Scribd Logo
Upload

Welcome to Scribd!

  • 30 views

    Sol ps2 PDF

    Uploaded by

    Jacob
    This problem set reviews multivariate regression analysis and requires students to answer questions about an article on gasoline demand. 1) The authors estimate a demand equation using log forms of gasoline consumption, price, and income along with month and year fixed effects. Table 1 reports estimates of the price elasticity, income elasticity, and constant term for different time periods. 2) Alternate linear specifications are considered and used to calculate implied elasticities for different months. A quantity weighted average is argued to be more reasonable for measuring overall elasticity. 3) A model with an interaction between price and income is specified, where the price elasticity depends on the log of income and vice versa. Estimates of this interaction term

    Copyright:

    © All Rights Reserved
    Download as PDF, TXT or read online from Scribd

    Sol ps2 PDF

    Uploaded by

    Jacob
    30 views5 pages
    This problem set reviews multivariate regression analysis and requires students to answer questions about an article on gasoline demand. 1) The authors estimate a demand equation using log forms of gasoline consumption, price, and income along with month and year fixed effects. Table 1 reports estimates of the price elasticity, income elasticity, and constant term for different time periods. 2) Alternate linear specifications are considered and used to calculate implied elasticities for different months. A quantity weighted average is argued to be more reasonable for measuring overall elasticity. 3) A model with an interaction between price and income is specified, where the price elasticity depends on the log of income and vice versa. Estimates of this interaction term

    Original Title

    sol_ps2.pdf

    Copyright

    © © All Rights Reserved

    Available Formats

    PDF, TXT or read online from Scribd

    Did you find this document useful?

    Is this content inappropriate?

    This problem set reviews multivariate regression analysis and requires students to answer questions about an article on gasoline demand. 1) The authors estimate a demand equation using log forms of gasoline consumption, price, and income along with month and year fixed effects. Table 1 reports estimates of the price elasticity, income elasticity, and constant term for different time periods. 2) Alternate linear specifications are considered and used to calculate implied elasticities for different months. A quantity weighted average is argued to be more reasonable for measuring overall elasticity. 3) A model with an interaction between price and income is specified, where the price elasticity depends on the log of income and vice versa. Estimates of this interaction term

    Copyright:

    © All Rights Reserved
    Download as PDF, TXT or read online from Scribd
    Download as pdf or txt
    30 views5 pages

    Sol ps2 PDF

    Uploaded by

    Jacob
    This problem set reviews multivariate regression analysis and requires students to answer questions about an article on gasoline demand. 1) The authors estimate a demand equation using log forms of gasoline consumption, price, and income along with month and year fixed effects. Table 1 reports estimates of the price elasticity, income elasticity, and constant term for different time periods. 2) Alternate linear specifications are considered and used to calculate implied elasticities for different months. A quantity weighted average is argued to be more reasonable for measuring overall elasticity. 3) A model with an interaction between price and income is specified, where the price elasticity depends on the log of income and vice versa. Estimates of this interaction term

    Copyright:

    © All Rights Reserved
    Download as PDF, TXT or read online from Scribd
    Download as pdf or txt
    You are on page 1of 5

    14.44/14.

    444 Energy Economics, Spring 2006


    Problem Set 2

    Due Friday February 23, 2006 in class or Arthur Campbell’s mail folder

    Late problem sets are not accepted

    This problem set reviews your knowledge of multivariate regression analysis. It re­
    quires you to answer questions related to the article on the reading list ”Evidence of
    a shift in the short-run price elasticity of gasoline demand” by J. Hughes, C. Knittel
    and D. Sperling. Some of you may also find the article ”An introduction to regression
    analysis” by A. Sykes, also on the reading list, is a useful reference for the material.

    1. (60) In section 2 the authors estimate the following demand equation:

    ln Gjt = β 0 + β 1 ln Pjt + β 2 ln Yjt + εj + εjt

    where Gjt is per capita gasoline consumption in gallons in month j and year t, Pjt
    is thee real retail price of gasoline in month j and year t, Yjt is real per capita
    disposable income in month j and year t, εj represents unobserved demand factors
    that vary at the month level and εjt is a mean zero error term.

    a) (10) What have the authors assumed about the price elasticity of demand when
    they wrote down the demand equation in this form? Remember the price
    ∂G P
    elasticity of demand Ep = ∂P G .

    Ans: The assumption made by the authors is that consumer demand for petrol
    has a constant elasticity of demand at all points along the demand curve.

    b) (10) Go to table 1 in the appendix, now assuming the authors have obtained
    unbiased estimates of the parameters β 0 ,β 1 and β 2 what do they mean? (eg.
    the coefficient β 1 is -0.335 in the period 1975-1980, this represents...)

    1
    Ans: The coefficient β 1 is the price elasticity of demand, in the period 1975­
    1980 it is -0.335, this means that for a 1% increase in price in this period
    demand will fall by 0.335%, similarly in the period 2001-2006 it is -0.042 so
    for the same price change demand falls by 0.042%.

    The coefficient β 2 is the income elasticity of demand in the period 1975­


    1980 it is 0.467, this means that for a 1% increase in income in this period
    demand will increase by 0.467%, similarly in the period 2001-2006 it is 0.530
    so for the same change in income demand increases by 0.530%.

    The coefficient β 0 is the constant in the demand specification. This


    scales the demand relation in such a way (holding demand elasticity con­
    stant) larger values of β 0 indicate larger per capita demand for the same price
    across different time periods. The value of β 0 decreases from -0.615 to -1.697
    from 1975-1980 to 2001-2006. This represents a fall in per-capita gasoline
    consumption which is consistent with the data presented in Figure 1 of the
    paper.

    c) (10) Interpret the values of the monthly unobserved demand factors (εj )? What
    are these relative to? What can you say about the yearly pattern of gasoline
    demand from these coefficients?

    The values of the monthly unobserved demand factors can be interpreted


    in much the same way as the constant term β 0 . That is greater values of the
    coefficient relative to other time periods or months indicates relatively higher
    demand. The yearly pattern of gasoline demand is characterized by a peak
    during August and a low during February and intermediate levels inbetween.
    The monthly factors are relative to the level in December which is normalized
    to 0 by including the constant term and excluding the month dummy for
    December.

    d) (10) From the information presented in this table calculate the appropriate t-
    statistics for each of the β � s to test if it is different from 0. You will need the
    standard errors for each coefficient which are presented in brackets below the
    respective coefficient value in the table. For instance the standard error for
    the coefficient β 1 in the period 1975-1980 is 0.024.

    2
    T-statistics are calculated by dividing the value of the co-efficient by the
    standard deviation. They are used to test whether a coefficient is significantly
    different from 0. For example the first entry in the table below is calculated
    −0.615
    by 0.929 = −0.662.
    1975-1980 2001-2006

    β0 -0.662 -2.891

    β1 -13.958 -4.667

    β2 4.865 9.138

    e) (10) What do the *** next to some of the entries in the table indicate? How
    are they related to the t-statistics you calculated?

    This indicates that the coefficient is significant at a level of 1%. The t-statistics
    are used along with the student-t distribution to determine at what level the
    coefficient is significant

    f ) (10) The table presents the adjusted R-squared statistic for the two regressions.
    What does this number mean? If we calculated the unadjusted R-squared
    values, can we say whether these are larger or smaller than the adjusted R-
    squared values of 0.84 and 0.94 in this table?

    The adjusted R-squared values represent the percentage of the variation


    in per-capita consumption is captured by the regression adjusted for the num­
    ber of variables in the regression. The unadjusted R-squared values will be
    greater than the adjusted values.

    2. (30) In table 2 and table 3 in the appendix, two alternate specifications for the
    demand equation are compared with the original double-log model.

    a) (10) Under the linear specification for the period 1975-1980 the coefficient on the
    Price variable is -7.252. What is the implied elasticity of demand, assuming
    the linear model, if during a June month per capita demand was 40 gallons,
    and price was $1.70?
    ∂G P 1.7
    Ep = ∂P G = −7.252 × 40 = −0.308

    b) (10) What is the implied elasticity of demand if during July demand is 5 gallons
    higher (due to a month specific effect) and price is the same?

    3
    ∂G P 1.7
    Ep = ∂P G = −7.252 × 45 = −0.274

    c) (10) Under the linear demand specification the demand elasticity varies within
    each period. Therefore the authors calculate an ”average” elasticity of demand
    across each period. Do you think a time weighted or quantity weighted average
    is more reasonable and why?

    A quantity weighted average seems more reasonable since we are dealing


    with elasticities. In terms of the amount of quantity demanded across the
    entire period and how this is affected by the overall price level then additional
    weight should be attached to periods/months during which there was more
    demand. In the extreme case where only a couple of gallons are demanded
    in a month the very small changes which occur during this month should not
    have as large an impact on the overall demand elasticity as say a month with
    very large gasoline demand.

    3. (20) In section 3.3 the authors’ specify a model where there is an interaction be­
    tween price and income

    ln Gjt = β 0 + β 1 ln Pjt + β 2 ln Yjt + β 3 ln Pjt ln Yjt + εj + εjt

    where β 3 is the coefficient on the interaction term.

    a) (10) Derive the price elasticity of demand using this model specification. You
    should get Ep = β 1 + β 3 ln Yjt . What would be the expression for the income
    elasticity of demand?

    Differentiating the demand relationship with respect to Pjt :

    1 ∂G β β ln Yjt
    = 1 + 3
    G ∂P Pjt Pjt

    which can be rearranged to get

    P ∂G
    Ep = = β 1 + β 3 ln Yjt
    G ∂P

    similarly differentiating with respect to Yjt

    1 ∂G β β ln Pjt
    = 3 + 3
    G ∂Y Yjt Yjt

    once more rearraging to get

    Y ∂G
    EY = = β 2 + β 3 ln Pjt
    G ∂Y

    b) (10) Interpret the values of β 3 presented in table 7 in the appendix. What do


    these tell you about the effect of income on the price elasticity of demand?
    How has this changed from the 1975-80 period to today?

    β 3 is -2.879 during 1975-80 and -1.014 during 2001-06. This tells us


    that the interaction between price and income has become weaker. That is
    % changes in price and income have relatively less impact on the elasticity of
    demand wrt the other variable. The negative sign indicates that increases in
    income decrease the elasticity of demand wrt price, because price elasticities
    are normally negative this in fact makes demand more elastic. On the other
    hand increases in prices decrease the elasticity of demand wrt income which
    is positive here and thus makes it more inelastic.

    4. (40) (14.444 students only) In section 3.2 the author’s find that when the simul­
    taneity bias is accounted for in the period 2001-2006 the estimate for the elasticity
    of demand changes from -0.042 to -0.077 and this change is statistically significant.
    The authors’ conclude that this is encouraging and that the effects of simultaneity
    are small relative to other factors. Give a brief (1/2 a page max.) critique of the
    authors’ methodology for addressing the simultaneity bias.

    So in this section I was just after a bit of a discussion of both the instrument
    the authors’ used to address the simulataneity bias and then what some of the assump­
    tions are that would make their approach valid in the event that they had not introduced
    a simultaneity bias through using a regression of quantity directly on price. In particular
    this assumes that variation observed in the data is not coming from demand side shocks
    rather there are supply side shocks which provide the variation in prices and quantities
    which allow us to identify the demand elasticity.

    You might also like