Economic Questions (Calculation Based)

flyingpig

Class10.pdf

Home >Business & Finance homework help >Economics homework help >Economic Questions (Calculation Based)

TECON 480 – Class 10

August 19, 2019

Part 3 – Undertaking the analysis

• Instructions available on Canvas for the following:

• Assignment #3 (due tonight at midnight)

• Assignment #4 (due Wed, Aug 21 at midnight)

• Final report (due Sat, Aug 24 at midnight)

• Includes template to use as starting point

• Final Exam due Sun, Aug 25 at midnight

• No class on Fri, Aug 23 2

Announcements / Reminders

Today’s Class

• Discuss how to obtain inputs in the analysis

• Discuss two approaches for estimating willingness to pay

• Revealed preference

• Stated preference

Predicting & Monetizing Impacts

• In Steps 4 & 5, we must predict the quantitative impacts of a policy/project and use appropriate values to convert these

“physical” impacts into monetary values

• We now discuss the most common sources of information for predicting policy impacts

• We will then turn our focus to valuing these impacts with shadow prices:

• Stated preference techniques

• Revealed preference techniques

Predicting Impacts

• Our goal when analyzing a project or a policy is to provide a comprehensive CBA

• Often, the primary basis for prediction is what has happened in the past

• Can use historical or statistical data to assess the consequences of past policy changes

• Similar policy interventions that have been evaluated can provide some guidance for prediction

• However, three major sources of error are possible using this approach

Predicting Impacts

1. Omission errors

• The exclusion of impacts with substantial costs and benefits

2. Forecasting errors

• Arise simply because we cannot predict the future with certainty

3. Valuation errors

• We often do not have precise estimates of appropriate shadow prices for converting each predicted impact into an opportunity cost or a willingness to pay

Predicting Impacts

• Analysts must often be clever and creative to complete comprehensive CBAs

� “Art” & “Science”

• Should also anticipate (and explicitly acknowledge!) the errors inherent in their efforts and consciously assess them to the

greatest extent possible

• Practically speaking: do as well as you can with the knowledge you have, but also attempt to detect sources of errors and try

to evaluate their potential consequences on the CBA

Predicting Impacts

• How do we predict the quantitative impacts of a policy/project?

• Can simplify things by predicting incremental impacts relative to the status quo

• The status quo can be the best alternative when none of the other alternatives dominates it

• Therefore, the status quo must be included in the analysis

• Consequently, impacts should be predicted relative to the status quo

• This means in practice that analysts do not need to assess impacts that will be the same under the status quo

� Only differences have to be included

Predicting Impacts

• Can also predict impacts using data from an ongoing policy

• Sometimes the relevant question is whether a policy in place should be continued, terminated, or replicated

• Inferences about impacts of the ongoing policy can serve as the basis for predicting future impacts

Predicting Impacts

An example: suppose that you are evaluating the impact of

fertilizer on blueberry production

• Consider two groups:

1. Consumers who purchased the fertilizer from Lowe’s in the

spring

2. Residents who were mailed a free sample in the spring

Do you think that a comparison of the output of these two groups is a valid

estimate of the impact of the fertilizer?

Predicting Impacts

• Best basis for prediction:

inferences drawn from an experimental design with random

assignment of subjects into treatment and control groups

Predicting Impacts

• Suggest how you could come up with treatment and control groups to estimate the impact of the following:

1. A new drug treatment for pancreatic cancer

2. An increase in the tax on cigarettes

3. Class sizes in elementary schools

Predicting Impacts

• More often, inferences must be made in the absence of a true control group through a quasi-experimental design

• Often, we cannot have a control group or we cannot have random assignments in groups (why do you think this might be the case?)

• These designs address the fundamental question about the observed impacts of a policy:

• “Compared to what?”

• That is, what would have happened in the absence of the policy? • The counterfactual

Predicting Impacts

• Another approach: predict impacts based on an evaluation of a similar policy

• A very similar policy may have already been evaluated

� Could be a good basis for comparison

• Example: forecasting effects of ST3 by using measured effect of past transit investments in other cities

• Predictive power of this comparison basis depends on how closely it matches the policy being considered and how well the evaluation was executed

� This approach requires external validity 14

Predicting Impacts

• Three questions that you should ask yourself in order to determine the degree of comparability:

1. Does the policy have the same underlying model?

• Usually, only comparisons of policies with the same underlying model will be appropriate

2. How closely do the details of the policies conform?

• Even if the underlying models are the same, the details of the policy may differ in terms of types and intensity of intervention

3. What is the quality of the evaluation of the similar policy?

• Is it based on an experiment? If not, does the quasi-experimental design provide a good basis for inference?

Predicting Impacts

• Even when the evaluation design is good, there is still a high level of risk in basing prediction on a single study for several

reasons:

1. Published study may show an unrepresentative effect

• Academic journals are biased towards publishing studies with statistically significant results

• May be (many?) other studies that did not find statistically significant effects

Predicting Impacts

2. Optimism bias

• People tend to bring cognitive biases to their decision making, including forecasting

� Tends to lead to overoptimistic predictions

� Some analysts routinely discount the size of effects from studies

in which the evaluator was associated with the design or the

implementation of the program

• Example: do you believe Sound Transit’s forecasts on ridership levels for ST3?

Predicting Impacts

3. Technically sound statistical inference about what has

happened may not correspond to the best prediction of

what will happen

• When sample size is small or the statistical model fits the data poorly, a better prediction may result from adjusting the values

estimated from the particular data set

• Considering impacts as zero if they are not statistically significant is not the correct approach for CBA

� The estimated effect and its standard error should be used in

Monte Carlo simulations to refine the impact prediction 18

Predicting Impacts

• Common approach: predictions based on meta-analyses of similar policies

• Meta-analysis: seeks to use information from all available studies to find an effect size and its variation

• Drawing information from multiple evaluations reduces the chances that the overall result will suffer from the limitations

of any single evaluation 20

Predicting Impacts

Meta-analyses have the three following elements:

1. Identification of relevant evaluations

2. Application of a standardized measure of size effect to

facilitate comparisons across studies

3. Computation of an overall effect and its standard deviation

�� This estimated effect can then be used as the basis for

predicting the impact of a similar program 22

• Example of meta- analysis process:

• Example of meta-analysis results:

• Another example of meta-analysis results:

Predicting Impacts

• When we cannot find relevant evaluations of the policy itself, it may be possible to predict impacts using generic elasticities

• Can search in the general economics literature and/or the specialized literatures to find empirical estimates of elasticities

• The policy may effectively change the price of a good for some target population

• A price elasticity of demand for the good could then be used to predict a change in the quantity of the good consumed

Predicting Impacts

Example:

• ST3 decreases the cost of transit by reducing wait and access times

� Can use the cross-price elasticity of demand between transit

and auto to estimate the impact of ST3 on transit and auto

ridership

Predicting Impacts

• Unfortunately, in some cases the only available comparison may be statistics on the general population or findings from research done for other purposes

• (Even more) unfortunately, in some cases we cannot find any quantitative evidence to help predict an impact

� You will have to turn to logic and theory to identify a plausible range of values

• You should not use a value of “0” because you do not know the value!

• Use the most reasonable value to you � “guesstimate”

Monetizing Impacts

• CBA requires the analyst to evaluate impacts in terms of a common unit of measurement

� Whenever possible, we want to convert physical units into

monetary units

� Allows us to add up and compare all of the relevant costs and

benefits

• Depending on the impact that you evaluate, monetization can be straight-forward/certain or indirect/uncertain 29

Monetizing Impacts

• Policies that change consumption quantities are relatively straight-forward to monetize:

• When the market is undistorted, the change in quantity of the input is monetized with the market price

• When the market is distorted, you will have to use a shadow price to get the appropriate value

Monetizing Impacts: Missing Markets

• You are probably not the first one needing to monetize a particular impact, so take advantage of researchers’ previous efforts to estimate shadow prices

• Many important shadow prices have already been estimated many times

• We will discuss many of these shadow prices later in the course

• When, you cannot find existing estimates, you will have to do it yourself

• We will discuss some of these techniques later in the course

Monetizing Impacts: Missing Markets

• As with predicting impacts, it is important to take into account the quality of the research design that is the basis for the

estimates of shadow prices

• Similarly, you should take into account how closely the shadow price fits the impact you are valuing

• Also, you should take advantage of existing meta-analyses when they are available

Valuing Market Goods

• Fairly straight-forward to quantify benefits in competitive markets

• Normally, in CBA, use CS or WTP to measure benefits

�market prices a good indicator of social value

If Market Price Exists: ∆CS

( ) ( )

0 1

CS PabP

E X P P X

∆ =

∆ ≈ − ∆ −

• For these market goods, we can estimate changes in consumer surplus

directly using estimates of the elasticity

of demand and observed market

transactions:

Qo Q1

∆WTP = Q0abQ1

If No Market Exists (Price = 0): ∆WTP

• Problem here is that this market demand curve is unobservable:

Valuing Non-Market Goods

• What if a project provides a good or service for which there is no market?

1. No market price or non-competitive price

2. No market demand curve

� Cannot directly measure changes in WTP or CS

• Two approaches:

1. Revealed preferences: implied value from individuals’ choices

2. Stated preferences: implied value from individuals’ statements

• What about the prices of goods in markets that aren’t competitive?

• Do public transit fares represent individuals’ willingness to pay for a trip (i.e. the social value)?

Valuing Non-Market Goods

• Revealed preference approach:

�implied social values through actual choices individuals make

• Example: Are commuters willing to pay a $4 toll to save 10 minutes in the HOT lane?

� If so, this implies that they value their time at least as much

as $4 * 6 = $24 per hour

Many (though not all) economists believe this approach is more “reliable”

than stated preferences

�� Challenge: are you measuring the correct effect?

Valuing Non-Market Goods

• Another revealed preference example:

My mom recently came to WA from BC to visit for a few days. Her travel options:

1. Drive her car: ≈ 7 hours each direction, cost ≈ $100

2. Fly: ≈ 1 hour flight each direction, cost ≈ $800

- Since she decided to fly, I estimate her value of time as:

� Is this a reasonable estimate?

( ) ( )

800 100 $58.33 /

14 2 hr

− ≥ =

−

Valuing Non-Market Goods

• Two revealed preference techniques we will mention:

• Hedonic pricing

• Travel cost

• Note that there are other revealed preference approaches, including:

• Production or cost functions (intermediate goods)

• Averting behavior

• Cost of illness 40

Hedonic Pricing

• View demand (i.e. willingness to pay) as implicit demand for a bundle of characteristics that are bound together within a good or service

• Collect information from a sample of households with different magnitudes of these characteristics

• Housing is viewed as a bundle of attributes

House price = f(structural characteristics, locational characteristics,

individual characteristics)

Structural: Locational:

• Square footage Schools

• Number of bathrooms Amenities

• Age, etc. Environmental quality, etc.

Individual: income, age, family size, etc.

Challenge: many of these attributes are positively correlated with one another, and we

want to identify the effect of each attribute separately

Hedonic Pricing

• Interested in “implicit prices” (i.e. “shadow prices”) of these attributes, ai:

• Objective: isolate the price responsiveness of consumers to each particular attribute � revealed preferences

• Often used as a form of valuing non-market goods

• Air pollution

• Noise damages 43

Houses experiencing higher levels of such negative

externalities should sell for a lower price, ceteris paribus

House value

i a

∆

Hedonic Pricing

• Since we observe house prices directly, we can do statistical analysis to estimate the effects of attributes on house prices

• This is called hedonic regression analysis

$220,000 $216,000

Implied value of the trees

= $220,000 - $216,000

= $4,000

Hedonic Pricing

• Useful result:

• Many public policies can be evaluated by assessing the change in property values attributable to the policy

• Example: the benefit of large-scale public transit investment could be measured through changes in property values in regions that are accessible to this transit system

� Can use the change in housing values before and after ST3 for houses of varying distance to transit stations

� Travel time savings and the benefits of enhanced accessibility due to transit (if any) will be capitalized in the value of the home

Hedonic Pricing

“In general, people

living near the planned

bus and rail stations

and neighborhoods

dominated by

apartments, gave ST3

its strongest support…”

Hedonic Pricing

� Example: if ST3 provides travel time savings of 5 minutes per

trip for a given home location, this annual benefit is…

Annual value of travel time savings

= hr/trip * trips/day * days/week * weeks/year * value/hr saved

= (5/60) * 2 * 5 * 50 * $20 = $833

� With a discount rate of 5% per year and an expectation that

these benefits will last for 20 years, the house value should

increase by… 20

$833 house value $11, 219

(1.05) t

∆ = =∑

�� Calculate benefit of ST3 by adding up all of these benefits for affected houses

Hedonic Pricing

• Current Tacoma property values:

How will property

values change

due to ST3?

Hedonic Pricing

• Example of hedonic analysis using realtor.com data

http://www.realtor.com/soldhomeprices/Seattle_WA

• Variables included in our model:

• Number of bedrooms

• Number of bathrooms

• Square feet

• Year built

• Average school rating

• Minutes to drive to Downtown Seattle

• Type of building (1 = single-family, 0 = multiple-family) 50

Hedonic Pricing

# obs Mean Std Dev Min Max

Sale Price 150 $490,046 $386,385 $80,000 $2,800,000

Bedrooms 150 3.1 1.0 1 6

Bathrooms 150 2.1 0.8 1 5

Square Feet 150 1866.4 921.8 649 6278

Year Built 150 1973.3 33.4 1900 2017

School

Ratings 150 6.0 2.2 1.7 10

Min from

Downtown

Seattle

150 47.1 21.3 2 85

% Single-

family 150 0.74 0.44 0 1

$ per square

foot 150 $283 $229 $78 $2,222

Sample of house sales from January/February, 2017

Hedonic Pricing

Seattle Renton Tacoma

# Sales 50 50 50

Sale Price $759,807 $446,030 $264,300

Bedrooms 3.04 3.34 3.06

Bathrooms 2.11 2.29 1.77

Square Feet 1922 1984 1693

Year Built 1966.9 1987.0 1966.0

School Ratings 6.8 6.7 4.4

Min from

Downtown Seattle 22.9 47.0 71.3

% Single-family 0.66 0.72 0.84

$ per square foot $456 $235 $159

Sample of transactions from January/February, 2017

Hedonic Pricing

• Our (simple) hedonic model:

• Estimated coefficients represent the “marginal value of the attribute”

� To estimate the effects of ST3:

measure the distance to a light rail station for each home

before and after ST3 and include this as an additional variable

1 2 3

4 5 6

Sale price bedrooms bathrooms square feet

year built school rating distance to downtown

single-family

α α α

α ε

= ⋅ + ⋅ + ⋅

+ ⋅ + ⋅ + ⋅

+ ⋅ +

Hedonic Pricing

sale price WTP to be one mile closer to downtown

distance to downtown α

∆ ≈ ≈ ∆

Source | SS df MS Number of obs = 150

-------------+------------------------------ F( 7, 142) = 30.82

Model | 1.3415e+13 7 1.9164e+12 Prob > F = 0.0000

Residual | 8.8296e+12 142 6.2180e+10 R-squared = 0.6031

-------------+------------------------------ Adj R-squared = 0.5835

Total | 2.2245e+13 149 1.4929e+11 Root MSE = 2.5e+05

------------------------------------------------------------------------------

sale | Coef. Std. Err. t P>|t| [95% Conf. Interval]

-------------+----------------------------------------------------------------

bed | -54304.4 32577.79 -1.67 0.098 -118704.5 10095.73

bath | 83043.51 45134.16 1.84 0.068 -6178.197 172265.2

sqft | 189.0796 38.86164 4.87 0.000 112.2575 265.9017

yr | -706.9033 805.2062 -0.88 0.381 -2298.644 884.8371

school | 17930.79 10760.18 1.67 0.098 -3340.063 39201.64

commute | -8809.225 1138.791 -7.74 0.000 -11060.4 -6558.051

single | 60415.45 69243.74 0.87 0.384 -76466.34 197297.2

_cons | 1794939 1555129 1.15 0.250 -1279257 4869135

$ per mile to light-rail station???

• Testing our model:

( ) ( )

1 2 3

4 5 6 7

ˆ ˆ ˆPredicted sale price constant bedrooms bathrooms square feet

ˆ ˆ ˆ ˆ year built school rating distance to downtown single-family

= 1794939 + 54304.4

α α α

α α α α

= + ⋅ + ⋅ + ⋅

+ ⋅ + ⋅ + ⋅ + ⋅

− ( ) ( )

( ) ( ) ( ) ( )

3 83043.5 1 189.1 1570

706.9 1953 17930.8 9 8809.2 38 60415.5 1

$518, 424

⋅ + ⋅ + ⋅

+ − ⋅ + ⋅ + − ⋅ + ⋅

actual sale price

predicted sale price

Hedonic Pricing

• Statistically estimate the transit accessibility “value gradient”:

57Miles to transit

station

Increase in

home value

Hedonic Pricing

• Estimate aggregate value by linking value gradient to distribution of houses affected by ST3:

Hedonic Pricing

• Can you think of some other ways to use hedonic analysis with real estate values?

1. What other benefits could you quantify?

2. What costs (or disbenefits) could you quantify?

Hedonic Pricing

• How could we use this approach to estimate the increase in property values attributable to the transit station?

Hedonic Pricing

Contingent Valuation Methods

• In some cases, we may not be able to rely on observations of the choices of individuals

• Alternative approach:

�Interview individuals to elicit their preferences for different states of the world

• Valuation based on individuals’ stated preferences to different hypothetical conditions or “contingencies”

• Measures willingness to pay (WTP) or willingness to accept (WTA) for these hypothetical conditions 66

Stated preference approach:

• Surveys

• Questionnaires “Contingent valuation”

What are the pros and cons of this approach? Do you “believe” the results?

Contingent Valuation Methods

• Compensating Variation (CV)

• How much would individuals need to pay or be paid to leave them just as well of as before the project was implemented?

• Equivalent Variation (EV)

• How much would individuals need to pay or be paid to be as well of as if the project were implemented?

Contingent Valuation Methods

68While these are theoretically appropriate measures of net benefits, we

typically use consumer surplus (which is usually “close” to CV and EV)

Contingent Valuation Methods

Equivalent

Variation

Compensating

Variation

Negative

impact of

project

WTP to avoid project

implementation

WTA to have project

implemented

Positive

impact of

project

WTA to avoid project

implementation

WTP to have project

implemented

The appropriate measure depends on the nature of the CBA being conducted

WTP and WTA

• WTP and WTA are (usually) not identical

• Maintaining initial utility level vs. moving to new utility level

• Different income effects

• WTP is constrained by income level

• WTA is not constrained

• Example: what is your WTP to have a life-saving operation vs. your WTA to not receive the operation?

WTP and WTA

• This difference can lead to bias toward status quo

• Possible construction of new airport

• Residents’ WTA for accepting the noise pollution may be too high for the project to be undertaken

� So if currently no airport, none will be built

• If airport already exists

• Residents’ WTP to close airport and eliminate noise pollution not sufficient to compensate the airlines for closure

� So if airport already exists, will not be closed

Bias toward Status Quo

• Diminishing marginal utility of income: the increase in utility from gaining an additional $100 is less than the loss in utility

from giving up $100 from current income

Bias toward Status Quo

• If outcomes from project are uncertain, consumers prefer certainty of status quo to project with same level of expected income:

73 Income, Expected Income ($)

Utility of status quo

Expected utility of project

Outcome

Risk averse: from starting point, losses weigh more heavily than gains

Bias toward Status Quo

Scenario 1:

• You just won $30

• You can now take the following bet: 50% chance of winning $9, 50% chance of losing $9

� Would you take this bet?

Scenario 2:

• Option 1) You can have $30 guaranteed

• Option 2) You can take the following bet: 50% chance of winning $21, 50% chance of winning $39

� Which option do you prefer?

Steps of CVM

Step 1: Construction of a hypothetical market

• Construct a scenario which corresponds as closely as possible to a real-world situation

• Reason for payment

• Explain payment mechanism (tax increase, increased price of good, etc)

• Construct provision rule for the good/service 75

Steps of CVM

Step 2: Obtaining the data

• Determine sample and implement survey via:

1. Personal interview, person-to-person

2. Personal interview session using an interactive medium (computer)

3. Mail questionnaire (with follow-ups)

4. Telephone interview 76

Steps of CVM

Step 2: Obtaining the data (cont.)

• Try to determine WTP or WTA using:

a) “Bidding game”: Ask a sequence of questions until maximum is found.

May suffer from lack of incentive compatibility, starting point bias, and

fatigue effects.

b) Payment card: Card indicates range of possible values, one of which is

pointed out by interviewee. May have problems of starting point bias.

c) Open-ended question: No anchor. High degree of individual

impreciseness, and sometimes systematic bias, may be a problem.

d) Closed-ended single-bounded referendum: “Yes” or “no” for specific

value provided.

e) Double-bounded referendum: Same as d, but with an additional follow-

up question of maximum WTP.

Steps of CVM

Step 3: Estimating average WTP/WTA

• Straightforward with open-ended and bidding-game formats

• More difficult with single-bounded referendum

� Must estimate probability functions, requiring more data

Steps of CVM

Step 4: Estimating bid curves

• Define bid curve for individual i as:

WTP(i) = f(Y(i), E(i), A(i), X(i), Q, U(i), e(i))

• Y = income

• E = education

• A = age

• Q = environmental quality

• X = vector of other background variables we want to control for

• U = individual use of the environmental asset/object

• e = random error term

Steps of CVM

Step 5: Aggregating the data

• Convert mean bids from the sample to population aggregates

Strengths of CVM

• Able to directly evaluate the issue in question

• Questions can be tailored to represent unique attributes

• Can be reasonably low-cost to conduct

Criticisms of CVM

• Strategic bias • Proponents of project may overstate WTP

• Starting point bias: “anchoring” • Vehicle bias • Mental account or scope bias

• Individuals make assessments of hypothetical situations • Do not face real budget constraints • May state very high WTP for positive impacts

• Do not fully assess the opportunity cost of the foregone income

• Embedding • Major differences in WTP vs. WTA • Informational biases & dependence on how questions are formed • If considering existence value, then who should be sampled? • Non-response bias

• Are respondents a “representative sample”? • How to treat “outliers”?

CVM Example

“If an additional $__ was added to your yearly local income tax, the riparian buffers around Brandywine Creek and all of its tributaries could be installed, repaired and maintained. This would result in a temperature decrease of 5° F, in turn creating a 25% increase in dissolved oxygen levels. The riparian buffers will also decrease the nitrate concentration by 50%.”

< The blank in the above text was replaced with a randomly selected amount between $2 and $200. >

Please answer the following questions honestly.

1. If the increase in taxes will be ___ $ per year, would you support this program? YES NO

2. What is the maximum your household would be willing to pay per year to preserve our watershed? $________

CVM Example

n Mean Std Dev Lower CL (90 %) Upper CL (90%)

110 $40.37 $42.95 $33.58 $47.17

Estimated Population WTP $4.5 million $6.3 million

Conservative Estimate $1.3 million $1.8 million

Bivariate Logit Coeff Std Err Low CL Up CL Population WTP

in Millions

Conservative

Population WTP in

millions

Tax -.0167 0.007

Constant 1.9853 0.339

Median $118 $63 $375 Low CL Up CL Low CL Up CL

Mean $126 $69 $395 $9.2 $52.7 $2.6 $14.9

Truncated Mean $87 $34 $129 $4.6 $17.2 $1.3 $4.8

Table 4. Maximum Willingness to Pay Open-Ended Response

Table 5. Willingness to Pay Discrete Choice

Assignment #3

• Any remaining questions?