The gender gap in the ownership of promising land

Significance

Prior studies show that, in terms of investment rate of return (ROR), females outperform males in securities. Using millions of observations from the public administrative data of Taiwan, we identify a unique opposite pattern in real estate investment: males outperform females in land ROR. Through statistical analysis, we further show that this ROR difference cannot be explained by individual factors mentioned in the literature. We hypothesize that it is parental gender preferences which are behind the scene: males have higher land ROR because their parents help them acquire more “promising” land. We propose several tests to support our hypothesis. Our finding establishes the quantity-and-quality double disadvantage of real estate that impacts females in many societies with patriarchal traditions.

Abstract

Using millions of observations compiled from the public administrative data of Taiwan, we find a surprising gender inequity in terms of real estate: Men own more land than women, and the annual rate of return (ROR) of men’s land outperform women’s by almost 1% per year. The latter finding of gender-based ROR difference is in sharp contrast to prior evidence that women outperform men in security investment, and also suggests a quantity-and-quality double jeopardy in female land ownership which, given the heavy weight of real estate in individual wealth, has important implications for wealth inequality among men and women. Our statistical analyses suggest that such a gender-based difference in land ROR cannot be attributed to individual-level factors such as liquidity preferences, risk attitudes, investment experience, and behavioral biases, as described in the literature. Rather, we hypothesize parental gender bias—a phenomenon that is still prevalent today—to be the key macrolevel factor. To test our hypothesis, we partition our observations into two groups: an experimental group in which parents can exercise gender discretion, and a control group in which parents cannot exercise such discretion. Our empirical evidence shows that the gender difference with respect to land ROR only exists in the experimental group. For many societies with long-lasting patriarchal traditions, our analysis provides a perspective to help explain gender differences in wealth distribution and social mobility.

It is well documented that men own a larger proportion of real estate than women almost everywhere in the world, and this phenomenon is more pronounced in countries in which gender preferences are prevalent (1–3). Using millions of observations compiled from the public administrative data of Taiwan (the datasets and compiling methods are explained in SI Appendix, section A), we find, not surprisingly, the same pattern in Taiwan, a society with a typical patriarchal tradition (4, 5): Of note, 63% of land is owned by men and only 37% is owned by women. The same datasets, however, also uncover an unexpected finding: The annual rate of return (ROR) on land investment for men exceeds that of women by 1%.^* This finding is in sharp contrast to findings in prior research that shows that women’s ROR on stock investment is about 1% higher than men’s ROR (6), which we also find in our data. Our paper addresses the following questions: why is Taiwan’s ROR difference in land between men and women so different from that of stocks? Can previous gender-difference explanations with respect to financial investment tools be applied? Further, can we identify statistical tests that establish a plausible hypothesis for this phenomenon?

Previous research on differences in security investment linked to gender have raised several possible explanations: biological conditions (7), psychological variations (8), risk attitudes (9), and investment knowledge (10). In our data, if land ROR differences between men and women can be attributed to any of these factors [e.g., females with low levels of testosterone (7)], then we arrive at a conclusion that females’ real estate investment decisions are determined by that factor. Using various statistical methods, however, we find that none of the previously mentioned individual-level factors suffice to provide a coherent explanation for the land ROR difference that we observe. Rather, the most important factor that explains such a difference is parental gender bias; that is, parents tend to help their sons acquire more “promising” land. This implies that, in terms of real estate, males in a more traditional society that shows preferences for sons benefit both quantitatively and qualitatively, which suggests a compounding double inequality with respect to female wealth. As a result, Taiwanese men’s privilege appears to invoke the sentiment in Exodus Song that “though I am just a man, when you are by my side, with the help of parents, I know I can be strong;” that said, the same cannot be sung by women in Taiwan.

Our investigation has implications for modern societies all over the world: The large differences in wealth by gender remain a prevailing phenomenon, even though the equal-right inheritance laws and the portio legitima rule have been legislated for long. In this research, we present a mechanism for parents to exercise their gender preferences: The differential quality of real estate in parents’ inter vivos transfers to children.

The remainder of this paper is arranged as follows. We first discuss how our study contributes to the literature. Next, we describe our data sources and summary statistics for wealth and ROR on land and stocks for both men and women. Finally, we discuss our empirical design to rule out other explanations and to explore how parental gender bias explains the intriguingly higher land ROR to men. We leave most technical details to our SI Appendix.

Literature

As an extensive literature on differential treatments of men and women exists, we here only review briefly the papers most related to our analysis, leaving more detailed references in SI Appendix, section B. Prior studies have examined parents’ differential treatments of sons and daughters with respect to pregnancy, nutrition, education, and inheritance (2, 5, 11–13). In addition to long-term civilizational differences, gender inequality may be driven by sociohistorical processes and political mobilizations. Sometimes such differential treatments must be compromised, because modernization often facilitates the enactment of equal-right laws (14, 15). Our study explores a certain angle: How wealthy parents can use real estate transactions to transfer their wealth to sons before their deaths, so they may avoid the rule of portio legitima in inheritance laws and sustain their preferences for sons.

Regarding gender differences with respect to investment performance, previous studies emphasize individual-level explanatory factors, such as biological conditions, psychological variations, and risk attitudes (7, 8, 16). In our paper, we propose that a culture-based factor—parental gender preference—explains the phenomena that we observe. On the other hand, the literature on discrimination in housing and mortgage markets focuses mostly on racial and ethnic discrimination (17–19) and leaves gender discrimination unexplored (1). This gap in the literature reveals a noteworthy difference with regard to social background: Racial and ethnic backgrounds receive much more attention in the United States, but these differences are not considered as important in East Asian countries.

More broadly, our study fills a gap in the literature on gender differences with respect to wealth allocation (20–24) by investigating inequality in real estate transactions. Because real estate accounts for more than 50% of individual wealth in Taiwan (and worldwide), the female double disadvantage in real estate certainly exacerbates wealth inequality linked to gender.

Finally, while the adverse effect of discrimination on social mobility and equality has been extensively discussed in the literature (25, 26), how to quantify such an effect in a society is challenging due to the lack of large-scale data on wealth and income. Our use of millions of administrative observations allows us to precisely measure wealth disparity between men and women, and hence our empirical evidence has implications for both tax policies and social welfare.

Materials and Methods

Data.

We compile the following publicly accessible government datasets to implement a comprehensive analysis of the wealth composition and transfers in Taiwan: a) the Family Member database, which includes anonymized individuals’ date of birth, marriage status, and information about parents;^† b) the Land Value Increment Tax database, which includes details of all taxable land transfers; c) the Personal Income Tax database, which includes all taxable salary, bonuses, dividends, interests, and other personal income sources; and d) the Nationwide Personal Property database, which includes information about individuals’ savings/deposits, stocks, housing, land, and car ownership.^‡ We provide the details of our data compiling, variable definitions, and summary statistics in SI Appendix, section A.

By combining these governmental databases, we obtain the wealth data of 10,566,455 individuals from 2005 to 2015, which accounts for 68.18% of the total population (2005 data); the remaining 31.82%, including children and elderly, do not have taxable wealth or income records. We focus on 3,533,780 individuals in the age cohort from 35 to 55 who have at least one living parent in 2005. Individuals in this group are not only in their prime earning age and mature enough to trade and manage wealth, but also have parents who are old enough to prepare wealth transfers for their descendants. Land accounts for over 55% of total wealth in our observations, indicating the dominant role that land plays with respect to wealth and asset allocation. Given the fact that land value weighs significantly in real estate in Taiwan (which accounts for 79% of real estate value in our data), our measures of land value and ROR appropriately reflect individuals’ real estate wealth and payoffs. In addition, 70.23% of households owned their real estate in 2005. This is consistent with common beliefs of the elderly in Chinese societies that land is the only “real” asset (27). Moreover, wealthy parents tend to invest a large portion of their wealth in real estate, as shown in SI Appendix, section A.

In Fig. 1, we present the accumulated value of land acquired by men and women in 2005. We find that men acquired significantly more land (63% in total value) in 2005 than women did; more importantly, the gap between men’s and women’s land value widened over the years, suggesting an increasing disparity in real estate wealth across genders.

Fig. 1.

The Value of Land Acquired in 2005. In this figure, we present the time-series patterns of the total values of land acquired by men and women (whose ages are between 35 and 55 and have at least one living parent) in 2005. The vertical axis denotes the total assessed values of the land from 2005 to 2015 (in millions NTD).

The RORs on Land and Stocks.

To examine gender differences in investment performance, we calculate each individual’s annual ROR on land and stocks, respectively (SI Appendix, section C). We calculate each individual’s ROR on a piece of land in one’s holding period (i.e., from the acquiring year to the selling year), and then calculate one’s land ROR as the weighted average ROR for all the land that one held in each year. Similarly, we calculate each individual’s ROR on a stock in one’s holding period, and then calculate one’s stock ROR as the weighted average ROR for all stocks that one held in each year.

Fig. 2A illustrates gender differences in land ROR by presenting average men’s and women’s land ROR (in solid and dashed lines, respectively) for each of the 50th to 99th percentiles ranked by parental wealth. We find a clear pattern that men’s ROR is consistently higher than that of women, especially whose parents are above the 95th percentile. More interestingly, the gap in ROR widens with wealth: It is as high as 1.8% and 2.1% per year in the 98th and 99th percentiles, respectively. Such a gap, when combined with the fact that men own more land, results in a twofold disadvantage with respect to women’s real estate wealth.

Fig. 2.

(A) Land ROR for Men and Women. In this figure, we present the average ROR of all men and women by parents’ wealth percentile. The ROR is based on individual ROR ( $R_i$ ), as defined in SI Appendix, section C. We focus on individuals whose ages are between 35 and 55 and have at least one living parent in 2005. The horizontal axis denotes the percentile of parental wealth, and the vertical axis denotes the average ROR of men and women (in solid and dashed lines, respectively) for each wealth percentile. (B) Stock ROR for Men and Women. In this figure, we present the average stock ROR of all men and women by parents’ wealth percentile. The individual stock ROR is defined in SI Appendix, section C. The horizontal axis denotes the percentile in parental wealth, and the vertical axis denotes the average ROR of men and women (in solid and dashed lines, respectively) for each wealth percentile.

In Fig. 2B, we present average men’s and women’s stock ROR (in solid and dashed lines, respectively) in each of the 50th to 99th percentiles ranked by parental wealth. We find a clear pattern for individuals in all percentiles: Men’s RORs are consistently lower than those of women’s. Our finding of women’s advantage in stock investment thus provides evidence from Asia to support the literature.

One may wonder if women’s advantage in stocks can offset their disadvantage in real estate. Using the 99th percentile group as an example, we find that the profits from land over our sample period are as high as NTD 42 million among men and NTD 9 million among women. These numbers amount to 79 and 17 times the average annual income per capita in Taiwan (NTD 533,752 in our sample period). The difference in real estate profits is in sharp contrast to and exceeds women’s advantage in stock profits, thus calling for further investigation.

Our question is why do Taiwanese men outperform women with respect to real estate investment, even though their stock investment shows a reverse pattern? We first examine whether previous explanations suffice to explain the differences that we observe. Given that these explanations fail, we propose a test to help identify the possible explanations.

Individual-Specific Explanations from Prior Studies.

Prior studies have proposed several individual-specific factors to explain gender differences in investment approaches. Women’s land ROR may be lower than men’s because i) women invest in more liquid or less risky real estate as they are more risk averse (9, 16), ii) women’s knowledge and experience in investment is lower than men’s (10, 28), and iii) women are more pessimistic and are hence more willing to sell land at lower prices during economic downturns (29–31).

Taking advantage of the large number of observations in our administrative data, we identify several variables that reflect the explanatory factors in the literature. We partition our observations into groups based on one of these variables, and examine men’s and women’s land ROR (and the gender difference) in each group. If a factor can explain men’s advantage in land ROR, then we should observe a correlation between cross-group gender differences in ROR and the variable that reflects the specific factor. If we cannot find such a correlation, then the factor fails to explain the ROR difference.

For example, one’s risk attitude can be identified by the proportion of risky assets in one’s total wealth. For both men and women, we can partition individuals into two groups: One is more risk-taking and the other is less so. We can then test whether the “risk attitude” factor helps explain ROR differences between men and women.

We use one’s land holding period length, cash-to-wealth ratio, and the volatility of past land ROR to measure one’s liquidity preference and risk aversion. We also use one’s past land trading frequencies to measure one’s knowledge and experience, as well as use one’s timing in selling land to measure one’s optimism and loss aversion. We present our detailed analyses in SI Appendix, section D. Our results suggest that none of these factors can satisfactorily explain the gender gap in land ROR.

Testing Method.

Having found that aforementioned microlevel factors cannot explain land ROR differences linked to gender, we suspect that such a difference may be due to a macrolevel factor: parents’ preferences for sons. Specifically, men’s land may have higher ROR simply because their parents help them acquire more promising land. How do we test this hypothesis?

We note the following two facts. First, there are four major types of land acquisitions by the young generation: 1) inter vivos transfers from their parents, 2) parent-supported purchases, 3) unsupported purchases, and 4) bequests.^§ Only the first two acquisition types are related to parental discretion; bequest allocations among children in category 4) are subject to inheritance laws and the portio legitima rule, which limits parental discretion. Also, category 3) by definition does not include any role played by parents. Thus, if the gender difference in land ROR is related to parental discretion, then we expect it to appear only in types 1) and 2), but not in 3) and 4).

Second, if the ROR difference linked to gender is due to parental differential treatments, then this discretion can happen only in families that have both girls and boys. In contrast, for families that either have all girls or all boys or that have only one child, there is no way for parents to exercise differential treatments based on gender. Thus, if we partition our observations according to the composition of a family’s children, we then expect to observe a more significant gender-based ROR difference in the families with sibship types that parents have the room of gender discrimination, but also expect to observe an insignificant difference in families with alternative types of siblings.

We note that the above-mentioned partition of land acquisition types and types of siblings in a family do not have any a priori relationship with real estate value appreciation. Thus, this observation partition very much serves as a random experiment, in which certain types of land-acquisition/siblings can be viewed as an “experimental group,” and the other types are treated as the “controlled group.” When we control for other independent variables, this experiment gives us a test statistic that helps us identify whether gender preferences indeed offer a sufficient explanation of the ROR difference.

Results

In our regression analysis, we focus on individuals with parents in the top 5% of parents’ wealth ranks because i) as we observe from Fig. 2A, this is the wealth range in which land ROR differences arise; and ii) only individuals in the top 5% wealth rank have much higher land-trading frequencies and much shorter holding periods (i.e., the lower 95% own real estate mainly for residential purposes; see SI Appendix, Figs SI.2 and SI.5). We estimate the following equations using ordinary least squares regressions:^¶

$$\begin{array}{c}{R}_{i\left(f\right)jt}^{}=\alpha +{\beta }_1{\mathit{Male}}_{i\left(f\right)}+\mathit{\beta }{\mathit{Controls}}_{i\left(f\right)jt}+ \sum {Year FE}_t\hfill \\ +\sum {Family FE}_f+{\epsilon }_{i\left(f\right)jt}\hfill \end{array}$$ [1]

in which $R_{i(f)jt}^{}$ denotes the annual ROR of land j sold in year t by individual i from family f. ${\mathit{Male}}_{i(f)}$ is an indicator variable that equals one if individual i is male, and zero otherwise. ${\mathit{Controls}}_{i(f)jt}$ denotes a set of control variables including fixed effects for birth order, the owner’s age, the owner’s marital status, as well as indicator variables for the owner’s city of residence and city of birth.^# $\sum {Year FE}_t$ denotes the fixed effects for selling years to control for real estate cycles, and $\sum {Family FE}_f$ denotes the fixed effects for families to help us absorb family heterogeneity and to enable us to estimate the “within-family” effect in gender differences. We cluster our SEs by land location and transaction year to correct for estimation errors.

We first estimate (Eq. 1) for all transactions to assess whether there is any difference between men’s and women’s land ROR. The results are reported in Panel A of Table 1. Model 1 without control variables shows the estimated coefficient on Male to be 1.05% (SE 0.52%), and Model 2 that includes ${\mathit{Controls}}_{i(f)jt}$ in the regression presents the estimated coefficient on Male to be 0.98% (SE 0.47%). Moreover, the P-values of Male are 0.042 and 0.036 in Models 1 and 2, suggesting that the likelihoods for Male to be irrelevant in regressions are only 4.2% and 3.6%, respectively. These estimates suggest that, on average, men’s ROR is about 1% per year higher than women’s in Taiwan, which is substantial given that the mean ROR ranges between 4% and 11% (SI Appendix, Table SI.1 Panel C). In addition, we find consistent results when we further include the coefficient of variation of ROR of all land in the same area in regressions to control for risk premium in land trading (SI Appendix, section D6).

Table 1.

Tests based on types of land acquisition

	β (%)	St. errors (%)	P-value	Observations
Panel A: All transactions
Model 1
Male	1.05	0.52	0.042	260,471
Model 2
Male	0.98	0.47	0.036	260,471
Age-at-Start	0.31	0.17	0.062
Age-at-Sell	−0.22	0.20	0.255
Marriage-at-Start	−0.04	0.90	0.961
Marriage-at-Sell	1.12	0.87	0.201
Out-of-Taipei (Addr)	0.67	0.49	0.168
Out-of-Taipei (Born)	2.64	1.07	0.014
Panel B: All transactions subject to parental discretion
Model 1
Male	1.53	0.89	0.087	59,313
Model 2
Male	1.84	1.06	0.082	59,312
Age-at-Start	0.02	0.20	0.930
Age-at-Sell	0.17	0.30	0.567
Marriage-at-Start	−3.32	1.07	0.002
Marriage-at-Sell	3.24	1.03	0.002
Out-of-Taipei (Addr)	1.23	0.93	0185
Out-of-Taipei (Born)	2.28	1.73	0.187
Panel C: All transactions not subject to parental discretion
Model 1
Male	0.78	0.62	0.204	198,365
Model 2
Male	0.66	0.55	0.234	198,365
Age-at-Start	0.27	0.19	0.159
Age-at-Sell	−0.19	0.22	0.368
Marriage-at-Start	1.00	1.38	0.471
Marriage-at-Sell	0.31	1.37	0.819
Out-of-Taipei (Addr)	0.41	0.63	0.522
Out-of-Taipei (Born)	2.83	1.49	0.057

We estimate the following equations using least squares regressions: $R_{i\left(f\right)jt}=\alpha +{\beta }_{1}Mal{e}_{i\left(f\right)}+\beta Control{s}_{i\left(f\right)jt}+\sum YearF{E}_t+\sum FamilyF{E}_f+{\epsilon }_{i\left(f\right)jt}$  in which $R_{i(f)jt}$ denotes the annual ROR of land j sold in year t by individual i from family f. ${\mathit{Male}}_{i(f)}$ denotes an indicator variable that equals one if individual i is male, and zero otherwise. $\sum {Year FE}_t$ denotes the fixed effects for years to control for real estate cycles, and $\sum {Family FE}_f$ denotes the fixed effects for families. In Model 1 of each panel, we report the estimation results without control variables. ${\mathit{Controls}}_{i(f)jt}$ denotes a set of control variables including owners’ age, marital status, dummies for birth order, and indicator variables for the owner’s city of residence and city of birth. Age-at-Start is the owner’s age upon acquiring the land, and Age-at-Sell is the owner’s age upon selling the land. Marriage-at-Start and Marriage-at-Sell are indicator variables that equal one if the landowner is married, and zero otherwise. Out-of-Taipei (Addr) equals one if the landowner resides outside of Taipei city, and zero otherwise. Out-of-Taipei (Born) equals one if the landowner is born outside of Taipei city, and zero otherwise. Panel A includes all transactions, Panel B includes transactions subject to parental discretion, and Panel C includes transactions not subject to parental discretion.

To implement our first test, we estimate (Eq. 1) for the group with parental discretion and the group without, and present our estimation results in Panels B and C of Table 1, respectively. In Model 1 of Panel B, we find that the coefficient on Male is 1.53% (SE 0.89%) in the group subject to parental discretion, suggesting that men’s ROR is higher than women’s by more than 1.5% per year in land that they acquired with parental help. We obtain a similar finding in Model 2, in which we include ${\mathit{Controls}}_{i(f)jt}$ in our regressions. The P-values of Male are 0.087 and 0.082 in Models 1 and 2, suggesting that the likelihoods for Male to be irrelevant in regressions are only 8.7% and 8.2%, respectively. For the group without parental discretion in Panel C, the coefficients on Male are 0.78% (SE 0.62%) and 0.66% (SE 0.55%) in Models 1 and 2, respectively. Their high P-values suggest that these coefficients are not statistically significant.

We also implement interaction regression in which to test if the cross-group difference in the coefficients on Male is statistically significant. For Table 1 in which we separate the group of land acquired with parental discretion and the group of all others, we implement the interaction regressions by pooling two groups together and interacting Male with an indicator variable that is one for the group subject to parental discretion and zero otherwise.^|| We find a significant coefficient on the interaction term (2.55% with SE 0.84%) in SI Appendix, section G. These results support that parental discretion benefits men with respect to real estate acquisition, as wealthy parents give more profitable real estate to sons and give less profitable real estate to daughters.

To implement our second test, we further partition our observations in the group subject to parental discretion into two subgroups: the mixed-sex sibling subgroup and the same-sex sibling subgroup. The former includes land sellers who have at least one different-gender sibling, and the latter includes land sellers who are single children and whose siblings are all of the same gender. We then estimate Eq. 1 within each subgroup and report our estimation results in Panels A and B of Table 2, respectively. Since we cannot control for family fixed effects in this setting (due to collinearity), we add additional control variables, including Out-of-Taipei (parents), Parent’s wealth (in log term), Parents’ wealth in lands, Parent’s wealth in stock, and number of siblings to mitigate the influence of other family characteristics. In Models 1 and 2 of Panel A for the mixed-sex sibling subgroup, we find that the coefficients on Male are 1.60% (SE 0.61%) and 2.11% (SE 0.82%), respectively, both of which are statistically significant because their P-values are 0.008 and 0.011, respectively. On the other hand, for the same-sex sibling subgroup in Panel B, the coefficients on Male are insignificant [1.33% (SE 0.83% and P-value 0.110) and 1.06% (SE 0.80% and P-value 0.186)].^** Our analyses provide evidence that men’s land ROR may be higher than women’s when they have sisters, which further supports the role that parental discretion plays in explaining men outperforming women with respect to real estate in Taiwan.

Table 2.

Tests based on sibling composition

	β (%)	St. errors (%)	P-value (%)	Observations
Panel A: Mixed-sex sibling subgroup
Model 1
Male	1.60	0.61	0.008	53,178
Model 2
Male	2.11	0.82	0.011	53,177
Age-at-Start	0.13	0.12	0.280
Age-at-Sell	−0.13	0.14	0.338
Marriage-at-Start	−1.36	1.34	0.308
Marriage-at-Sell	0.08	1.42	0.958
Out-of-Taipei (Addr)	0.50	0.67	0.452
Out-of-Taipei (Born)	0.17	0.81	0.833
Parent’s wealth in stock	−4.48	1.91	0.019
Parent’s wealth in lands	1.23	2.20	0.576
Parents’ wealth	−0.31	0.34	0.355
Out-of-Taipei (parents)	1.42	0.97	0.142
Number of siblings	1.11	0.38	0.003
Panel B: Same-sex sibling subgroup
Model 1
Male	1.33	0.83	0.110	9,619
Model 2
Male	1.06	0.80	0.186	9,619
Age-at-Start	0.35	0.24	0.146
Age-at-Sell	−0.52	0.21	0.012
Marriage-at-Start	0.21	1.30	0.871
Marriage-at-Sell	0.44	1.71	0.796
Out-of-Taipei (Addr)	−0.67	0.91	0.461
Out-of-Taipei (Born)	0.40	1.12	0.724
Parent’s wealth in stock	−4.90	2.23	0.028
Parent’s wealth in lands	−0.36	1.82	0.845
Parents’ wealth	−0.12	0.68	0.863
Out-of-Taipei (parents)	0.70	1.08	0.518
Number of siblings	0.03	0.40	0.945

We focus on all lands that were acquired through parent-supported purchases plus gifts. We split our observations into two groups based on the structure of siblings within a particular family. The same-sex sibling subgroup includes individuals from families of all sons, all daughters, or single children, and the mixed-sex sibling subgroup includes individuals who have at least one sibling of different gender. We estimate the following equations using least squares regressions: $\begin{array}{c}{R}_{i\left(f\right)jt}^{}=\mathrm{ \alpha }+{\beta }_1{\mathit{Male}}_{i\left(f\right)}+\mathit{\beta }{\mathit{Controls}}_{i\left(f\right)jt}+\sum Year F{E}_t+{\mathrm{ \epsilon }}_{i\left(f\right)jt}\end{array}$  in which $R_{i(f)jt}^{}$denotes the annual ROR of land j sold in year t by individual i from family f.${\mathit{Male}}_{i(f)}$ denotes an indicator variable that equals one if individual i is male, and zero otherwise. $\sum {Year FE}_t$ denotes the fixed effects for years to control for real estate cycles. ${\mathit{Controls}}_{i(f)jt}$ denotes a set of control variables including an owner’s age, marital status, dummies for birth order, indicator variables for the owner’s city of residence and city of birth, and variables related to the owner’s family characteristics. Age-at-Start is the owner’s age upon acquiring the land, and Age-at-Sell is the owner’s age upon selling the land. Marriage-at-Start and Marriage-at-Sell are indicator variables that equal one if the landowner is married, and zero otherwise. Out-of-Taipei (Addr) equals one if the landowner resides outside of Taipei city, and zero otherwise. Out-of-Taipei (Born) equals one if the landowner is born outside of Taipei city, and zero otherwise. Variables capturing the owner’s family characteristics are as follows. Parents’ wealth in stock is the ratio of parental stock wealth to their total wealth. Parents’ wealth in lands is the ratio of parental land wealth to their total wealth. Parent’s wealth is expressed in log term. Out-of-Taipei (parents) is an indicator variable that equals one when the individual’s father was born outside Taipei City, and zero otherwise. Number of siblings is the number of siblings of the landowner.

Moreover, we examine if the effect of parental discretion varies by parents’ wealth in real estate and fathers’ city of birth; we report these results in SI Appendix, section E. We find that men’s advantage in land ROR tends to be more pronounced among individuals whose parents have a greater percentage of their wealth invested in real estate. This is intuitive because when parents have more wealth in real estate, such parents not only are more experienced with real estate transactions, but also own more real estate that they may use to facilitate their wealth transfer to their sons. On the other hand, we find that men’s advantage in land ROR tends to be more pronounced among individuals whose fathers were not born in Taipei City (i.e., parents who are more likely to insist on patriarchal traditions).

Discussion

It is well documented worldwide that a majority of real estate is owned by men. Against this gender-based difference in terms of quantity, we examine a possible difference in terms of quality that is often neglected in the literature. To do so, we combine several publicly accessible government databases to build a comprehensive dataset on individual wealth in Taiwan. We first show that men own 63% of land, which explains a great portion of gender disparity with respect to individual wealth. We further find that men’s ROR from land appreciation significantly outperforms women’s—by about 1% per year. Thus, women are confronted by “double jeopardy” in that they not only own less real estate, but less profitable real estate. These findings contrast with women’s advantage in security investment ROR and cannot be explained by several factors proposed in the literature.

In our empirical analysis, our test results provide evidence that men’s higher returns might be higher for those owners whose land is parents’ gifts or purchased with parents’ financial support, but not for other groups (i.e., those in which parents have no room to exercise their discretion). We then show that men’s higher ROR gained from land is more pronounced for the group of land owners with siblings of different genders, but not for other groups (i.e., those with parents have no objectives for exercising their gender preferences). Our empirical evidence is consistent with our hypothesis regarding discriminatory wealth transfers.

This study offers important implications with respect to social and wealth inequities (32, 33). Because real estate value represents more than half of individual wealth, women in patriarchal societies often face a twofold disadvantage (both quantitative and qualitative) with respect to wealth. From a broader perspective, we highlight how parents use inter vivos transfers of undervalued assets to children, which may explain why gender gap in wealth remains in modern economies despite the existence of equal-right inheritance laws.

Supplementary Material

Appendix 01 (PDF)

Dataset S01 (PDF)

Data, Materials, and Software Availability

Some study data available [Our empirical analyses are based on publicly accessible income/property tax datasets from the Ministry of Finance (MOF), Taiwan. These datasets and labs are indeed open to the public, as long as the operation complies with the administrative rules of the Fiscal Information Agency of MOF. We have provided all codes with detailed explanations in this submission].