Why Designs Matter in the Romer Model
A new design is not itself a finished good. In the Romer variety-expansion model, a design makes a distinct intermediate input available, and final-good producers combine those inputs with labor to produce output. The distinction matters: more designs can increase productivity even when the total quantity of intermediate inputs is unchanged, but research also uses resources that could have supported current production.
The model separates three activities. Final production combines labor with intermediate inputs. Intermediate production supplies the inputs associated with existing designs. Research uses labor to create additional designs. The analysis below focuses on how those designs enter production, rather than on the market arrangements that reward inventors or on the economy’s long-run growth rate.
The production function assumes that final producers can use the available designs and that each intermediate input has diminishing returns: because \(0<\alpha<1\), doubling one input while holding other inputs fixed less than doubles its contribution to the integral. The factor \(L_Y^{1-\alpha}\) captures the contribution of final-production labor. In this simplified setup, intermediate inputs are produced using capital one-for-one, so their quantities sum to \(K\); research labor is not included in \(L_Y\).
The design stock \(A\) is treated as a continuous measure in the aggregate model, even though an individual design is conceptually distinct. The research equation says that the rate at which designs are created is higher when research employs more labor and when there is a larger existing stock of designs. It is a model of idea creation; by itself, it does not specify who owns a design or how research is financed.
How Variety Enters Final Production
A useful benchmark is symmetric allocation: every available design receives the same quantity of intermediate input. If total input \(K\) is spread evenly over \(A\) designs, then \(x_i=K/A\). Substituting that allocation into the production function gives a compact expression for output.
With equal allocation, each design supplies \(x_i=K/A\).
Each of the \(A\) designs contributes \((K/A)^\alpha\), while final labor contributes \(L_Y^{1-\alpha}\).
Multiplying the identical contributions by \(A\) yields \(A(K/A)^\alpha=A^{1-\alpha}K^\alpha\).
This expression shows the variety effect precisely. Holding \(L_Y\) and \(K\) fixed, a larger \(A\) raises output because the same total intermediate-input stock can be spread across more distinct inputs. The exponent \(1-\alpha\) is positive, so output rises with variety. Equivalently, output per unit of intermediate capital is
At fixed \(L_Y\) and \(K\), this ratio also rises with \(A\). It is a productivity comparison conditional on those inputs, not a claim that research is costless. When labor moves from final production into research, current \(L_Y\) falls; whether the future output benefit justifies that cost is a separate question.
Worked Example: More Designs With the Same Intermediate Stock
Suppose \(\alpha=1/2\), final-production labor is \(L_Y=9\), and the intermediate stock is \(K=16\). Compare output with \(A=4\) designs and \(A=9\) designs, assuming equal allocation in each case. Calculate the change in output and output per unit of intermediate capital.
Method. Use the symmetric-allocation production formula, keeping \(L_Y\) and \(K\) fixed. This isolates the effect of variety.
With \(A=4\), each design receives \(x_i=16/4=4\). The sum of intermediate contributions is \(4\sqrt{4}=4(2)=8\), and \(L_Y^{1/2}=\sqrt{9}=3\). Therefore \(Y=3(8)=24\), and \(Y/K=24/16=1.5\).
With \(A=9\), each design receives \(x_i=16/9\), not 4. Each contributes \(\sqrt{16/9}=4/3\), so all nine contribute \(9(4/3)=12\). Therefore \(Y=3(12)=36\), and \(Y/K=36/16=2.25\).
Output rises by \(36-24=12\), or \(12/24=0.5\), which is a 50% increase. Output per unit of intermediate capital also rises from 1.5 to 2.25. The check is that total intermediate input remains 16 in both cases: \(4(4)=16\) and \(9(16/9)=16\).
Research Labor Creates Designs, Not Current Final Output
The research equation describes a flow of new designs rather than the final-good production function. In continuous time, \(\dot A\) is the rate of change in the design stock. For example, if \(A=4\), \(L_A=4\), and \(\eta=0.05\), then \(\dot A=0.05(4)(4)=0.8\) design units per year. This is an instantaneous rate under the model, not a statement that exactly 0.8 indivisible products appear at the end of each year.
Because total labor is divided between research and final production, additional research labor has two distinct effects in this setup. It raises the rate of design creation, while reducing the labor currently available for final production if total labor is fixed. The research equation alone measures the first effect; the production function measures current output given the current design stock and allocations.
Worked Example: A Labor Shift Toward Research
An economy has total labor \(L=20\), intermediate capital \(K=16\), and \(A=4\) designs. Let \(\alpha=1/2\) and \(\eta=0.05\). Compare the current output and design-creation rate when \(L_A=4\) and when \(L_A=5\). Assume equal allocation of intermediate capital across designs.
Method. First use \(L_Y=L-L_A\) to find final-production labor. Then calculate current output from the production function and the innovation flow from the research equation. Keep the current design stock fixed for this comparison.
When \(L_A=4\), final-production labor is \(L_Y=20-4=16\). Each design receives \(16/4=4\) units of intermediate capital, so the sum of the square-root contributions is \(4\sqrt{4}=8\). Current output is \(Y=\sqrt{16}(8)=4(8)=32\). The design-creation rate is \(\dot A=0.05(4)(4)=0.8\) design units per year.
When \(L_A=5\), final-production labor is \(L_Y=20-5=15\). The design stock and capital allocation are unchanged, so the sum of intermediate contributions remains 8. Current output is \(Y=\sqrt{15}(8)\), approximately 30.984. The design-creation rate is \(\dot A=0.05(5)(4)=1.0\) design unit per year.
The one-worker shift reduces current output by approximately \(32-30.984=1.016\) while increasing the design-creation rate by \(1.0-0.8=0.2\) design units per year. These calculations compare current production and research flows; they do not determine whether the shift is worthwhile, since that would require assessing the value of future output and the cost of research.
What a Variety Comparison Does—and Does Not—Show
A variety comparison is meaningful only when the other inputs being compared are clear. Holding \(K\) fixed asks whether more designs make a given intermediate stock more productive. Holding input per design fixed asks a different question: adding designs then also requires adding intermediate capital. In the latter comparison, an output increase cannot be attributed to variety alone.
Worked Example: Separating Variety From More Capital
Let \(\alpha=1/2\), \(L_Y=4\), and suppose each design receives \(x_i=4\). Calculate output and output per unit of intermediate capital for \(A=4\) and \(A=9\). Explain whether the output difference isolates the variety effect.
Method. Apply the production function to the stated per-design input, then calculate total capital as \(K=Ax_i\). The key is not to hold total capital fixed when the question holds input per design fixed.
For \(A=4\), total capital is \(K=4(4)=16\). The sum of contributions is \(4\sqrt{4}=8\), and \(\sqrt{L_Y}=\sqrt{4}=2\), so \(Y=2(8)=16\). Thus \(Y/K=16/16=1\).
For \(A=9\), total capital is \(K=9(4)=36\). The sum of contributions is \(9\sqrt{4}=18\), so \(Y=2(18)=36\). Thus \(Y/K=36/36=1\).
Output rises from 16 to 36, but capital also rises from 16 to 36. Output per unit of capital is unchanged. The calculation therefore does not isolate the effect of variety; it compares economies with more designs and proportionally more intermediate input.
Common Mistake: Counting the Same Capital Twice
A frequent error is to increase \(A\) while keeping the input per design unchanged, then describe the resulting output change as a pure variety effect. In the first worked example, retaining \(x_i=4\) after moving from \(A=4\) to \(A=9\) would give the incorrect calculation \(Y=3(9\sqrt{4})=54\). But that allocation uses \(9(4)=36\) units of capital, not the fixed stock \(K=16\).
With \(K=16\), the correct input per design is \(16/9\), and output is \(3(9\sqrt{16/9})=36\). The incorrect calculation silently adds 20 units of intermediate capital. Always identify which quantity is held fixed: total capital \(K\), input per design \(x_i\), or neither.
Check Your Understanding
For each problem, use the stated production and research assumptions. In production calculations, assume equal allocation across designs unless the problem says otherwise.
- Let \(\alpha=1/2\), \(L_Y=25\), \(A=4\), and \(K=36\). Find the input per design, final output, and output per unit of intermediate capital. (Answer: \(x_i=9\), \(Y=60\), and \(Y/K=5/3\).)
- Let \(\alpha=1/2\), \(L_Y=16\), and \(K=36\). Compare output at \(A=4\) and \(A=9\), holding \(L_Y\) and \(K\) fixed. Give the percentage increase. (Answer: output rises from 48 to 72, a 50% increase.)
- Total labor is \(L=24\). If \(L_A=6\), \(A=10\), and \(\eta=0.01\), calculate \(L_Y\) and the design-creation rate. (Answer: \(L_Y=18\) and \(\dot A=0.6\) design units per year.)
- Suppose \(\alpha=1/2\), \(L_Y=16\), \(K=36\), and output is \(Y=48\) under symmetric allocation. Find the number of designs and the input per design. (Answer: \(A=4\) and \(x_i=9\).)
- With \(\alpha=1/2\), \(L_Y=9\), and \(x_i=1\) for every design, compare \(A=4\) with \(A=9\). Calculate output and capital in each case, and state whether output per unit of capital changes. (Answer: at \(A=4\), \(K=4\), \(Y=12\), and \(Y/K=3\); at \(A=9\), \(K=9\), \(Y=27\), and \(Y/K=3\). Output per unit of capital is unchanged.)