Convex Preferences and Bowed Indifference Curves

Why Convexity Matters for Indifference Curves

Suppose two consumption bundles are equally ranked. One might contain relatively more of good 1, while the other contains relatively more of good 2. Convex preferences express a preference for balance: a weighted average of those bundles is at least as good as either one. On an indifference map, that preference is represented by curves bowed toward the origin, subject to the assumptions that make the map interpretable.

The central test is not whether one particular average is preferred. Convexity is a condition on every pair of bundles and every weight between zero and one. A midpoint calculation is useful because it is easy to picture and calculate, but it is only one special case of that full condition.

Definition: Preferences are convex if, whenever bundles \(A\) and \(B\) are each weakly preferred to bundle \(C\), every weighted average \(\lambda A+(1-\lambda)B\), for \(0\leq\lambda\leq1\), is also weakly preferred to \(C\). Equivalently, for any given bundle \(C\), the set of bundles weakly preferred to \(C\) is convex. Preferences are strictly convex if a weighted average with \(0<\lambda<1\) of two distinct, indifferent bundles is strictly preferred to those bundles.

A weighted average is calculated separately for each good. For two goods, if \(A=(x_{1A},x_{2A})\) and \(B=(x_{1B},x_{2B})\), then

\lambda A+(1-\lambda)B = \bigl(\lambda x_{1A}+(1-\lambda)x_{1B},\; \lambda x_{2A}+(1-\lambda)x_{2B}\bigr).

At \(\lambda=\tfrac12\), the weighted average is the midpoint. Convexity requires the ranking condition for all \(\lambda\), not just for \(\tfrac12\). Consequently, finding a midpoint that is worse than two indifferent bundles disproves convexity. Finding one midpoint that is better, however, does not establish convexity.

From Preferred Averages to Bowed Curves

An indifference curve contains bundles that are equally ranked. Consider two distinct bundles on the same curve, with one containing more of good 1 and the other containing more of good 2. Under convex preferences, any weighted average of these bundles is weakly preferred to them. In particular, their midpoint must be at least as good as either endpoint.

To translate that ranking into a curve shape, assume goods are desirable in the sense that more of good 2, holding good 1 fixed, is strictly preferred, and that the relevant indifference curve can be drawn as a single-valued graph of good 2 against good 1. At the midpoint quantity of good 1, the curve cannot lie above the midpoint bundle: if it did, that curve bundle would have more good 2 and be strictly preferred to the midpoint, contradicting the midpoint’s weak preference to bundles on the curve. Thus the curve lies on or below the straight chord connecting the two bundles. This is the geometric meaning of being bowed toward the origin.

With strictly convex preferences and the same graph and monotonicity assumptions, the average is strictly preferred, so the indifference curve lies strictly below the chord between distinct points. Weak convexity permits equality: a straight segment may appear on an indifference curve. Therefore, “convex preferences” does not always mean every indifference curve is strictly curved.

\text{two indifferent bundles} \quad\Longrightarrow\quad \text{their weighted average is weakly preferred} \quad\Longrightarrow\quad \text{the curve is weakly below their chord}.

Why the Full Weight Condition Matters

The midpoint condition says that the average of two weakly preferred bundles is weakly preferred as well. It is necessary for convexity, since \(\lambda=\tfrac12\) is one of the allowed weights. But it is not, by itself, a definition of convexity: a condition checked only at midpoints can hold without the corresponding condition holding at all weights. In mathematical terms, a set that contains the midpoint of every pair of its points need not be convex without additional regularity assumptions.

Continuity is one familiar regularity condition under which midpoint convexity can be extended to convexity. For preference analysis, the safe test is to use the definition directly: check all weighted averages, or establish that the relevant upper contour sets are convex. A finite set of midpoint checks can reveal a violation, but cannot prove the full condition.

Identify the ranking.
Confirm that both endpoint bundles are indifferent, or that each is weakly preferred to the same reference bundle.

Calculate the average.
Apply the same weight to each good’s quantity; do not average one coordinate and leave the other unchanged.

Compare its ranking.
Convexity requires the average to be weakly preferred. A worse average is a counterexample.

Read the curve carefully.
With monotonicity and a graph-shaped indifference curve, compare the curve’s height at the average quantity with the chord’s height.

Worked Example: Testing a Curved Indifference Map

A hypothetical consumer has an indifference curve described by \(x_2=24/x_1\) for positive quantities. Bundles \(A=(3,8)\) and \(B=(8,3)\) lie on this curve. Assuming more of either good is preferred when the other is held fixed, use their midpoint to assess the curve’s convexity implication.

Method. Calculate the coordinate-wise midpoint, then compare its good-2 quantity with the curve’s quantity at the same amount of good 1. The calculation tests the required midpoint implication for these two bundles; it does not alone prove convexity for all pairs and weights.

Step 1: Verify the endpoints. At \(A\), \(24/3=8\). At \(B\), \(24/8=3\). Both bundles therefore lie on the stated curve.

Step 2: Calculate the midpoint. The good-1 quantity is \((3+8)/2=5.5\), and the good-2 quantity is \((8+3)/2=5.5\). The midpoint is \((5.5,5.5)\).

Step 3: Find the curve’s height at good 1 quantity 5.5. The curve gives \(x_2=24/5.5=48/11\), approximately \(4.3636\). The midpoint has \(x_2=5.5\), which is greater than \(48/11\).

Under the stated monotonicity assumption, the midpoint lies above the indifference curve and is preferred to the bundles on it. The curve is below the chord at this midpoint, consistent with convex preferences. The arithmetic avoids a common mistake: comparing the midpoint’s good-2 quantity with an endpoint’s quantity instead of with the curve at the midpoint’s good-1 quantity.

Weak Convexity and Straight Segments

A straight indifference curve is not automatically evidence against convex preferences. When preferences are weakly convex, the weighted average of indifferent bundles may itself remain indifferent. A straight segment is therefore consistent with convexity, even though strict convexity would require a distinct interior average to be strictly preferred.

Worked Example: A Linear Indifference Segment

On a hypothetical indifference map, bundles \(A=(2,10)\) and \(B=(8,4)\) lie on the same straight indifference line. The map’s equal-ranking line through them is \(x_1+x_2=12\), with bundles on higher-sum parallel lines ranked above it. Calculate a weighted average and determine what it shows.

Method. Use the midpoint to check whether this average remains on the same indifference line. The ranking rule is stipulated by the map, so no additional preference representation is needed.

Step 1: Check the endpoints. For \(A\), \(2+10=12\). For \(B\), \(8+4=12\). Both are on the stated line.

Step 2: Calculate the midpoint. Its coordinates are \(((2+8)/2,(10+4)/2)=(5,7)\).

Step 3: Check its ranking. The coordinate sum is \(5+7=12\), so the midpoint remains on the same indifference line. It is not strictly preferred to the endpoints.

This example is consistent with weak convexity but does not satisfy strict convexity for these bundles. A common error is to label the preferences nonconvex simply because the curve is straight. Convexity permits equality; strict convexity does not.

Using a Midpoint to Find a Violation

The midpoint test is especially decisive when it finds an average that is worse than its indifferent endpoints. Such a result directly contradicts convexity. On a graph with more-is-better preferences, this can happen when an indifference curve lies above the chord between two of its points.

Worked Example: Detecting a Curve Bowed Away from the Origin

A hypothetical indifference map shows \(A=(2,9)\) and \(B=(10,3)\) on the same curve. At good-1 quantity \(6\), the curve requires good-2 quantity \(7\). Assume more of either good is strictly preferred when the other is held fixed. Does the map satisfy the convexity implication for these points?

Method. Compare the midpoint bundle with the curve’s bundle at the same good-1 quantity. If the midpoint lies below that curve, monotonicity makes it worse than the indifferent endpoints, violating convexity.

Step 1: Calculate the midpoint coordinates. Good 1 is \((2+10)/2=6\). Good 2 is \((9+3)/2=6\). Thus the midpoint is \((6,6)\).

Step 2: Compare with the curve. At good-1 quantity \(6\), the curve has good-2 quantity \(7\). Since \(7>6\), the curve bundle \((6,7)\) has more good 2 than the midpoint \((6,6)\), with good 1 held fixed.

Step 3: Apply the ranking. Strict monotonicity makes \((6,7)\) strictly preferred to \((6,6)\). But \((6,7)\) is indifferent to \(A\) and \(B\), while convexity requires their midpoint to be weakly preferred to them. The map therefore violates convexity for this pair.

The relevant chord also confirms the geometry: at good-1 quantity \(6\), the chord’s good-2 quantity is the midpoint value \(6\), while the curve requires \(7\). The curve is above the chord, rather than weakly below it. The mistake to avoid is treating the endpoints’ indifference as sufficient: convexity also imposes a ranking condition on their average.

Common Mistakes and Exam Interpretation

Using only one midpoint to claim convexity. One failed midpoint disproves convexity; one successful midpoint does not establish the condition for every pair and every weight.
Confusing weak with strict convexity. Weak convexity allows an average to be indifferent to its endpoints. Strict convexity requires a distinct interior average of indifferent bundles to be strictly preferred.
Reading the picture without stating assumptions. To infer that a curve lies below a chord from preference rankings, identify the monotonicity assumption and confirm that the curve is a graph at the quantity being compared.
Comparing the wrong quantities. At an average amount of good 1, compare the average bundle’s amount of good 2 with the curve’s amount of good 2 at that same good-1 quantity.
Assuming all convex preferences look strictly curved. A straight indifference segment can be consistent with weak convexity.

For a candidate studying material aligned with the SOA VEE Economics (Validation by Educational Experience) syllabus, the reliable procedure is to state the ranking condition, calculate the weighted bundle, and interpret the result under the stated assumptions. Convexity describes rankings of averages; the bowed-curve picture is its geometric consequence when the indifference map and monotonicity conditions support that interpretation.

Key takeaway: Convex preferences require every weighted average of two bundles weakly preferred to a reference bundle to remain weakly preferred to it. For two indifferent bundles, this places their average at least as high in the ranking. With monotonicity and a graph-shaped indifference curve, the curve lies weakly below the chord between them; strict convexity gives a strict bow.

Check Your Understanding

For each problem, calculate the average before deciding what the preference or curve information implies.

Two bundles \(A=(4,12)\) and \(B=(10,6)\) are indifferent. What is their midpoint? If it is strictly worse than both, are the preferences convex? (Answer: The midpoint is \((7,9)\); no, that outcome violates convexity.)
Two indifferent bundles are \(A=(2,8)\) and \(B=(8,2)\). Their midpoint is also indifferent. Does this contradict convexity, and does it establish strict convexity for this pair? (Answer: It does not contradict weak convexity; it does not satisfy strict convexity for this pair.)
A curve is \(x_2=30/x_1\), and indifferent endpoints are \((3,10)\) and \((10,3)\). Find the midpoint and the curve’s good-2 quantity at its good-1 quantity. Is the midpoint above or below the curve? (Answer: The midpoint is \((6.5,6.5)\); the curve quantity is \(30/6.5=60/13\), approximately \(4.6154\); the midpoint is above the curve.)
Indifferent bundles are \((1,11)\) and \((9,3)\). At their average good-1 quantity, the curve has good-2 quantity \(8\). Under strict monotonicity in good 2, does this support or violate convexity? (Answer: The midpoint is \((5,7)\), below the curve bundle \((5,8)\); it violates convexity.)
A map passes midpoint tests for several pairs of indifferent bundles. Is that enough to conclude the preference relation is convex? State the reason. (Answer: No. Convexity requires the ranking condition for all pairs and every weight from zero to one; finitely many midpoint checks are not sufficient.)

What You'll Learn