Total Effects

Total effect of a price change = substitution effect + income effect
This is what is captured in the Slutsky equation for a change in the demand for good \(x_i\) in response to a change in price of good \(x_j\):
\(\frac{\partial x_i (p, w)}{\partial p_j} = \frac{\partial h_i(p, u)}{\partial p_j} - \frac{\partial x_i (p, w)}{\partial w} d_j (p, w)\)
where  \(h(p, u)\)is the Hicksian demand (holding utility constant) and \(x(p, w)\) is the Marshallian demand (which holds income constant), at the vector price levels \(p\), wealth (or income) level \(w\), with a fixed utility level  \(u \) given by the original maximized utility function subject to the budget constraint. 
Alternately, the Frisch elasticity of labor supply holds the marginal utility of wealth constant. This is accomplished by capturing the elasticity of hours worked to the wage rate :
\(\frac{\% \Delta L_S}{\% \Delta w}\)
and then keeping a constant marginal utility of wealth.
* Fun aside: Keynes (in Economic Possibilities for our Grandchildren) predicted that our living standards would quadruple (yes) and that we'd work 15hrs/wk. (ha ha.. no). What can economic theory tell us about this fact? Answer: Don't forget that while we work the same amount of hours per week, we also created childhood and retirement, so less lifetime hours potentially. Otherwise, this tells us that the substitution effect must be much stronger than the income effect in the elasticity of labor supply. 

Log Utility

To find a neoclassical model where labor supply doesn't depend on productivity, what do I need to believe?
Consider:
\(u = \frac{z L ^{1 - \theta}}{1 - \theta} - \gamma L\)
where \(z\) is a measure of productivity, \(\theta\) is intertemporal time preference, and \(\gamma\) represents the disutility of labor. First-order conditions yield:
\(L^* = \gamma ^{- \frac{1}{\theta}} z ^{\frac{1}{\theta} - 1}\)
as we can see above, when \(\theta = 1\), productivity has no impact on the labor supply. In other words, the income and substitution effects will cancel out. Another way to see this:
\(u = \ln z L - \gamma L\)
\(\frac{\partial u}{\partial L} = \frac{1}{zL} \cdot z - \gamma\)
The first fraction represents the substitution effect and the second "z" is the income effect, where both of these will be eliminated.  

Cases other than Log Utility

Keynes's prediction meant that income effect would win. What does this look like? Begin again with:
\(L^* = \gamma ^{- \frac{1}{\theta}} z ^{\frac{1}{\theta} - 1}\)
\(\ln L^* = \frac{1}{\theta} \ln \gamma + (\frac{1}{\theta} - 1) \ln z\)
Therefore, if \(\theta > 1\), a rise in z means the income effect wins and there are mass vacations (Keynes was right). However, if \(\theta < 1\), the substitution effect wins and people will work more. This is also called quasi-linear utility. When considering \(\theta\), the real question to ask is, how strong is diminishing marginal utility of consumption?

CES and near-Leontief Behavior

Now we will examine "z" as a share parameter, that will sometimes mean a "lifting tide raises all boats," and sometimes be factor-specific. This will depend on the type of production function that we use. If we want to use Cobb-Douglas, our solution will be the latter:
\(u = zK^\alpha L^{1 - \alpha}\)
\(\ln u = \ln z + \alpha \ln K + (1 - \alpha) \ln L\)
Thus the productivity share parameter extends to all factors, and this is a case of CES production functions where each input is a perfect complement. However, looking at a case of constant elasticity of substitution of the Leontif type:
\(u = (zK)^\alpha + L^{1 - \alpha}\) 
Here, z represents factor-biased technological change. For example, we would see this when machines are increasingly getting more productive though labor remains at a stable level.
In general, if you have

Corner Solutions & Dixit

When should you spend all your money on one good?  The evidence of this would be a corner solution. Our answer relies on the equimarginal principle: 
If marginal utility per dollar for good X is less than marginal utility per dollar for good Y (regardless of the amount of good Y already bought), spend all your money on good Y. Or, 
\(\frac{MU_X}{\$} > \frac{MU_Y}{\$} \Longrightarrow \$ \rightarrow Y\)
Again, this is a classic corner solution. Don't buy any of a certain good if the marginal utility of that good is always lower than the rest. Example: 
How high does "z" have to be for me to spend all my income on good 2?
max \(u = c_1 + z \sqrt c_2\)
subject to \(c_1 + c_2 = 100\)
\(\frac{MU_1}{\$} = 1\)
\(\frac{MU_2}{\$} = \frac{1}{2} z c_2 ^{- \frac{1}{2}}\)
Recall that the marginal utility of a good is the partial derivative with respect to that good of the total utility function. Optimizing will always mean setting these marginal utilities equal to one another, and then the question gives us the assumption that all of income is spent on good 2:
\(1 = \frac{1}{2} z 100 ^{- \frac{1}{2}}\)
\(z = 20\)
Calculus are the wrong way to solve a corner solution. If we would have workout out the Lagrangian, the solution would have been:
\(c_1 = -300 , c_2 = 400\)
This tells us that we should be buying good 1 and turning it into good 2. In other words, it's a clue that our marginal utility will never be high enough for good 1. 

Intertemporal Choice (in discrete time)

We can treat consumption now vs. consumption later as 2 different goods. Typically, intertemporal choice problems will involve adding up these separable utility functions. Mathematically:
\(u(c_1, c_2) = u^A (c_1) + u^B (c_2)\)
Here, the various goods have no dependence on each other. A counter-example to this is habit formation, where past decisions impact present utility. Now, consider the most popular general case of intertemporal choice:
max \(u = \frac{c_1 ^{1 - \theta}}{1 - \theta} + (\frac{1}{1 + \rho}) \frac{c_2 ^{1 - \theta}}{1 - \theta}\)
subject to \(c_2 = (y - c_1)(1 + r)\)
Recall that \(\rho\) is the rate of time preference (or valuation place on having a good at an earlier rather than later date), \(r\) signifies the real interest rate, and \(\frac{1}{\theta}\) is the elasticity of intertemporal substitution (or measure of responsiveness of the growth rate of consumption to the real interest rate). The constraint reads: second period consumption is equal to the present income, minus present consumption, multiplied by one plus the real interest rate.  
Substituting in the constraint and solving for the optimal proportion of present to future consumption will yield
\(u = \frac{c_1 ^{1 - \theta}}{1 - \theta} + (\frac{1}{1 + \rho}) [\frac{(y - c_1)(1 + r)}{1 - \theta}]^{1 - \theta}\)
\(\frac{\partial u}{\partial c_1} = c_1 ^{- \theta} + \frac{1}{1 - \rho} c_2 ^{- \theta} (-1)(1 + r) = 0\)
\((\frac{c_2}{c_1})^* = (\frac{1 + r}{1 + \rho})^{\frac{1}{\theta}}\)
Exponents are elasticities! See why \(\frac{1}{\theta}\) is the elasticity of intertemporal substitution. This final equation states that the optimal proportion of future consumption to present consumption depends on those three parameters. For example, if \(\frac{1}{\theta} = \frac{1}{10}\), then your consumption is hugely responsive to interest rates (will be pushed to the future if the interest rate is relatively high). 
Finally, a real-world example. If our economy grows at about 3%, and consumption is two-thirds of GDP growth, then it grows at 2%. What is the discount rate if we know the interest rate is 4% and assume log-utility?
\(\ln 1.02 = \ln 1.04 - \ln (1 + x) \)
\(x = 0.02 \Longrightarrow \rho = 2 \%\)
There are two reasons for present consumption to be equal to future consumption:
  1. \(r = \rho\) : the effects will cancel out
  2. \(\theta \rightarrow \infty\) (or even 4, 5): you demand perfect equality of consumption across time, without regard to the interest or discount rates. Mathematically, this makes their fraction matter infinitely less.