As a key symbol of the Fourth Industrial Revolution, the proliferation of industrial robots presents a fascinating duality. While their adoption is often viewed as a response to demographic pressures, my research investigates the less-explored reverse pathway: how the use of China robots actively shapes the demographic landscape itself. Utilizing data from the International Federation of Robotics (IFR), the China Family Panel Studies (CFPS), and national census microdata, this analysis delves into the impact of industrial automation on aging in China, unraveling the underlying mechanisms and deriving critical policy implications.
The central findings are threefold. First, the use of industrial robots significantly accelerates the aging trend. Second, this effect operates primarily through the channel of altered marital and fertility behaviors—lowering marriage and birth rates while increasing divorce rates. Third, the fundamental driver is the transformation of intra-household dynamics: China robots narrow the gender economic gap, elevating the economic status and decision-making power of women (wives), which in turn depresses traditional family formation and fertility, culminating in a more aged population structure.
Theoretical Framework: From Robots to Demographics
To logically structure this relationship, I develop a theoretical model linking robot adoption to aging through the intermediation of gender economics and family decisions.
1. Robot Adoption and the Gender Wage Gap
I begin with a production model featuring manual and cognitive tasks. Output \(Y\) requires four inputs: physical labor (\(L_A\)), mental labor (\(L_B\)), robot capital (\(R\)), and traditional non-robot capital (\(K\)). I assume robots substitute more easily for physical labor than for mental labor, and that men have a comparative advantage in physical skills while women excel in mental skills. The production function is specified as:
$$ Y = \left[ \beta (R + L_A)^{\gamma} + (1-\beta) L_B^{\gamma} \right]^{\frac{\alpha}{\gamma}} K^{1-\alpha} $$
where \(\beta \in (0,1)\) is the output weight of physical-task-centric production, \(\alpha\) is the output elasticity of the composite labor-robot bundle, and \(\gamma \in (-\infty, 1)\) governs substitutability between labor types (\(\gamma < 1\) implies substitution).
The wage rates for physical and mental labor are determined by their marginal products. The relative skill premium \(\pi\) of mental to physical labor is:
$$ \pi = \frac{\omega_B}{\omega_A} = \left( \frac{1-\beta}{\beta} \right) \left( \frac{R + L_A}{L_B} \right)^{1-\gamma} $$
Given \(\gamma < 1\), the derivative \(\partial \pi / \partial R > 0\). The premium for mental skills increases with robot capital. Assuming men supply more physical labor (\(L_A^m > L_A^f\)) while both genders supply equal mental labor (\(L_B^m = L_B^f\)), the gender wage gap \(G\) is defined as the ratio of male to female wages. The model yields:
$$ \frac{\partial G}{\partial R} = \frac{\partial G}{\partial \pi} \frac{\partial \pi}{\partial R} < 0 $$
Thus, the application of China robots acts to narrow the gender wage gap. This forms our first hypothesis: H1: The adoption of industrial robots reduces the gender wage gap.
2. Gender Wage Gap and Marriage/Fertility Decisions
Next, I model how a shrinking gender wage gap influences family formation. Following frameworks on household bargaining, consider a couple (i, j) deciding on marriage and time allocation between market work (\(1-t\)) and home production (\(t\)). The marital utility function is:
$$ U_{ij} = \max_{t_i, t_j} \left[ (1-t_i)\omega_i + (1-t_j)\omega_j + \alpha_i \log(n) + \beta (t_i + t_j) + q_{ij} \right] $$
where \(\omega\) denotes wage, \(n\) represents children, \(\alpha\) captures gender norms regarding the valuation of spousal consumption, and \(q_{ij}\) is the perceived quality of the match. Under traditional norms where men’s market advantage is pronounced (\(\omega_m > \omega_f\)), a reduction in the gender gap \(G = \omega_m / \omega_f\) increases the reservation quality threshold \(q^*\) for both men and women to enter marriage. Intuitively, as women’s relative earnings rise, the opportunity cost of marriage and specialization in home production increases, making marriage less attractive unless matched with a exceptionally high-quality partner. This leads to a lower equilibrium marriage rate and a higher propensity for divorce among existing marriages.
For fertility, consider a household maximizing a weighted sum of utilities subject to a joint budget and a household production function for children, \(b = (h_m + h_f)^2\), where \(h\) is time spent on childcare. The optimal number of children \(b^*\) can be expressed as a function of the gender wage gap \(G\):
$$ b^* = \frac{2}{1+\phi} \left( 1 + \frac{\phi}{G} \right) $$
where \(\phi\) is a parameter related to the weight of children in utility. The derivative is:
$$ \frac{\partial b^*}{\partial G} = \frac{2\phi}{(1+\phi)G^2} > 0 $$
This indicates that a smaller gender wage gap (lower \(G\)) leads to a lower optimal fertility rate \(b^*\). The rising opportunity cost of women’s time dominates any income effect. Combining this with H1, I arrive at the second hypothesis: H2: Industrial robot adoption reduces the gender wage gap, which in turn decreases marriage rates and fertility, while increasing divorce rates.
3. Robot Adoption and Aging: The Direct Demographic Link
Finally, I formalize the direct link between robots and an aging population structure. The workforce \(L\) comprises young workers (\(L_A\)) and old workers (\(L_O\)). Two sectors exist: a robot-using sector employing young workers and robots, and a non-robot sector employing only old workers. The robot-using sector’s production is:
$$ Y_R = \left[ R^{\frac{\sigma-1}{\sigma}} + L_A^{\frac{\sigma-1}{\sigma}} \right]^{\frac{\sigma}{\sigma-1}} $$
where \(\sigma > 0\) is the elasticity of substitution between robots and young labor. The non-robot sector is \(Y_N = L_O\). Aggregate output combines both sectors. Solving the cost-minimization problem and defining the robot exposure measure as \(ETR = R / L\), I derive the relationship between \(ETR\) and the share of old workers \(l_O = L_O / L\):
$$ ETR = \left( \frac{\gamma_2}{\gamma_1} \rho^{\sigma} \right)^{\frac{1}{\sigma-\epsilon}} l_O (1 – l_O)^{\frac{\epsilon-1}{\sigma-\epsilon}} $$
where \(\rho\) is the robot price, \(\epsilon\) is the elasticity of substitution between sectoral outputs, and \(\gamma\) are sector shares. Assuming \(\sigma > \epsilon\), it follows that \(\partial l_O / \partial ETR > 0\). This yields the third hypothesis: H3: Increased use of industrial robots exacerbates population aging.
Empirical Analysis and Key Results
To test these hypotheses, I construct a panel dataset at the regional (prefecture-city) level in China from 2006 to 2019. The core explanatory variable, Exposure to Robots (ETR), measures regional industrial robot density, calculated by combining IFR data on industry-level robot stocks with baseline regional employment structures from the 2000 census.
| Variable | Description | Mean | Std. Dev. |
|---|---|---|---|
| Aging | Population aged 65+ / Total Population (%) | 9.51 | 2.30 |
| ETR | Regional Industrial Robot Exposure | 1.37 | 2.05 |
| Marriage Rate | New Marriages / Total Population (‰) | 8.21 | 1.89 |
| Divorce Rate | Divorces / Total Population (‰) | 2.15 | 1.05 |
| Fertility Rate | Births / Women of Childbearing Age (%) | 38.12 | 8.77 |
| Wife’s Income Share | Wife’s Income / (Husband’s + Wife’s Income) | 0.42 | 0.28 |
| Wife’s Work Hour Share | Wife’s Work Hours / (Husband’s + Wife’s Hours) | 0.48 | 0.18 |
Baseline Impact on Aging
The baseline regression model is:
$$ Old_{ct} = \alpha_0 + \alpha_1 ETR_{ct} + \alpha_2 Z_{ct} + \epsilon_{ct} $$
where \(Old_{ct}\) is the aging indicator in region \(c\) at year \(t\), and \(Z_{ct}\) is a vector of controls (sex ratio, GDP, urbanization, education spending, etc.).
| Dependent Variable | (1) Aging | (2) Aging | (3) Econ. Dependency | (4) Econ. Dependency |
|---|---|---|---|---|
| ETR | 0.0836*** | 0.0198*** | 0.0936*** | 0.0223*** |
| (14.15) | (3.30) | (13.85) | (3.38) | |
| Controls | No | Yes | No | Yes |
| Region/Year FE | Yes | Yes | Yes | Yes |
| Observations | 2,252 | |||
| R-squared | 0.354 | 0.542 | 0.330 | 0.524 |
Note: *** p<0.01, ** p<0.05, * p<0.1; t-statistics in parentheses.
The results confirm H3. The adoption of China robots has a statistically significant positive effect on the aging level. To address potential reverse causality (aging regions adopting more robots) and other endogeneity concerns, I employ an instrumental variable (IV) approach, using the lagged industrial robot exposure in the United States as an instrument for China’s exposure. The first-stage F-statistic is strong (F=869.33), rejecting weak instrument concerns.
| Second Stage: Aging | First Stage: ETR | |||
|---|---|---|---|---|
| Variable | (1) | (2) | (3) | (4) |
| ETR | 0.1194*** | 0.0831*** | — | — |
| (20.21) | (6.11) | |||
| IV: US Robot Exposure | — | — | 0.1682*** | 0.1051*** |
| (29.48) | (11.45) | |||
| Controls | No | Yes | No | Yes |
| Observations | 2,252 | |||
The IV estimate is larger than the OLS estimate, suggesting OLS may underestimate the effect due to measurement error. Quantitatively, during the 2006-2019 sample period, the increase in the application of China robots explains approximately 26.97% of the observed increase in the aging ratio.
Unpacking the Mechanism: Marriage and Fertility
I now test the mechanism proposed in H2. Using household-level data from CFPS, I estimate models where the dependent variable is an indicator for new marriage, new divorce, or new birth in a family.
| Dependent Variable | New Marriage | New Divorce | New Birth | Wife’s Income Share | Wife’s Work Hour Share |
|---|---|---|---|---|---|
| ETR (Region) | -0.0025* | 0.0047*** | -0.0126*** | 0.0330*** | 0.0279*** |
| (-1.72) | (3.16) | (-3.48) | (2.84) | (4.91) | |
| Household Controls | Yes | Yes | Yes | Yes | Yes |
| Region/Year FE | Yes | Yes | Yes | Yes | Yes |
| Observations | 45,874 | 45,874 | 45,874 | 20,790 | 43,347 |
Note: Estimates from IV models using household-level CFPS data. t-statistics in parentheses.
The results are clear and consistent. Greater exposure to China robots at the regional level significantly reduces the probability of new marriages and new births within households, while increasing the probability of divorce. This confirms the first part of the mechanism chain. The effect is particularly pronounced among prime-age adults (35-45 years), the core group for family formation.
The Core Driver: Narrowing the Gender Economic Gap
The final link in the chain is verifying H1—that robots narrow the gender economic gap. The last two columns of Table 4 present compelling evidence. Regional robot exposure significantly increases the wife’s share of total household income and her share of total household work hours. Further analyses using broader female-specific metrics (e.g., female income share in the household) yield congruent results. The diffusion of China robots rebalances economic contributions within the household, enhancing women’s relative economic power.
This shift in bargaining power has profound implications. As women’s economic independence and opportunity cost of home time increase, the traditional gains from marriage centered on gender specialization diminish. Women demand higher-quality matches to enter marriage, and within marriages, fertility becomes less desirable relative to market participation. This logic directly leads to the observed “two lows and one high” pattern—low marriage, low fertility, high divorce—which is the proximate demographic driver of accelerated aging.
Conclusion and Policy Implications
This research establishes a critical feedback loop in China’s development trajectory: population aging incentivizes firms to adopt industrial robots, but the proliferation of these China robots, by transforming labor markets and intra-household economics, subsequently dampens marriage and fertility, thereby further intensifying the aging trend. The empirical evidence strongly supports this pathway, with robots explaining a substantial portion of aging dynamics.
The policy implication is stark but clear. Attempting to break this cycle by restricting technological adoption is neither feasible nor desirable, as robots are essential for productivity and competitiveness. The key leverage point lies in addressing the depressed fertility outcome. Policies must creatively offset the economic disincentives for childbearing that automation inadvertently strengthens. This could involve substantial financial subsidies for childbirth and child-rearing, significantly reducing educational costs, and promoting flexible work arrangements that support working parents. Furthermore, revitalizing cultural narratives that celebrate family and “the joy of children” could help counterbalance the purely economic calculus. The goal is not to reverse technological progress but to ensure societal structures adapt to harness its benefits while sustaining demographic resilience.