Mathematics for Economists 3610-E/3610-E-V
Black-board exercises
Chapters 4-5
1. Suppose that the relationship between gross income Y and total income
tax T , for taxpayers with incomes between 80K and 120K, is given by
T = a(bY + c)p + kY,
where a, b, c, p, k are positive constants. Find an expression for the marginal
tax rate, dT /dY . Evaluate it when Y = 100000 and the constants are
a = 0.0000338, b = 0.81, c = 6467, p = 1.661, k = 0.053.
2. Let C(Q) be the total cost of producing Q units of a commodity. The
quantity C(Q)/Q is called the average cost of producing Q units. Find an
d
expression for the rate of change of the average cost, i.e., dQ
(C(Q)/Q).
3. Consider the following macroeconomic model for determining national income in closed economy: Y = C + I and C = f (Y ), where Y is the
national income (goes either to consumption C or to investment I) and
C is the consumption function. The two equations define Y as a differentiable function of I. Assume that f 0 (Y ) (0, 1). Find expressions for
dY /dI and d2 Y /dI 2 (i.e., Y 0 and Y 00 ).
4. Let D(p) denote the demand function for a product. By selling D(p) units
at price p, the producer earns revenue R(p) given by R(p) = pD(p). Find
an expression for the elasticity of R(p) w.r.t. p, ER , as a function of the
elasticity of D(p) w.r.t. p, ED .
5. Poisson random variable X with parameter > 0 is characterized by the
so-called probability mass function
f (x; ) = P rob(X = x) =
x e
x!
for x = 0, 1, 2, ..., where x! = 1 2 . . . x (0! = 1). For example, X =
the number of phone calls received at an info desk per hour, and = the
Most
of the problems come from the textbook.
average number of phone calls received per hour.
In econometrics, the likelihood function, based on a sample of n independent observations from this distribution (or any other distribution with a
probability mass (or density) function f ), x1 , x2 , . . . , xn , is given as
l() = f (x1 ; ) f (x2 ; ) . . . f (xn ; ) =
n
Y
f (xi ; )
i=1
that is, it is the product of the n individual probability mass functions
(the so-called joint probability), where the parameter is considered as
an argument of the function l. The log-likelihood function is simply
L() = ln(l()).
(a) Show that L() can be written as
n + ln()
n
X
xi
i=1
n
X
ln(xi !)
i=1
(b) Based on the previous part, find the derivative of L w.r.t. .
(c) Assume you observe the following five values from a variable that
follows a Poisson distribution with parameter (which is unknown)
x1 = 8, x2 = 9, x3 = 11, x4 = 13, x5 = 10. For these five observations, sketch a graph (by hand or in Mathematica/calculator)
of the log-likelihood function for [0.01, 20]. For what is the
log-likelihood function maximized (roughly)?
(this value is obtained
Note: Log-likelihood is maximized for = x
by solving the equation L0 () = 0). The idea behind the maximumlikelihood estimation is to choose in such a way that it maximizes
the likelihood of observing the data we actually observed.
6. Differentiate the following functions of t (a, b, c, d are constants, n is a
natural number):
(a)
f (t) =
(b)
at + b
ct + d
g(t) = tn (a t + b)
(c)
h(t) =
at2
1
+ bt + c
7. Suppose that y is a differentiable function of x given by
x + y 3 = y 5 x2 + 2y
for all x in a given interval I. Find an expression for y 0 =
2
dy
dx .
8. It has been shown that the demand Q for butter in Stockholm during
1925-1937 was related to price P by the equation
Q P 1/2 = 38
Find Q0 = dQ/dP by implicit differentiation.
9. Bernoulli random variable X with parameter p (0, 1) is defined as
1 with prob. p
X=
0 with prob. 1 p
hence its probability mass function can be written as
x
p
,
f (x; p) = px (1 p)1x = (1 p)
1p
for x = 0, 1. Imagine you observe a sample of size n from this distribution.
(a) Show that the log-likelihood function, L(p), (see Exercise 5 of Section
1) can be written as
L(p) = n ln(1 p) +
n
X
xi (ln(p) ln(1 p))
i=1
(b) Based on the previous part, find the derivative of L w.r.t. p.
(c) Solve equation L0 (p) = 0 for p.
(d) Assuming you observe the following n = 4 values from a variable that
follows a Bernoulli distribution with parameter p (unknown) x1 = 0,
x2 = 1, x3 = 0, x4 = 0, sketch a graph of L(p) for p [0.01, 0.99].
For which p is the log-likelihood function maximized (roughly)?
Answers:
1.
T 0 (Y ) = abp(bY + c)p1 + k
and for a = 0.0000338, b = 0.81, c = 6467, p = 1.661, k = 0.053,
T 0 (100000) = 0.137. Extra: In Figure 1, functions T (Y ) and T 0 (Y ) are
shown.
2.
d
dQ
C(Q)
Q
=
C 0 (Q)Q C(Q) 1
Q2
marg. cost
z }| {
1
C 0 (Q) C(Q)
Q
Q
| {z }
ave. cost
Note: the rate of change of the average cost is positive if the marginal cost
exceeds the average cost.
3
Figure 1: Left: T (Y ) from Exercise 1 of Section 1 for a = 0.0000338, b = 0.81,
c = 6467, p = 1.661, k = 0.053. Right: T 0 (Y ) with the same parameters.
3. To find Y 0 , we first differentiate both sides w.r.t. I:
Y
= f (Y ) + I
= f 0 (Y )Y 0 + 1
Y f (Y )Y
Y0
1
= (1 f 0 (Y ))1
1 f 0 (Y )
Now Y 00 :
Y 00
1(1 f 0 (Y ))2 (f 00 (Y )Y 0 )
Y 00
(1 f 0 (Y ))2 f 00 (Y )Y 0
=(1f 0 (Y ))1
Y 00
Y 00
z}|{
Y0
Y 00
= (1 f 0 (Y ))3 f 00 (Y ) =
(1 f 0 (Y ))3
=
Answers:
Y0 =
(1 f 0 (Y ))2 f 00 (Y )
1
1 f 0 (Y )
and Y 00 =
f 00 (Y )
(1 f 0 (Y ))3
Note that Y 0 is always greater than 1 (1 unit increase in the investment
will lead to more than 1 unit increase in the national product).
4. According to the definition
ER
p d
R(p)
R(p) dp
p
d
R(p)
pD(p) dp
=
=
Lets find
d
dp R(p)
d
R(p) = D(p) + pD0 (p)
dp
4
Hence
ER
p
(D(p) + pD0 (p))
pD(p)
z }| {
p
1 + D0 (p)
D(p)
ED
Answers:
ER = 1 + ED ,
ED = D0 (p)
where
p
D(p)
5. (a) We have
L()
n
Y
ln(l()) = ln
!
f (xi ; )
= ln
i=1
=
=
=
ln e
n
Y
xi
n
Y
xi e
xi !
i=1
!
n
n
Y
xi
= ln e
+ ln
xi !
x!
i=1 i
xi
n
X
n +
ln
= n +
[xi ln() ln(xi !)]
xi !
i=1
i=1
i=1
n
X
n + ln()
n
X
xi
i=1
n
X
ln(xi !)
i=1
(b) From the previous part we see that
n
L0 () = n +
1X
1X
xi 0 = n +
xi
i=1
i=1
(c) See Figure 2. For = 10.2 (the sample mean).
6. (a)
0
f (t)
=
Answer: f 0 (t) =
at + b
ct + d
0
=
a(ct + d) (at + b)c
ad bc
=
2
(ct + d)
(ct + d)2
adbc
(ct+d)2
(b)
g 0 (t)
0
tn (a t + b) = (atn+1/2 + btn )0
a(n + 1/2)tn1/2 + bntn1
Answer: g 0 (t) = a(n + 1/2)tn1/2 + bntn1
In[2]:=
-lambda * 5 + Log @lambdaD * Hx1 + x2 + x3 + x4 + x5 L -
L@lambda_ , x1_, x2_, x3_, x4_, x5_D :=
HLog @Factorial @x1DD + Log @Factorial @x2DD + Log @Factorial @x3 DD +
Log @Factorial @x4DD + Log @Factorial @x5 DDL
In[3]:=
Plot@L@lambda, 8, 9, 11, 13, 10D, 8lambda, 0.01, 20<D
-10
-20
-30
Out[3]=
-40
-50
-60
5
10
15
20
Figure 2: Log-likelihood function from Exercise 5 of Section 1.
(c)
h0 (t)
=
1
at2 + bt + c
0
= ((at2 + bt + c)1 )0
= 1(at2 + bt + c)2 (2at + b) =
2at + b
+ bt + c)2
(at2
Answer: h0 (t) = (at22at+b
+bt+c)2
7. Step 1) (differentiate w.r.t. x)
1 + 3y 2 y 0 = 5y 4 y 0 2x + 2y 0
Step 2) (solve for y 0 )
1 + 2x = (5y 4 + 2 3y 2 )y 0
And hence
y0 =
Answer: y 0 =
1 + 2x
5y 4 + 2 3y 2
1+2x
5y 4 +23y 2 .
8. After differentiating both sides w.r.t. P , we obtain
1
Q0 P 1/2 + Q P 1/2
2
Q0
1
= QP 1 multiply the RHS by P 1/2 P 1/2 = 1
2
6
=38
Q0
Q0
z }| {
1
= QP 1/2 P 1/2 P 1
2
1
19
= 38P 3/2 = 3/2
2
P
Answer: Q0 = 19/P 3/2
9. Bernoulli distribution
(a) We have
L(p) = ln(l(p))
ln
ln
ln
ln
x1
x2
xn
p
p
p
(1 p)
(1 p)
. . . (1 p)
1p
1p
1p
!
xi
n
Y
p
(1 p)
1
p
i=1
xi !
n
Y
p
n
(1 p)
1p
i=1
Pn xi !
i=1
p
property : ax ay = ax+y
(1 p)n
1p
= n ln(1 p) +
n
X
xi (ln(p) ln(1 p))
i=1
(b) Based on the previous part
n
L0 (p)
X
n
+
xi
1 p i=1
X
1
n
+
xi
1 p p(1 p) i=1
1
1
+
p 1p
(c) We have
n
1
+
1 p p(1 p)
L0 (p) = 0
n
X
xi = 0
i=1
n
1X
n +
xi
p i=1
1X
xi
p i=1
Pn
i=1
xi
=x
(d) See Figure 3. For p = 0.25, which is the sample mean.
7
Figure 3: Log-likelihood function from Exercise 9 of Section 1.
Chapters 6-7
1. Consider the graphs (i)-(vi) of function f (x) from Figure 4. Based on the
figure, fill out the missing parts of Table 1.
limxa f (x)
limxa+ f (x)
Value of f at a
Is limxa f (x) = f (a), i.e.,
is f left-cont. at a?
Is limxa+ f (x) = f (a), i.e.,
is f right-cont. at a?
Is f cont. at a?
limx f (x)
(i)
(ii)
Graph
(iii)
(iv)
Skip
Skip
Skip
(v)
(vi)
Skip
Table 1: Questions/Answers to Exercise 1 of Section 2 based on Figure 4. NA
= Not Applicable, Y = Yes, N = No.
2. Calculate the following limits:
(a) limx2
3x2 +3x18
x2
Figure 4: Graphs of f (x) for Exercise 1 of Section 2.
(b) limh0
(c) limx
(d) limx2
h+11
h
3x2 +x1
x2 +1
x4 4x3 +5x2 4x+4
x3 2x2 4x+8
(e) limx
x5 x4 x
3. Find the mistake in
x2 + 3x 4
2x + 3
2
1
= lim
= lim =
x1 2x2 2x
x1 4x 2
x1 4
2
lim
4. Assume that a, b are constants. Find
ax + b cx + b
lim
x0
x
5. Show that
lim
p
n
a
1
xn + a1 xn1 + . . . + an1 x + an x =
n
6. Find limx f (x), where f (x) =
xp
ax ,
a > 1, p fixed.
7. Suppose f (x) is a continuous function from [a, b] to [a, b]. Based on The
Intermediate-Value Theorem, justify that the equation f (x) x = 0 has
at least one solution in [a, b]. (Hint: define g(x) = f (x) x and apply the
theorem to g.)
8. If a function f is continuous in the interval I and differentiable in the
interior of I, then we have the following implication:
if f 0 (x) < 0, then f is strictly decreasing on I
Provide the proof of this statement. (Hint: check the proof of a similar
statement for the strictly increasing f that you can find in the theory
slides from Ch.7.)
10
