Why Small Changes Should Not Cause Abrupt Reversals
Consumers often compare bundles that differ only slightly: a little more of one good, a little less of another, or a small change in both. A preference ranking is continuous when comparisons behave stably as the bundles change by small amounts. If one bundle is strictly preferred to another, sufficiently small changes to either bundle should not suddenly reverse that ranking.
Continuity is about nearby comparisons, not about whether a consumer's preferences are sensible in every other respect. It does not say that a ranking stays the same after a large change in a bundle. Nor does it require every pair of bundles to have a strict ranking. It rules out abrupt reversals at the boundary between bundles that are close together.
A sequence of bundles converges to a bundle when its quantities get arbitrarily close to the corresponding quantities in that bundle. For two goods, write \(x=(x_1,x_2)\), where \(x_1\) and \(x_2\) are the quantities of the two goods. Convergence means that both coordinates approach their limiting quantities. This definition also applies when more than two goods are being compared.
A useful consequence is that a strict preference is locally stable. If \(x \succ y\), then there is some sufficiently small neighborhood around \(x\) and \(y\) in which every bundle near \(x\) remains strictly preferred to every bundle near \(y\). The size of that neighborhood can depend on the particular pair. A preference may be robust to small changes around one comparison but less robust around another.
What Continuity Does—and Does Not—Require
Continuity does not mean that bundles change without affecting rankings. A sufficiently large change can certainly alter which bundle is preferred. Instead, continuity says that there is no strict preference reversal caused by an arbitrarily small change at a fixed comparison. The distinction is between a ranking changing as bundles move across a meaningful distance and a ranking flipping immediately when a bundle moves by any positive amount.
The definition also does not require a strict ranking whenever two bundles are compared. If \(x \sim y\), meaning \(x \succeq y\) and \(y \succeq x\), nearby comparisons might resolve that tie in either direction. Continuity protects strict rankings from nearby reversal; it does not promise that a tie will remain a tie.
A continuous utility function is a convenient way to establish continuity. If a consumer's preferences are represented by a continuous utility function \(U\), the ranking rule is \(x \succeq y\) exactly when \(U(x) \ge U(y)\). When bundles approach \(x\) and \(y\), their utility values approach \(U(x)\) and \(U(y)\). Therefore, a ranking that holds along the sequence cannot fail in the limit.
The middle implication follows from continuity of \(U\): if \(x_n \to x\) and \(y_n \to y\), then \(U(x_n) \to U(x)\) and \(U(y_n) \to U(y)\). Limits preserve a weak inequality. A continuous utility function is a sufficient way to demonstrate continuous preferences; the definition of continuous preferences itself is about rankings and does not require a particular utility formula.
Worked Example: Checking a Robust Linear Ranking
A consumer ranks bundles using \(U(x_1,x_2)=2x_1+x_2\). Compare \(A=(4,2)\) with \(B=(3,3)\). Then suppose each coordinate of each bundle can change by at most \(0.1\). Determine whether the strict ranking is guaranteed to persist. Use the utility method and state the mistake it avoids.
Step 1: Compare the original bundles. The utility of \(A\) is \(2(4)+2=10\). The utility of \(B\) is \(2(3)+3=9\). Thus \(A\succ B\), with a utility difference of \(10-9=1\).
Step 2: Find the lowest possible utility near \(A\). Each coordinate can fall by at most \(0.1\), and the coefficient on \(x_1\) is 2. The maximum utility decrease is \(2(0.1)+0.1=0.3\). Therefore, any allowed \(A'\) has \(U(A')\ge 10-0.3=9.7\).
Step 3: Find the highest possible utility near \(B\). Each coordinate can rise by at most \(0.1\), so the maximum utility increase is also \(2(0.1)+0.1=0.3\). Therefore, any allowed \(B'\) has \(U(B')\le 9+0.3=9.3\).
Step 4: Compare the bounds. Since \(9.7>9.3\), every allowed \(A'\) has higher utility than every allowed \(B'\). The strict ranking persists. The method avoids the mistake of checking only one chosen perturbation and assuming it represents all nearby pairs.
Check: Even in the worst case for \(A'\) and the best case for \(B'\), the utility gap is at least \(9.7-9.3=0.4>0\).
Why Strict Rankings Persist Locally
The stability result follows directly from the definition and completeness of preferences. Completeness means that for any two bundles, at least one is weakly preferred to the other. Suppose \(x\succ y\), but imagine that there is no neighborhood in which bundles near \(x\) remain strictly preferred to bundles near \(y\). Then one could choose increasingly close pairs \(x_n,y_n\) for which \(x_n\not\succ y_n\). By completeness, \(y_n\succeq x_n\) for each pair. As those bundles approach \(x,y\), continuity implies \(y\succeq x\), contradicting \(x\succ y\).
Assume \(x\succ y\), so \(x\succeq y\) and it is not true that \(y\succeq x\).
Then arbitrarily close pairs \(x_n,y_n\) can be found for which \(x_n\not\succ y_n\). By completeness, \(y_n\succeq x_n\).
Because \(x_n\to x\), \(y_n\to y\), and \(y_n\succeq x_n\) for every \(n\), continuity implies \(y\succeq x\).
The conclusion \(y\succeq x\) conflicts with \(x\succ y\). Therefore, some neighborhood must preserve the strict ranking.
This argument explains why continuity matters for bundle comparisons. A strict ranking has a positive local margin in the sense that it cannot be overturned by comparisons arbitrarily close to the original pair. The argument does not calculate a numerical neighborhood size; that requires more information, such as a particular utility function and a bound on the changes in quantities.
Worked Example: A Ranking Reversal Under Lexicographic Preferences
Suppose a consumer ranks two-good bundles lexicographically: first compare the quantity of good 1; the bundle with more good 1 is preferred, regardless of good 2. If the amounts of good 1 are equal, compare good 2. Test continuity using \(A=(2,1)\) and \(B=(2,0)\). Name the method and the common mistake it avoids.
Step 1: Rank the original pair. Both bundles contain 2 units of good 1, so compare good 2. Since \(1>0\), \(A\succ B\).
Step 2: Make a sequence of nearby bundles. For each positive integer \(n\), let \(A_n=(2-1/n,1)\). The second coordinate stays at 1, while the first approaches 2. Thus \(A_n\to A\).
Step 3: Compare every nearby bundle with \(B\). For every \(n\), \(B\) has 2 units of good 1 and \(A_n\) has \(2-1/n<2\). Lexicographic ranking therefore gives \(B\succ A_n\), even though the difference in good 1 becomes arbitrarily small.
Step 4: Apply the continuity test. The sequence has \(B\succeq A_n\) for every \(n\), and its limit is the comparison \(B\) with \(A\). Continuity would require \(B\succeq A\). But \(A\succ B\), so that weak ranking is false. These preferences are not continuous.
Check: For example, at \(n=100\), \(A_n=(1.99,1)\), only \(0.01\) unit of good 1 below \(A\), yet \(B\succ A_n\) while \(A\succ B\). The sequence method avoids the mistake of judging continuity from a single small change rather than checking what happens as the change tends to zero.
Using a Continuous Nonlinear Utility Function
The utility test also works when the utility formula is nonlinear. For products of nonnegative quantities, a product comparison can sometimes establish the ranking without calculating square roots. When a strict gap remains between the relevant utility values, small changes in both quantities can be bounded to show that the ranking survives.
Worked Example: Comparing Nearby Bundles With Product Utility
A consumer's preferences are represented by \(U(x_1,x_2)=x_1x_2\) for nonnegative quantities. Compare \(A=(9,4)\) and \(B=(4,4)\). If every coordinate of both bundles can vary by at most \(0.1\), determine whether all such nearby comparisons preserve the ranking. Use bounded utility values and identify the mistake the method avoids.
Step 1: Compare the original utilities. \(U(A)=9(4)=36\), while \(U(B)=4(4)=16\). Therefore \(A\succ B\).
Step 2: Bound utility near \(A\) from below. In the allowed neighborhood, both coordinates of \(A'\) are at least \(8.9\) and \(3.9\), respectively. Since the quantities are nonnegative, \(U(A')\ge 8.9(3.9)=34.71\).
Step 3: Bound utility near \(B\) from above. Both coordinates of \(B'\) are at most \(4.1\). Thus \(U(B')\le 4.1(4.1)=16.81\).
Step 4: Compare the bounds. The lowest possible utility near \(A\) exceeds the highest possible utility near \(B\), since \(34.71>16.81\). Every allowed nearby pair therefore preserves \(A'\succ B'\).
Check: The minimum guaranteed utility gap is \(34.71-16.81=17.90>0\). Bounding both neighborhoods avoids the mistake of checking only one direction of change—for example, lowering \(A\) without also considering that \(B\) could rise.
Common Mistakes and Exam Traps
- Treating continuity as no ranking changes at all. Continuity concerns arbitrarily small changes near a comparison. Large changes can alter a ranking without violating continuity.
- Checking only one nearby bundle. A stability claim concerns nearby comparisons. If both bundles can move, consider changes to both, or use the sequence definition.
- Confusing a tie with a reversal. A strict ranking that becomes a tie for arbitrarily nearby bundles may still be a failure of strict local stability. A tie at the original comparison, however, is not itself a strict ranking to preserve.
- Assuming every preference rule is continuous. Lexicographic preferences provide a counterexample: any positive change in the first good can outweigh a difference in the second, however small that first-good change is.
- Assuming a utility formula proves continuity without checking it. A continuous utility representation is sufficient, but a discontinuous utility formula does not establish continuous preferences merely because it assigns numbers to bundles.
Check Your Understanding
For each problem, decide what the stated comparison or sequence implies about continuity. Answers are provided so you can check your reasoning.
- A consumer has \(U(x_1,x_2)=x_1+3x_2\). Compare \(A=(5,1)\) and \(B=(3,1)\). If each coordinate of each bundle changes by at most \(0.1\), does the strict ranking necessarily persist? (Answer: Yes. The original utilities are 8 and 6; the maximum change in either utility is \(0.1+3(0.1)=0.4\), so the worst-case utilities are 7.6 and 6.4.)
- Under lexicographic preferences, compare \(A=(1,4)\) with \(B=(1,2)\). If \(A_n=(1-1/n,4)\), what is the ranking between \(B\) and \(A_n\), and what does the sequence show? (Answer: \(B\succ A_n\) for every \(n\), although \(A_n\to A\) and \(A\succ B\); the preferences are not continuous.)
- A consumer has \(U(x_1,x_2)=x_1x_2\). Compare \(A=(6,5)\) and \(B=(5,4)\) when each coordinate of each bundle can vary by at most \(0.1\). Is the ranking guaranteed to persist? (Answer: Yes. The minimum utility near \(A\) is \(5.9(4.9)=28.91\), and the maximum near \(B\) is \(5.1(4.1)=20.91\), so \(A'\succ B'\) throughout those neighborhoods.)
- Suppose \(x_n\succeq y_n\) for every \(n\), \(x_n\to x\), and \(y_n\to y\). What ranking must hold if preferences are continuous? (Answer: \(x\succeq y\).)
- A consumer has \(U(x_1,x_2)=x_1+2x_2\). Bundles \(A=(4,3)\) and \(B=(3,3)\) are each perturbed in each coordinate by at most \(0.2\). Find a lower bound for utility near \(A\), an upper bound near \(B\), and determine whether the ranking is guaranteed. (Answer: The bounds are \(10-0.2-2(0.2)=9.4\) and \(9+0.2+2(0.2)=9.6\); the bounds overlap, so this calculation does not guarantee that every nearby pair preserves the ranking.)