Assessing the Stringency of Land-Use Regulation in a High-Density City: Evidence from Land Values and FAR Constraints in Guangzhou, 2003–2025

Comments 1: [For the sake of transparency, I would provide all data sources in an additional column of Table 1. ]

Response 1: [Thank you for this helpful suggestion. We agree and have revised the manuscript accordingly. Specifically, we have added a table note labeled “Data Source” to Table 1 to clearly indicate the origin of each variable, thereby improving transparency and reproducibility. This revision can be found on page 7, Table 1 of the revised manuscript.]

“[Revised text (Table note):All variables reported in Table 1 are derived from the China Land Market Network, which provides official transaction-level information on land parcels, including prices, parcel characteristics, FAR limits, and conveyance methods. Variables were constructed by the authors.]”

Comments 2: [In addition to straight-line distance measure, analyses often use travel time or actual travel distance. This is because barriers like rivers, railroads, or hills can make some neighborhoods less accessible and less connected to the city center.]

Response 2:
[Thank you for this suggestion. We agree that straight-line distance may not fully capture accessibility. To address this concern, we conducted a robustness check using road-network distance, and the results remain qualitatively unchanged. This stability is consistent with the spatial structure of Guangzhou, where urban development has expanded in parallel with major river corridors, while mountain ranges lie to the north rather than across the urbanized area. In addition, railway corridors run parallel to highways rather than forming transverse barriers. As a result, natural or infrastructural obstacles rarely disrupt intra-urban accessibility, suggesting that Euclidean distance provides a reasonable approximation of spatial separation. We have revised the manuscript to clarify this point; the discussion appears on Page 14, Lines 508–510]: [“We also test alternative distance measures using road-network distance as a robustness check, and the main results remain unchanged.”]

Comments 3:[In Table 1, I would present transaction price in million or even billion CNY to make it readable. In addition, minimum and maximum values can be informative. Given that the number of observations is constant across variables, the corresponding column can be dropped and the information on the number of observations place in a note to the table.]

Response 3:

[Thank you for this helpful suggestion. We have revised Table 1 by: (i) presenting the transaction price in billion CNY to improve readability; (ii) adding minimum and maximum values for each variable; and (iii) moving the number of observations (552) to a footnote. These changes can be found on page 7, Table 1 of the revised manuscript.]

Comments 4:[Figure 1 must be mentioned and interpreted in the text. The content of each panel must be explained. For instance, it is not clear to me what “Far” means. If it is a distance to the city center, why then points that are close to it are dark red?]

Response 4:

[Thank you for this helpful suggestion. We have revised Figure 1 and the interpenetration can be found on Page 6,Lines 241-249: Figure 1 presents the spatial distribution of the main variables used in the analysis. Panel (A) shows the geographic locations of the matched land transaction sample. Panel (B) indicates the locations of the city center (GTT) and district-level subcenters, which serve as reference points for distance-based measures. Panel (C) illustrates the spatial distribution of maximum allowable floor area ratio (FAR), where darker colors represent higher permitted development intensity. Panel (D) displays land transaction unit prices, with deeper shades indicating higher land values. Together, these panels provide a visual overview of the spatial structure of the sample and the key variables used in the empirical analysis.]

Comments 5: [Please, provide a thorough interpretation of your key indicator defined in equation (4). What does a larger or smaller value of it implies in terms of regulation stringency. Why stringency of 60–67% means that Guangzhou displays “below-average FAR stringency among Chinese cities during the 2010s”.]

Response 5: Thank you for the comment. We revised the manuscript to clarify the interpretation of the key indicator in Equation (4) and the meaning of the 60–67% result.

[In our framework, corresponds to statutory FAR, while y(M∗)represents the counterfactual free-market FAR. Equation (4) yields the ratio , which measures regulated FAR as a share of the market level. A smaller ratio indicates stronger Far stringency, as statutory FAR is more compressed relative to the market benchmark. A value closer to one implies weaker binding.

Our “below-average” characterization is based on comparing Guangzhou’s estimated FAR coefficient (0.575) with the cross-city national average (0.7331) reported by Brueckner et al. Within the same framework, a lower coefficient implies a higher  ratio and thus weaker binding. The implied range of 0.60–0.67 therefore indicates a moderate and relatively less restrictive level of FAR stringency compared with the cross-city average during the 2010s.]

These revisions have been incorporated in the revised manuscript as follows:

[Page 10, lines 352–358:As noted above, y(M) represents the floor area ratio (FAR), since it measures floor space per unit of land. Accordingly, Equation (4) can be interpreted as the ratio of regulated FAR to the counterfactual free-market FAR. A lower value of this ratio implies stronger regulatory stringency, as permitted density is more substantially constrained relative to the market-determined level. Conversely, a value closer to one indicates weaker regulatory binding, with statutory FAR approaching the unconstrained equilibrium outcome.]

[Page 12, lines 435–439:Within the theoretical framework, a larger coefficient indicates stronger regulatory binding, as land values respond more strongly to changes in FAR when regulatory constraints are tighter. Therefore, Guangzhou’s lower coefficient suggests relatively weaker regulatory stringency compared with the cross-city national average during the sample period.

Given that the sample period in Brueckner et al. [6] covers 2002–2012, while the present study uses data from 2003–2025, this consistency echoes their conclusion that, despite the government’s explicit policy objective of controlling growth in mega-cities, the degree of FAR stringency in Guangzhou has remained broadly similar over time. 

Page 12, lines 449–454:These values imply that the statutory FAR in Guangzhou is approximately 60%–67% of the counterfactual free-market FAR. Because smaller ratios correspond to stronger regulatory compression, this range indicates a moderate degree of FAR stringency rather than a highly restrictive regime. Consistent with the coefficient comparison above, this implied range suggests that FAR regulation in Guangzhou is less restrictive than the cross-city average reported for major Chinese cities.]

Comments 6:[It would be enough to report in all tables no more than three digits after decimal sign.]

Response 6:Thank you for this suggestion. We have revised all tables to report coefficients and standard errors with no more than three decimal places for clarity and consistency.

Comments 7:[It would be very convenient to have all possible values of stringency measure, Y (M

) /Y (M∗) corresponding to different model specifications in one summarizing table.]

Response 7: We agree that summarizing the implied stringency across specifications would be informative. [However, the theoretical mapping in Equation (4) is formally defined for 0<α<1. Several extended specifications yield coefficients exceeding unity, which cannot be directly translated into the structural ratio under this mapping. Nevertheless, these estimates remain economically meaningful, as they continue to capture the relative strength of FAR capitalization and the intensity of regulatory binding in a reduced-form sense across specifications.]

 [This revision can be found on page 10, lines 359–362:The mapping in Equation (4) is defined for 0<α<1, which characterizes the parameter range consistent with the underlying theoretical framework. This formulation provides an operational measure of FAR stringency by quantifying the extent to which planning regulation compresses market-driven development intensity.]

Comments 8：[You can find a quite comprehensive overview of the effects of land-use regulations in Kholodilin (2025).]

Response 8:Thank you for the suggestion.[ We have added a citation in Section 2.1 to situate our discussion within the broader literature on land-use regulation. The meta-study finds that regulatory constraints are often associated with higher housing prices and weaker supply responses, while emphasizing heterogeneity across contexts. We reference this work to position our analysis of FAR regulation within the wider evidence, focusing on measuring regulatory stringency at a fine intra-urban scale in a high-density city.]

[This revision appears on page 3, lines 102–104 of the revised manuscript: Kholodilin [29], in a meta-study of land-use regulation, finds that regulatory constraints are often associated with higher housing prices and weaker supply responses, while emphasizing substantial heterogeneity across institutional contexts.]

Comments 9：[Please, check for typos that are quite multiple. Section titles must start with capital letters.]

Response 9: Thank you for this comment. We have carefully proofread the manuscript to correct typographical errors and revised all section titles to ensure proper capitalization.

4. Response to Comments on the Quality of English Language

Response: We thank the reviewer for this comment. The manuscript has been carefully proofread and edited to improve clarity, grammar, and overall readability.

5. Additional clarifications

Comments 1: [1. Identification and Endogeneity Concerns

The key identification strategy relies on cross-sectional variation in statutory FAR and land prices. However, FAR is not exogenous. It is likely set in anticipation of land value, location attractiveness, infrastructure, and political priorities. Although district and year fixed effects are included, unobserved parcel-level amenities and expectations remain. This raises reverse causality concerns: High-value locations may receive higher FAR allowances, rather than higher FAR causing higher land value. The paper acknowledges omitted variable bias but does not attempt stronger identification strategies, such as Instrumental variables or regulatory discontinuities (?, not 100% sure). Without addressing this more rigorously, the elasticity estimate should be interpreted cautiously as associational rather than causal.]

Response 1: We thank the reviewer for the suggestion regarding stronger identification strategies, such as instrumental variables or regulatory discontinuities. Such approaches are appropriate when the objective is to identify the causal effect of exogenous FAR changes on land values. However, our research question is conceptually different.

[Following Brueckner et al., the coefficient on statutory FAR is interpreted as an equilibrium capitalization parameter that reflects how binding regulatory constraints are relative to market-determined development intensity. The purpose is not to estimate the causal impact of a regulatory shock, but to infer the degree to which statutory FAR limits compress market outcomes. Introducing an instrument to isolate exogenous variation in FAR would change the interpretation of the coefficient, converting it from a capitalization-based measure of regulatory stringency into a local treatment effect.

In the institutional setting examined here, statutory FAR limits are established through formal planning procedures and are relatively persistent. They are not continuously adjusted in response to short-term land price fluctuations. Although long-run spatial sorting may influence the allocation of FAR across locations, this does not undermine the reduced-form capitalization interpretation central to our framework.]

 [The revised text appears on Page 10, lines 367–373： We emphasize that our empirical specifications are designed to estimate the equilibrium capitalization of statutory FAR constraints into land values, rather than a causal effect of exogenous FAR changes. Accordingly, the estimated FAR coefficients are interpreted as reduced-form measures of how strongly FAR constraints are priced in the land market under binding regulation, and we use fixed effects and parcel controls to mitigate (but not fully eliminate) omitted-location heterogeneity.]

Comments 2: [Assumption that Built FAR Equals Maximum FAR

The paper assumes built FAR equals maximum allowable FAR. This is a contribution. In practice: Developers may under-build due to market demand or financing constraints. Such binding constraints may vary across time and districts. Some parcels may not reach full intensity immediately. If built FAR differs from statutory FAR, the theoretical mapping from elasticity to stringency becomes biased. This assumption requires empirical justification or sensitivity analysis.]

Response 2:  We thank the viewer for this important comment. [The theoretical mapping in Equation (4) rely on statutory FAR being binding in practice. In the institutional setting of urban China, residential land parcels are conveyed with clearly specified maximum FAR limits, and developers typically build close to the permitted intensity because land acquisition costs are high and under - utilization would reduce profitability.

Unlike contexts where zoning represents an upper envelope rarely reached, statutory FAR in Guangzhou functions as a binding regulatory ceiling that developers generally approach in completed projects.

Moreover, our empirical analysis is based on land transaction data at the time of conveyance, where land prices reflect expected development.]

[The manuscript has been revised accordingly, as detailed below:

Page 6, lines 258–264:

This reflects the institutional context in which residential land parcels are conveyed at substantial market prices and developers have strong incentives to utilize permitted development intensity. Although short-run deviations may occur due to construction timing or financing conditions, systematic long-run under-utilization of statutory FAR is uncommon in the residential land market. Therefore, this study assumes that built FAR approximates the maximum allowable FAR when interpreting the theoretical mapping.]

[Page 19, lines 692–699: In addition, the empirical framework assumes that built FAR approximates statutory FAR. In the institutional context of Guangzhou’s residential land market, developers typically build close to permitted intensity due to high land acquisition costs and binding planning conditions. Nevertheless, temporary or project-specific deviations between permitted and realized development intensity may occur. Future research using completed project-level data could further examine potential gaps between statutory and realized FAR. ]

Comments 3: [Interpretation of the “60–67%” Stringency Result：The conversion from elasticity to implied stringency depends critically on assumed β values (0.4–0.5). However, β is not estimated for Guangzhou. β is imported from other empirical contexts. Since the stringency calculation in Equation (4) is highly sensitive to β, the 60–67% estimate should be presented as illustrative rather than definitive. A sensitivity table (?) across a wider β range would strengthen credibilit.]

Response 3: Thank you for the comment. We agree that the implied stringency depends on the assumed production parameter β, which is not estimated specifically for Guangzhou. [In the revised manuscript, we clarify that β is calibrated from national-level evidence and that the 60–67% range should be interpreted as illustrative rather than definitive. We also report results over a broader range of plausible β values. The implied ratio remains within a moderate range, indicating that the qualitative conclusion does not rely on a narrow parameter choice.]

[The revised text appears on Page 12, lines 455–459 and in Table 3: The parameter β is calibrated using national-level estimates rather than city-specific estimation; accordingly, the 60%–67% range should be interpreted as an illustrative approximation rather than a definitive measure. Sensitivity analysis reported in Table 3 shows that the implied ratio remains within a moderate range across plausible parameter values.

Comments 4: [Spatial Gradient Interpretation

The finding that FAR stringency declines with distance from the centre is consistent with monocentric theory. However, the distance to GTT may proxy for unobserved amenity and accessibility gradients.

• The model does not include controls for infrastructure, transit access, school quality, or environmental amenities.

• Multicollinearity between centre and river distances suggests strong spatial clustering.

Therefore, the spatial gradient could reflect market fundamentals rather than regulatory targeting. More discussion distinguishing regulatory intent from market geography is needed.]

Response 4: Thank you for this insightful comment. We have revised the manuscript to clarify the interpretation of the spatial gradient. [Specifically, we acknowledge that distance may proxy for accessibility and amenity gradients, and therefore the results are interpreted as reduced-form spatial relationships reflecting the joint influence of market geography and regulatory implementation.

We now explicitly distinguish between the spatial gradient of FAR levels and that of FAR stringency. While market forces naturally generate higher development intensity near the urban core, our analysis focuses on the relative deviation between regulated and market-determined FAR. The negative distance coefficient thus reflects stronger regulatory compression relative to market fundamentals in central areas, rather than merely higher FAR levels near the center. In the absence of differentiated regulation, central locations would likely exhibit even higher density.]

[The manuscript has been revised accordingly, as detailed below:

Page 14, lines514–516:
“Importantly, this pattern pertains to the gradient of regulatory stringency rather than the gradient of FAR levels themselves.”

Page 14, lines 518–521:
“While market forces alone would naturally generate higher development intensity in central locations, the results suggest that the relative deviation between regulated and market-determined FAR is also greater in central areas, implying stronger regulatory compression near the urban core.”

Page 14, lines 480–481:
“The high correlation reflects the city’s spatial morphology, in which major development corridors follow river alignments, rather than model misspecification.”

Page 15, lines 538–540:
“Overall, the estimated spatial gradient should be interpreted as reflecting the joint influence of underlying market geography and the differentiated implementation of regulatory constraints across urban space.”]

Comments 5: [District-Level Interaction Interpretation：The interaction model in Table 5 produces district-specific FAR elasticities. However:

• The base Ln(FAR) coefficient is small or insignificant in some specifications.

• Interpretation of interaction terms requires clarity about reference categories.

• It is not always clear whether coefficients represent total elasticities or deviations from baseline.

A table presenting implied district-specific total elasticities explicitly would improve clarity]

Response 5: Agree. Thank you for this constructive comment. We have revised Section 4.5 to clarify the interpretation of the interaction model. [First, we now explicitly state that Nansha District is treated as the omitted reference category. Accordingly, the coefficient on ln(FAR) represents the effect of statutory FAR on land value in Nansha, while each interaction term captures the deviation from this base effect.Second, we clarify that the interaction coefficients represent deviations from the baseline, not total effects. The total FAR coefficient for each district is computed as the sum of the base ln(FAR) coefficient and the corresponding interaction term.Third, we report a separate table (Table 7) presenting the district-specific total FAR coefficients. This allows readers to directly compare the relative strength of FAR binding across districts.These revisions clarify the reference category and the interpretation of district-level heterogeneity in FAR regulation.]

[The manuscript has been revised accordingly, as detailed below:

Page 17, lines 603–608:Where z_i denotes the distance from a parcel to the subcentre of its district, and δ_c represents district fixed effects. In this specification, Nansha District is treated as the omitted reference category. The coefficient on ln(FAR), denoted α, therefore represents the effect of statutory FAR on land value in Nansha. The interaction term δ_c∗α₁ captures the district-specific deviation from this base effect. Accordingly, the total FAR coefficient in district c is given by α+δ_c×α₁.

Page 17, lines 619–622: Table 7 reports the implied district-specific total FAR coefficients calculated as the sum of the base ln(FAR) coefficient and the corresponding interaction term in column (2) of Table 6. These total coefficients reflect the relative strength of FAR binding across districts.]

Comments 6: [Conveyance Mechanism Interpretation

The claim that listing reflects stronger enforcement of FAR constraints may be overstated. Listing dominates the sample at 82%. Therefore, it may reflect the default institutional mode rather than stronger enforcement. Higher land prices under listing could simply reflect higher-quality parcels. The causal mechanism linking conveyance type to regulatory stringency is not fully known.

Also, there is a serious and unusual econometric issue in Table 2, titled Effects of FAR on Residential Land Values with Fixed Effects (2005–2025). The note indicates that robust standard errors are reported in parentheses, yet the standard errors appear as 0.000000. This is not plausible in applied microeconometric work using parcel-level land transaction data. Even with a large sample and fixed effects, standard errors cannot be zero, I supposed. Please check.]

Response 6: We thank the reviewer for this important comment. We agree that listing is the dominant conveyance mechanism in the sample and primarily reflects the standard institutional mode of land transfer rather than a distinct regulatory instrument. 

[Our revised manuscript no longer interprets the results as evidence of stronger regulatory enforcement under specific conveyance types. Instead, we clarify that differences across conveyance mechanisms capture heterogeneity in the institutional pricing environment under which land transactions occur. Since all residential land parcels are transferred through these formal mechanisms, land prices are formed within these institutional settings. The regression results indicate that statutory FAR constraints are capitalized to varying degrees across conveyance contexts, and the substantial increase in explanatory power after controlling for conveyance mechanisms suggests that institutional arrangements play an important role in shaping how FAR constraints are reflected in land values. We therefore interpret the results as evidence of institutional heterogeneity in FAR capitalization rather than as proof of differential enforcement intensity.

Regarding the econometric issue in Table 2, we sincerely apologize for the oversight. The values previously reported as 0.000 were due to a formatting error when transferring regression outputs into the manuscript. The robust standard errors are non-zero and have now been corrected to report appropriate values with adequate precision. We have carefully rechecked all regression tables to ensure accuracy and consistency throughout the manuscript.]

[The revised text appears on Page 13, lines 490–496：which may reflect heterogeneity in parcel characteristics, market demand, or administrative procedures associated with these mechanisms. Given that listing represents the dominant and default institutional mode of land transfer, its coefficient can be viewed as reflecting the average institutional context in which FAR constraints are priced in the market. Differences across mechanisms therefore indicate institutional heterogeneity in FAR capitalization rather than differential enforcement intensity.]

4. Response to Comments on the Quality of English Language

Response: We thank the reviewer for this comment. The manuscript has been carefully proofread and edited to improve clarity, grammar, and overall readability.

5. Additional clarifications

Comments 1: [First, although the authors now clearly state that the FAR coefficient should be interpreted as an equilibrium capitalization parameter rather than a causal estimate, the language in a few places still occasionally implies regulatory “effect” rather than capitalization. A careful pass to ensure consistent terminology throughout would prevent any residual ambiguity.]

Response 1: Thank you for your helpful comment. We understand the importance of using precise and consistent terminology to avoid ambiguity.

[In the revised manuscript, we carefully reviewed the relevant sections and replaced several expressions that previously referred to the “effect” of FAR regulation with language emphasizing capitalization. The FAR coefficient is now consistently interpreted as an equilibrium capitalization parameter, reflecting how statutory FAR constraints are capitalized into land values rather than representing a causal regulatory effect. We believe these revisions improve the clarity and precision of the manuscript.]

Comments 2: [Second, while the justification for the assumption that built FAR approximates statutory FAR is institutionally grounded, a brief descriptive check would enhance credibility. Even a short statement indicating whether a small subsample comparison, planning compliance record, or secondary source supports this claim would strengthen the empirical foundation.]

Response 2: Thank you for your valuable suggestion. We agree that providing additional context to support the assumption that built FAR approximates statutory FAR will enhance the credibility of our analysis.

 [The revised text appears on Page 6, Lines 258-263: This is consistent with the findings of Cai et al. (29), who investigate land developers’ compliance with FAR regulations using data from 30 major Chinese cities, including Guangzhou, matched with residential development projects, and note that despite occasional upward adjustments, developers in urban China are typically constrained by FAR regulations due to strong institutional incentives and compliance mechanisms.‘]

Comments 3: [Third, the sensitivity analysis on β is helpful, but the discussion could briefly comment on the economic meaning of the extreme values in the sensitivity table. This would help readers assess whether the broader parameter range is theoretically plausible or mainly illustrative.]

Response 3: Thank you for your helpful suggestion. We have revised the manuscript to include a brief discussion of the economic implications of the extreme values of β in the sensitivity analysis.  

[The revised text appears on Page 12, Lines 454-476: In the context of land development, β captures how capital investment translates into built floor area. A higher β indicates that capital is more productive in generating additional development density (FAR), while a lower β suggests stronger diminishing returns to capital, meaning that increasing FAR requires progressively more capital input per unit of built space.

Following previous empirical studies, we set the national benchmark value of β at 0.49 [26]. In the sensitivity analysis, we consider a range of β values between 0.35 and 0.6 to reflect potential variations in development technology across different districts and time periods in Guangzhou.

Higher β values may be more representative of core urban districts, such as Yuexiu and Liwan, where land scarcity and strong market demand make capital investment in vertical development relatively productive. In this case, the upper-bound assumption of β = 0.6 yields a regulated-to-market FAR ratio of approximately 0.52, implying a more stringent regulatory scenario in which statutory FAR limits substantially constrain the density that would prevail under market conditions.

Conversely, lower β values may better characterize peripheral areas with relatively abundant land supply, where development intensity is less responsive to capital investment. Under the lower-bound assumption of β = 0.35, the estimated ratio increases, indicating a less stringent regulatory environment in which statutory FAR is closer to the level that would be chosen under unconstrained market conditions.

Overall, the sensitivity analysis illustrates how plausible variations in development technology affect the inferred degree of FAR stringency while leaving the qualitative interpretation unchanged.]

Comments 4: [Fourth, in the district interaction section, although total coefficients are now reported, a short explanatory sentence clarifying economic magnitude, not only statistical significance, would improve interpretability.]

Response 4: Thank you for this insightful comment. We have added explanatory sentences that describe the economic implications of the total FAR coefficients.

[The revised text appears on Page 17, Lines 644-657: “To further illustrate district-level differences in FAR stringency implied by Equation (10), Table 8 reports the district-specific total FAR coefficients, calculated as the sum of the base ln(FAR) coefficient and the corresponding interaction term in column (2) of Table 7. These coefficients reflect the strength of FAR constraints across districts. Tianhe exhibits the highest FAR stringency at 1.385, followed by Yuexiu (1.287) and Baiyun (1.200). These coefficients suggest FAR constraints have the most impact on land values in central and economically active districts like Tianhe and Yuexiu. For instance, the 1.385 FAR coefficient in Tianhe implies that a 1% increase in FAR corresponds to an estimated 1.385% increase in land value. In contrast, Huangpu displays the weakest FAR stringency with a total FAR coefficient of 0.552, reflecting more flexible FAR constraints in peripheral areas. This suggests FAR constraints have less impact on land values in Huangpu. A 1% increase in FAR in Huangpu would result in only a 0.552% increase in land value, indicating FAR constraints are less binding.]