Why Interpretation Matters
A utility function is a way to represent preferences over bundles. Its numerical values make comparisons convenient, but they can also invite mistakes: a larger utility number is not necessarily twice as desirable, and a change in utility is not automatically a measure of how much better a bundle is. Similar care is needed with marginal utility and the marginal rate of substitution (MRS). Marginal utility is a derivative, while MRS is a ratio of derivatives whose sign must be distinguished from the slope of an indifference curve.
The focus here is interpretation rather than finding a consumer’s best affordable bundle. For that purpose, the key questions are: does a proposed change preserve preference rankings, is a derivative being mistaken for a total or finite change, and is MRS being reported as a positive trade-off or as a signed slope?
Utility Magnitudes Are Not Preference Intensities
Suppose two bundles receive utility values 12 and 6. The first bundle is ranked above the second under that representation. It does not follow that the first is “twice as good.” A strictly increasing transformation can change the values and differences while leaving every ranking unchanged. For example, if \(U\) represents preferences, then \(V=3U+5\) represents the same preferences: multiplying by a positive number and adding a constant preserves the order.
The qualification “strictly increasing” matters. A strictly decreasing transformation reverses rankings, while a transformation that is not strictly increasing may collapse distinct utility values or change rankings. Only transformations that preserve the order are guaranteed to be equivalent ordinal representations.
Worked Example: Same Ranking, Different Utility Magnitudes
A consumer’s bundles are scored by \(U(x,y)=2x+y\). Compare bundle A, \((3,2)\), with bundle B, \((1,8)\). Then compare the scores under \(V=3U+5\). Explain what can and cannot be concluded.
Method. Evaluate each representation at both bundles, then check whether the ordering is preserved. The common mistake is to interpret the numerical difference or ratio as a measure of how much more the consumer likes one bundle.
For A, \(U(A)=2(3)+2=8\). For B, \(U(B)=2(1)+8=10\). Since \(10>8\), B is ranked above A by \(U\).
Now transform each score. \(V(A)=3(8)+5=29\), and \(V(B)=3(10)+5=35\). Since \(35>29\), \(V\) gives the same ranking.
Under \(U\), the difference is \(10-8=2\) and the ratio is \(10/8=1.25\). Under \(V\), the difference is \(35-29=6\) and the ratio is \(35/29\approx1.2069\). The rankings agree, but the differences and ratios do not. The valid conclusion is that B is preferred to A, not that B is 25 percent better, or that the preference gap has a particular size.
Check. Both representations rank B above A, as required. The transformed numbers confirm that a change in utility scale can alter differences and ratios without altering the preference ordering.
Marginal Utility Is Not Total Utility
For a differentiable utility function, the marginal utility of \(x\) is the partial derivative \(MU_x=\partial U/\partial x\), holding \(y\) fixed. It describes the local rate of change in utility as \(x\) changes. Similarly, \(MU_y=\partial U/\partial y\), holding \(x\) fixed. These are local measures, not the total utility of the bundle and not necessarily the exact change in utility from a finite increase in a good.
For a small change \(\Delta x\) with \(y\) fixed, the approximation \(\Delta U\approx MU_x\Delta x\) uses the marginal utility at the starting point. If marginal utility changes as \(x\) changes, this approximation need not equal the exact difference \(U(x+\Delta x,y)-U(x,y)\). When \(U\) is differentiable, the exact difference can instead be found by evaluating the utility function at both endpoints.
Worked Example: Marginal Utility Versus a Finite Utility Change
Let \(U(x,y)=\sqrt{xy}\) for positive quantities. At bundle \((4,9)\), find total utility, both marginal utilities, and the exact utility change when \(x\) increases from 4 to 5 while \(y\) stays at 9. Compare the exact change with the initial marginal-utility approximation.
Method. First evaluate the utility and derivatives at the starting bundle. Then evaluate utility at the new bundle. The common mistake is to report \(MU_x\) itself as either total utility or the exact change from a one-unit increase.
At \((4,9)\), total utility is \(U(4,9)=\sqrt{36}=6\). Differentiating gives \(MU_x=\tfrac{1}{2}\sqrt{y/x}\) and \(MU_y=\tfrac{1}{2}\sqrt{x/y}\). Thus \(MU_x(4,9)=\tfrac{1}{2}\sqrt{9/4}=3/4=0.75\), while \(MU_y(4,9)=\tfrac{1}{2}\sqrt{4/9}=1/3\).
After \(x\) increases to 5, utility is \(U(5,9)=\sqrt{45}=3\sqrt{5}\approx6.7082\). The exact change is \(3\sqrt{5}-6\approx0.7082\). The approximation based on the initial marginal utility is \(MU_x\Delta x=0.75(1)=0.75\). It is close, but not exact.
Check. Total utility at the starting bundle is 6; \(MU_x=0.75\) is the local rate of change there; and the exact finite change is approximately 0.7082. The distinction is consistent with marginal utility declining as \(x\) increases with \(y\) fixed.
Signed Slope and Positive MRS
For a differentiable indifference curve \(U(x,y)=c\), where \(c\) is a fixed utility level, implicit differentiation gives its slope when \(MU_y\neq0\):
If both marginal utilities are positive, the slope is negative: keeping utility unchanged requires less \(y\) when \(x\) increases. A common convention defines the positive MRS of \(x\) for \(y\) as \(MRS_{xy}=MU_x/MU_y\). Under that convention, MRS is the absolute value of the indifference-curve slope, not the signed slope itself. Other contexts may describe the trade-off using the signed slope, so a careful answer states which quantity is being reported.
Worked Example: Reporting the Slope and MRS Correctly
For \(U(x,y)=xy\) with positive quantities, evaluate the marginal utilities, signed indifference-curve slope, and positive \(MRS_{xy}\) at \((2,6)\). Then use the same indifference curve to find the exact change in \(y\) when \(x\) rises from 2 to 2.2.
Method. Use the derivatives for the local slope, then hold utility fixed to find the exact finite change. The common mistake is to call the negative slope the positive MRS, or to treat a local slope as an exact trade-off over a finite change.
At \((2,6)\), \(MU_x=y=6\) and \(MU_y=x=2\). Therefore \(MRS_{xy}=MU_x/MU_y=6/2=3\), while the signed slope is \(dy/dx=-6/2=-3\).
The initial utility is \(2(6)=12\). On the same indifference curve, utility must remain 12. When \(x=2.2\), solve \(2.2y=12\), giving \(y=12/2.2=60/11\approx5.4545\). The exact change in \(y\) is \(60/11-6=-6/11\approx-0.5455\). The average decrease in \(y\) per unit increase in \(x\) over this interval is \((6/11)/0.2\approx2.7273\), not exactly 3.
Check. The signed slope is negative, the conventional MRS is positive, and the finite trade-off differs from the local MRS because the slope changes along the curve.
What a Monotonic Transformation Does to Marginal Utilities and MRS
A strictly increasing transformation preserves rankings, but it does not generally preserve the numerical marginal utilities. Let \(V=g(U)\), where \(g\) is differentiable and \(g'(U)>0\) at the bundle being considered. By the chain rule, \(MU_x^V=g'(U)MU_x^U\) and \(MU_y^V=g'(U)MU_y^U\). Each marginal utility is scaled by the same positive factor at that bundle. Their ratio is therefore unchanged, provided \(MU_y^U\neq0\).
This explains why marginal utility values depend on the chosen representation while the local MRS does not, under these conditions. The statement concerns a differentiable, strictly increasing transformation with positive derivative at the relevant utility level. If the derivative is zero, the transformed marginal utilities can both become zero at that point, and their ratio may not be defined by the transformed derivatives there.
Worked Example: Testing MRS Invariance Under a Transformation
Let \(U(x,y)=xy\) for positive quantities and define \(V(x,y)=\ln U(x,y)\). At \((2,4)\), calculate the marginal utilities and \(MRS_{xy}\) under both representations. Explain why the marginal utilities differ while the MRS agrees.
Method. Differentiate each representation directly and compare the ratios. The common mistake is to assume that equivalent utility representations must have identical marginal utilities.
For \(U=xy\), \(MU_x^U=y\) and \(MU_y^U=x\). At \((2,4)\), these are \(4\) and \(2\), respectively, so \(MRS_{xy}^U=4/2=2\).
For \(V=\ln(xy)=\ln x+\ln y\), \(MU_x^V=1/x\) and \(MU_y^V=1/y\). At \((2,4)\), these are \(1/2=0.5\) and \(1/4=0.25\), so \(MRS_{xy}^V=0.5/0.25=2\).
Since \(U(2,4)=8>0\) and \(g(u)=\ln u\) has \(g'(u)=1/u>0\) for \(u>0\), the transformation is strictly increasing and differentiable at the relevant utility level. Its scaling factor is \(g'(8)=1/8\), so it scales the original marginal utilities \(4\) and \(2\) to \(0.5\) and \(0.25\). The ratio stays the same.
Check. Both representations give \(MRS_{xy}=2\), although their marginal-utility values differ. The logarithm is valid here because \(xy>0\).
Common Errors to Avoid
- Reading utility as a cardinal score. If one bundle has utility 20 and another has utility 10, the ranking is meaningful under that representation; “twice as much utility” is not.
- Comparing utility differences across representations. A strictly increasing transformation can change the size of a difference without changing any ranking.
- Calling a marginal utility a total change. A derivative is a local rate. For a finite change, evaluate utility at both endpoints or use an appropriate integral.
- Dropping the minus sign from the slope. With positive marginal utilities, the indifference-curve slope is negative. The conventional MRS \(MU_x/MU_y\) is positive.
- Assuming a local MRS is an exact finite trade-off. The MRS can change along an indifference curve, so a finite movement generally has a different average trade-off.
- Assuming every transformation is equivalent. The transformation must preserve rankings; a strictly decreasing transformation reverses them, and a non-monotonic transformation may change them.
Check Your Understanding
For each problem, distinguish the numerical representation from the preference or slope conclusion it supports.
- A utility representation assigns bundle A a value of 7 and bundle B a value of 11. A second representation is \(V=2U-3\). Calculate both transformed values and state the ranking. (Answer: \(V(A)=11\), \(V(B)=19\); B ranks above A under both representations.)
- Let \(U(x,y)=x^2+y\). At \((3,4)\), find total utility and \(MU_x\). If \(x\) rises to 3.1 with \(y\) fixed, find the exact utility change and the initial-marginal-utility approximation. (Answer: \(U=13\), \(MU_x=6\), exact change \(=3.1^2-3^2=0.61\), approximation \(=6(0.1)=0.60\).)
- For \(U(x,y)=x+2y\), find \(MRS_{xy}\) and the signed indifference-curve slope where derivatives are defined. (Answer: \(MRS_{xy}=1/2\), signed slope \(=-1/2\).)
- Let \(U(x,y)=x^2y\) for positive quantities and \(V=\sqrt{U}\). At \((2,3)\), calculate \(MRS_{xy}\) using each representation. (Answer: under \(U\), \(MU_x=2xy=12\), \(MU_y=x^2=4\), so \(MRS_{xy}=3\); under \(V=x\sqrt{y}\), \(MU_x^V=\sqrt{3}\) and \(MU_y^V=1/\sqrt{3}\), so \(MRS_{xy}=3\).)
- A student replaces \(U\) by \(W=-U\) and claims that the new function is an equivalent representation. Is the claim correct? Explain what happens to the ranking of any two bundles with distinct original utility values. (Answer: No. The transformation is strictly decreasing, so it reverses the ranking; a bundle with higher \(U\) has lower \(W\).)