1
$\begingroup$

"It is well known" that the elasticity of $Q$ relative to $P$ is the constant $b$ in the equation $Q(P)=aP^b$. We can verify it using the convenient definition of elasticity $= \frac{dQ}{dP} \frac{P}{Q}$, we can verify it in Wikipedia.

Based on this, when I create a spreadsheet with parameters $b=-2$, $a=100$ and P ranging from 1-5 by 1 I expected to recover an elasticity of -2, where elasticity = % change in Q/ % change in P. This is not the case. Instead the elasticity varies substantially as the % change in P varies.

I'm forced to conclude that 'constant elasticity' does not mean a fixed value regardless of the magnitude of the price change, but only invariant with regards to the value of price OR that I have a spreadsheet error.

I've pasted an image of the spreadsheet results below. Obviously, without seeing the formulas, no one can verify that I don't have an error, but I'm hopeful that someone will give this a spin for themselves and get very different numbers and then I'll know my failure to verify the constant elasticity in this function is a spreadsheet error.

spreadsheet image

$\endgroup$
2
$\begingroup$

The problem with your calculations is that you have very big changes in price

If you made price increases of $1\%$ in price, you would find the quantity fell by about $2\%$ as shown in the table below leading to an elasticity of about $-2$

If you made the price changes even smaller, the calculated elasticity would get even closer to $2$, and in a sense you are looking for the limit of elasticity as the price change approaches $0$, which will indeed be the constant $b=-2$

 a       b      P       Q=a*P^b pct change Q    pct change P   elasticity

100     -2      1       100.00          
100     -2      1.01     98.03      -1.97%          1%          -1.97
100     -2      2        25.00          
100     -2      2.02     24.51      -1.97%          1%          -1.97
100     -2      3        11.11          
100     -2      3.03     10.89      -1.97%          1%          -1.97
100     -2      4         6.25          
100     -2      4.04      6.13      -1.97%          1%          -1.97
100     -2      5         4.00          
100     -2      5.05      3.92      -1.97%          1%          -1.97
$\endgroup$
1
$\begingroup$

Elasticity is the slope after taking a log transform. Throughout this post $\log$ refers to the natural logarithm. That is, the logarithm with base $e$, the ln function in Excel. The logarithm transforms multiplication into addition. Log differences are something akin to percent changes (and they're equivalent for small changes).

Let's say we have:

$$ Q_t = A P_t^b $$

Take the logarithm of both sides:

$$ \log Q_t = \log A + b \log P_t$$

Log differences are conceptually similar to percent changes (for small percent changes). The slope of the line using the log transformed variables is $b$. This is the elasticity.

$$ \frac{d \left( \log Q_t \right) }{\log P_t} = b $$

Why do people treat elasticities as percent changes?

$$\log(Y) - \log(X) \approx \frac{Y-X}{X} \quad \quad \text{ for }\frac{y}{x} \approx 1$$

Log differences are approximately the percent change. Why? Take the first-order taylor expansion of the log to approximate it with a line near 1.

\begin{align*} \log(Y) &\approx \log(1) + \frac{1}{\log(1)}\left(Y - 1 \right) \\ &\approx Y - 1 \end{align*} Hence $\log\left(\frac{Y}{X} \right) \approx \frac{Y-X}{X}$ for $Y \approx X$. This works well for small changes, eg. $\log(1.03) = .0296$, but it gets quite off for big changes $\log(1.5) = .4055$. This is probably where your stuff is getting off...

Example where logs are nice...

Let's say we have $1, 2, 4, 8$. Each is a 100% increase from the previous value. Taking logs we have $0, .693, .1386, 2.079$. The numbers increase linearly by $\log(2)$. Looking at percent increases though, it's less clean: $\frac{2-1}{1} = 100\%$ and $\frac{4-2}{2} = 100\%$ but $\frac{4-1}{1} = 300%$.

$$ \left(1 + R_1 \right) \left(1 + R_2 \right) = 1 + R_1 + R_2 + R_1R_2$$ Dealing with percent changes gets messy when $R_1R_2$ is big enough to matter. Taking logs though, let $r_1 = \log(1+R_1)$ etc...

$$ \log\left((1+R_1)(1+R_2) \right)= r_1 + r_2$$

What I would do if I were you:

Take a log and then it will be clean, and simple. Add a few columns to your spreadsheet.

  • $q_t = \log Q_t$.
  • $a = \log A$
  • $p_t = \log P_t$

Then the equation is: $$ q_t = a + b p_t$$

$$\frac{dq_t}{dp_t} = b $$

$$ \begin{array}{rrrrr} P & Q & \log P & \log Q & \Delta \log P & \Delta \log Q & \frac{\Delta \log P}{\Delta \log Q} \\ 1 & 100 & 0 & 4.605 & \\ 2 & 25 & .6931 & 3.2189 & .6931 & -1.3863 & -2.0\\ 3 & 11.11 & 1.0986 & 2.4079 & .4055 & -.8109 & -2.0\\ 4 & 6.25 & 1.3863 & 1.8326 & .2877 & -.5754 & -2.0\\ 5 & 4 & 1.6094 & 1.3863 & .2231 & -.4463 & -2.0 \end{array} $$

$\Delta$ means "change in", hence $\Delta \log Q_i = \log Q_i - \log Q_{i-1}$.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.