics will be covered in this lesson:Forecasting in policy analysis: aims of forecasting, limitations of forecasting; Types of futures: goals and objectives of normative futures, sources of goals / objectives / and alternatives; Approaches to forecasting: objects of forecasts, bases of forecasts, techniques of forecasting; Extrapolative forecasting: basic assumptions, classical time-series analysis, linear trend estimation, nonlinear time series, exponential weighting, data transformation, catastrophe methodology; Theoretical forecasting: theory mapping, theoretical modeling, causal modeling, regression analysis, point and interval estimation, correlation analysis; andJudgmental forecasting: the Delphi technique, cross-impact analysis, and feasibility assessment.Next slide. |

Slide 3 | Forecasting in Policy Analysis | We will begin this lesson with discussing forecasting in policy analysis. Forecasting is the ability to use historical data to analyze future data. In policy analysis, forecasting uses historical policy problems to analyze current expected outcomes of a policy problem. The process of forecasting has three principle forms: projection, prediction, and conjecture. Using historical data, the analyst can use the first principle to project what the data will be in the future. A prediction of forecasting is when the analyst uses theory to make a prediction about what may or may not happen. A conjecture forecast is based on the analyst assumption of what may happen in the future. The purpose forecasting in policy analysis is to understand how past policies have made an impact on today, which in turn, will make an impact on tomorrow. The consequences of past policies allow the policy analyst to make judgment on what to recommend or suggest as a course of action in the future. It also addresses the behaviors of various disciplines to take action and prevent chaotic policy situations.Next slide. |

Slide 4 | Forecasting in Policy Analysis, Continued | With any type of projection, prediction, or conjecture, there are limits to how much forecasting can be accomplished. This slide provides an overview of the limitations and strengths of forecasting.Next slide. |

Slide 5 | Types of Futures | There are three types of future societal states of policy forecast. The first is potential futures, which looks at what might happen in a particular society or group in the future. Plausible futures are based on the assumption that if the government does not intervene, certain events will happen. The last is normative futures which link the forecast to specific goals using the features of potential and plausible futures. As time changes, so does the goals and objectives of an agency. Analyst who use normative futures for forecast are met with a challenge because goals and objectives today may be far different from the future, which creates a challenge for the policy analyst and agency. Normative futures require the analyst to specifically define the goals and objectives today. The goals should be defined in a manner that allows flexibility if the agency changes. The same should be done for objectives. Keep in mind, objectives are measurable and can be altered for the future. Another way to help in the process is to look at the current goals and objectives and create alternatives that may address the future of the agency.Next slide. |

Slide 6 | Types of Futures, Continued | One way to help a policy analyst to use normative futures is to identify the source or sources of the goals, objectives, and alternatives. The first source is to find out the authority of experts. The second source is using insight, intuition, or tactical knowledge about a problem from individuals who are knowledgeable about the problem but not experts of the problem. These individuals could be public administrators whom work directly with the problem situation. To assist with finding alternatives, being creative by using various methods can assist. The next source is scientific theories. As we understand, scientific theories usually derive from some natural phenomenon or social science. Another source for alternatives is motivation where individual and groups moral values are taken into consideration. Parallel case is finding a similar policy problem in another state, country, city, or communication as a source for alternatives.The next source is analogy where similarities are found in amongst different policies. The last source is ethical systems. Ethical theories allow the analyst to look at theories regarding social justice, equity, and morality.Next slide. |

Slide 7 | Approaches to Forecasting | The next step in forecasting is to determine what approach the analyst will be taking. The analyst must decide what to forecast, select one or more bases for the forecast, and choose the appropriate techniques.We will start with the point of reference or objective of a forecast. The object of forecasts has four separate objects to consider. The first is consequences of existing policies if the government decides to do nothing to estimate future changes. Next, are the consequences of new policies. In this object, forecasts are considered on the adoption of new policies. The third object is contents of new policies. This object reviews certain aspects of new policies and forecasts its changes. The last object is behavior of policy stakeholders, which lends the analyst to estimate whether there will be support from stakeholders of the new policies.Next slide. |

Slide 8 | Approaches to Forecasting, Continued | Now that we have considered the point of reference, we are ready to select a basis in which we will carry out the forecast from the set of data. There are three bases to select.The first base is trend extrapolation. Trend extrapolation uses inductive logic to state what has happened in the future will happen in the future providing there are no new policies or unexpected events to change the course of outcomes. Inductive logic allows the policy analyst to observe the historical data and make general conclusions about its future. The second basis is theory. As we know, theories are the result of deductive logic and have been tested to explain and predict outcomes about an event on the basis of another event. Deductive logic is reasoning from general statements, laws or assumptions to a particular claim or event. When a policy analyst selects a theory for its basis, the analyst uses informed judgments regarding the estimation. Informed judgments derive from abductive logic whereas the process of estimation begins with the future and works itself backward to the claim. Next slide. |

Slide 9 | Approaches to Forecasting, Continued | Regardless of what object or basis selected to use, they’re hundreds of techniques to make the selection process quicker and easier. The slide provides a useful way to simplify the process.Next slide. |

Slide 10 | CYU | |

Slide 11 | Extrapolative Forecasting | In this section, we take an in-depth look into extrapolative forecasting. This particular forecasting approach is based on some form of time-series analysis by providing the averages from one point to the next from history to future projections. These future projections rest on the assumptions of persistence, regularity, and reliability and validity of data. If there is a persistent and regular trend in the historical data, the same will show evidence in future projections. Making projections on future data based on historical data using persistent and regular trends, it is important for the forecasting to be reliable and the data to be valid. None of the historical data used should show inconsistencies. If all three of these assumptions are met, extrapolative forecasting will give positive forecasting results. If not met, the forecasting will be misleading and inaccurate.The first appropriate technique is the classical time-series analysis. The time-series analysis has four components: secular trend, seasonal variations, cyclical fluctuations, and irregular movements. Secular trend is represented by smooth-long term growth or decline. Seasonal variation represents a series that is usually short in time such as holidays of production of work that is affected by an event like snow or bad weather. Cyclical fluctuations are similar to seasonal except the time is extended over a period of years. Irregular movements are when there is no particular set time and the trend has no set identifiable trend. Next slide. |

Slide 12 | Extrapolative Forecasting, Continued | The figure above depicts a classical time series analysis of the actual number of passengers over a one hundred and fifty month time period. The solid line represents the trend. Next slide. |

Slide 13 | Extrapolative Forecasting, Continued | The figure above represents the actual number passengers over a one hundred and fifty month time period in black and the predicted numbers of passengers in green. The blue circles represent the actual number of passengers based on extrapolative forecasting.Next slide. |

Slide 14 | Extrapolative Forecasting, Continued | The next technique is linear trend estimation. This technique is about the assumptions of persistence, regularity, and reliability and validity of data. It also includes regression analysis to provide reliable estimates in the forecast data as long as the estimates are not curvilinear. Curvilinear estimates produce forecasts with unlimited errors. The regression analysis in the linear trend estimation has two important properties: deviations cancel and squared deviations are a minimum. Deviations cancel is the difference between the observed value and the value on the regression line, which should always equal zero. When the observed value lies below the regression line, the deviation is always negative. In contrast, the deviation is positive when the value lies above the regression line. Square deviations are a minimum is the most efficient way to draw a trend or slop of regression line through a series of observed data points. The figure above provides a graphical view. Next slide. |

Slide 15 | Extrapolative Forecasting, Continued | Nonlinear time series falls outside of the conditions set for linear trend estimation and has five main classes. The oscillations class shows consistency but only within a specific time period and does not show an increase or decrease in the forecast. Cycles are unpredictable and do not carry a set trend, fluctuates between longer periods of time. A growth class shows an increase growth rate, decrease growth rate, or a combination of both, which forms the shape of an S. The decline class is similar to the growth class with the exception that the forecasting shows consistency in design with a moderate increase in a small time series. The last class is catastrophe and displays a sudden and sharp discontinuation of data. Although these techniques are based on linear regression, they require transformations of the time-series variable that may include roots, logarithms, or exponents. Next slide. |

Slide 16 | Extrapolative Forecasting, Continued | The next technique is exponential weighting. The purpose of this technique is to forecast how the data increases at an increasing rate by writing a new and explicit nonlinear regression. This technique is projected by squaring the value of x. The more the increase or decrease the higher the power necessary to represent it. The purpose of data transformation is to make the application of the assumptions easily to be read and understood. Many times, the data from the linear regression can be misleading and challenging to determine its accuracy. By transforming the data, it gives the analyst a clearer picture in the reliability and validity of the assumption. To transfer the data, the untransformed value is squared and then you take the logarithm of the squared value. The table on the left shows the transformation of data. The graph on the right shows how the data looks untransformed from the graph on the left and transformed from the graph on the right.Next slide. |

Slide 17 | Extrapolative Forecasting, Continued | In every extrapolative technique exist some type of catastrophe. Catastrophe methodology happens when there is an unexpected event. This small change will have a drastic impact on the results. The policy analyst should be aware of a few assumptions and applications of the catastrophe methodology.Discontinuous processes are when the physical, biological, or social domain has an unexpected event or sudden shift in the forecast. When there is a sudden shift in any system, such as a social system, it can change the structure of the system as a whole. This is call systems as wholes. Incremental delay in the catastrophe methodology involves ongoing small changes to routines and practices. The incremental changes present ongoing sudden changes. Some incremental changes are the result of incomplete information. It is also used to gain public support during political debates.When incremental delays affect existing policies, it may become challenging to make a decision until the incremental change stops and the results are catastrophic. At that time, policy makers are forced to make an immediate policy change.Next slide. |

Slide 18 | Theoretical Forecasting | In this section, we move our attention to focus on the future cause and effect of propositions and laws as it impacts the societal states using deductive reasoning. In policy analysis, deductive reasoning is most frequently used in connection with casual arguments to prove if one event happens another event will follow. This is a distinguishing feature in theoretical forecasting because predictions are deduced from propositions and laws. The argument of persuasion is greatly increased when deducted reasoning is observed over a long time period. There are several procedures that can assist analysis in making theoretical forecasts. The first is theory mapping. Theory mapping is similar to a flowchart that identifies and arranges assumptions in four types of casual arguments. These casual arguments are described as convergent, divergent, serial, or cyclic. Convergent arguments have two or more assumptions that support one conclusion. A divergent is when a single assumption supports more than one conclusion. A serial argument has one conclusion that is used as an assumption to support further conclusions. A cyclic argument is serial arguments where the last conclusion is connected with the first conclusion in the series. Because a theory may contain a mixture of casual arguments, there are several procedures that may be used to uncover the overall structure of an argument. The procedures are separate and number assumption, underline the words that indicate claims, when specific words have been omitted but implied, supply the appropriate logical indicators, and arrange numbered assumptions and claims in an arrow diagram that illustrates the structure of the argument. Next slide. |

Slide 19 | Theoretical Forecasting, Continued | Another procedure for theoretical forecasting is theoretical modeling. Theoretical modeling presents a diagram of theories regarding a particular claim. The theories allow the policy analyst to predict the outcomes of a claim. The figure on the right is an example of theoretical modeling.Casual modeling is an extension of theoretical modeling except it allows the policy analysts an attempt to explain and predict the causes and sequences of a claim. Casual modeling is mostly used in economic, social and political public policies. One main statistical procedure in casual modeling is the use of path analysis that uses multiple independent variables. The figure on the left is an example of casual modeling. Next slide. |

Slide 20 | Theoretical Forecasting, Continued | Regression analysis, which was previously discussed, has the same use in theoretical forecasting. Regression analysis provides summary measures of the pattern of a relationship between an independent variable and a dependent variable, assist the budget analyst to decide which variables is the cause of the other, provides estimates of predicted future relationships, and calculate the probable error in an estimation. Because the distances between the regression line and individual data points are minimized, the regression estimate contains less error than other kinds of estimates. On a regression analysis graph, the greater the scatter, the less accurate is the estimate. In the regression analysis, the standard error of estimate gives a probable estimation of the forecast data. The budget analyst has the opportunity to validate the standard of error by making a simple point estimation that produces a single value. The point estimate is the best value of the parameter. If the budget analyst is not satisfied with the single value, he or she can create an interval estimation which provides a measure of accuracy around the point estimate. Another feature of theoretical forecasting is the correlation analysis. Correlation analysis interprets the relationship, direction, and strength of the forecast data. There are two measures in the correlation analysis to help support the relationship. The coefficient of determination is a summary measure of the amount of variation in the dependent variables explained by the independent variable. The coefficient of correlation is the square root of the coefficient of determination. If the coefficient of correlation is zero, there is no relationship, if the coefficient of determination is between zero and one, then it is a positive correlation. Next slide. |

Slide 21 | Judgmental Forecasting | The last function of forecasting is judgmental. Judgmental forecasting are based on intuitive arguments from assumptions made by individuals. These assumptions lead to claims about the future. Because many governmental policies are ill-structured, many of its alternative policies and its consequences are unknown or cannot be determined and there are no relevant theories and or empirical data to make an accurate forecast. The Delphi technique is one way to perform judgmental forecasting. This technique was developed in 1948 by the researchers of the RAND corporation as a forecast strategy by a team of experts around the world. The Delphi technique brings together the opinion of a group of individuals to survey on a particular policy issue. Each of the participants remain anonymous, is selected based on interest and knowledge and not expertise, measures the answers based on disagreement and conflict, structures conflicts by exploring alternatives and their consequences, and, if possible, conducts a series of anonymous computer conferencing.An example of the Delphi technique is shown in the above figure. After the initial round, the number of remaining rounds is determined by the analysis of the initial round. If the results need to be more defined and or descriptive, another round of questioning is conducted. At the completion of the rounds, the results are made in preparation for the final report. Next slide. |

Slide 22 | Judgmental Forecasting, Continued | Also developed by the RAND corporation is the cross impact analysis. The purpose of the cross impact analysis is to identify how one event will assist or hinder the occurrence of related events. The figure above is a cross impact matrix that is used in the cross impact analysis. Based on the results or sum of the matrices, the policy analyst will be able to determine the positives or negatives of the linkages of the events. A linkage of events provides the direction and strength of the forecast and the amount of time between the occurrences of linked events. Review the above figure, each block must be completed with a number. For example, the impact of output on job quality is high. A number three is placed under the seventh column and first row. Once all the spaces are completed, the sum of the numbers is added for the X axis for active and the Y axes for passive. The results of the cross impact analysis provide a judgmental forecast. The advantage of the cross impact matrix is that it enables the analyst to discern interdependencies that otherwise may have gone unnoticed. As new events happen over time, the matrix can be redone and recalculated. Next slide. |

Slide 23 | Judgmental Forecasting, Continued | The final part of this section is feasibility assessment. The purpose of a feasibility assessment is to assist the analyst in producing estimates of probable consequences regarding policy alternatives as a result of political conflict and unequal distribution of power and other resources. It assesses the forecast behavior of stakeholders under conditions of political conflict and enables the analyst to consider the sensitivity of policy issues, its available resources, and changes in policy alternatives. It also helps to demonstrate the impact of stakeholders in support or opposing the adoption and or implementation of different policy alternatives. The feasibility assessment focuses on several aspects of political and organizational behavior. The figure above is an example of a feasibility assessment template. The criteria of the policy are listed in the first column. The options are the programs or policies being assessed. The score is the issue position and coded as supporting as a positive one, opposing as a negative one, or indifferent as a zero. The probability is the estimate which ranges from zero to one and indicates the importance of the issue to each stakeholder. The resources available column is ranked from zero to one and shows the subjective estimate of the resources available to each of the stakeholders in pursuing his or her respective position. The resource rank column, which also ranges from zero to one, determines the relative rank of each stakeholder with respect to his or her resources to support a policy. Based on the scores of the feasibility column, the highest scores provide the best guessed estimated of which policy or program to implement. Next slide. |

Slide 24 | CYU | |

Slide 25 | Summary | We have reached the end of this lesson. Let’s take a look at what we’ve covered.The start of the chapter discussed the policy analyst role in forecasting policy outcomes, its uses on various levels of the government, and how forecasts are characterized as potential, plausible, and normative. The approaches to forecasting are based on inductive, deductive, and abductive reasoning. The most widely used reasoning in forecasting is deductive because it provides reasoning from general statements, laws, or propositions to particular claims and allows the policy analyst to make informed judgments.There are three different types of techniques in forecasting, extrapolative, theoretical, and judgmental. Extrapolative forecasting provides the assumption that the data is persistence, regular and reliable and valid. A trend analysis is performed on transformed data to demonstrate the relationship between the data and its estimate outcomes.Theoretical forecasting using various mapping models to demonstrate the forecast outcomes of policy issues. A regression analysis, point and interval estimation, and correlation analysis are techniques to be used in assisting the policy analyst identifying outcomes.Judgmental forecasting uses the Delphi technique, cross-impact analysis, and feasibility study to forecast the behavior of stakeholders in ill-structured policy issues. These techniques assist the policy analyst in estimate true beliefs about the future of policy and help to eliminate wrong judgment.This completes this lesson |