Capital Structure in French Family Firms After COVID-19: A Pecking Order Reassessment

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

The study aims to reassess the ability of the financing arrangement theory to explain financing decisions in French family firms compared to non-family firms following the COVID-19 pandemic. This is achieved by expanding the time frame of the analysis, employing a difference-in-differences model with a three-way interaction between family status, the post-pandemic period, and financing shortfalls, and analyzing the impact of access to secured loans. The study is distinguished by its application of the Shim-Sunder and Myers model and multiple robustness tests, which contribute to a deeper understanding of the changes in financing behavior in family firms during the post-pandemic period.


Introduction

1. The introduction is too long and contains methodological details and preliminary results, usually placed in the "Methodology" or "Results" section rather than in the introduction. This makes it crowded and distracts the reader from the main objective. It is better to simplify the introduction to focus only on the research problem, context, research gaps, and objectives of the study.
2. Confusion between the introduction and the executive summary, where the introduction presents the research design (e.g., 2SLS, triadic interaction, event study) and presents preliminary results in detail. This deviates from the purpose of the introduction, which should pave the way for the reader rather than present the results in advance.
3. Repetition in the literature, where there is extensive citation of previous work (Myers & Majluf, Frank & Goyal, Anderson & Reeb, etc.), but some of these references are cited more than once and in similar ways. This repetition reduces the clarity of the narrative and makes the text cumbersome.
4. The introduction devotes significant space to the literature on Pecking Order Theory and family firms, but it does not devote enough space to explaining why France in particular is so important from a scientific and practical perspective. While these points are mentioned, they are in passing compared to the density of methodological details.
5. The style is geared toward expert readers (many technical and methodological terms), which may be difficult for a wider audience to grasp. The introduction should be more narrative and streamlined to clearly state the research question before delving into the technical details.
6. The lack of an explicit research gap; the introduction does not clearly highlight the gap the study fills. There is a reference to the research "re-evaluating" Pecking Order in family firms after COVID, but it is not clarified whether previous studies on France are rare or whether the crisis represents a "unique research opportunity." Highlighting this point would strengthen the academic argument.
7. The narrative moves quickly without clear separation between the sections of the introduction. It would be better to restructure it to:
   
   Introduction to the problem (financing in the face of cash shortages).
   The role of family businesses in this context.
   The importance of COVID-19 as a research shock.
   Research gap and main objective.

 

 

 

Theoretical Framework

1. Excessive length and detail. The section is filled with minute details (especially in 2.1 and 2.3), which makes the reader lose track of texts and references before reaching the core of the argument. It is preferable to simplify the theoretical presentation through more concise formulation, especially since some paragraphs resemble a lengthy literature review rather than a "theoretical framework guiding the hypotheses."
2. Confusion between the theoretical framework and the literature review. There is a clear overlap between what is supposed to be a "theoretical foundation" and a "review of previous studies." For example, the paragraph "Evidence on Family Firms" (2.2) appears more like a literature review than a theoretical foundation for the research hypothesis. It is better to separate what constitutes a theoretical foundation from what constitutes supporting empirical evidence.
3. Weak connection between the hypotheses and the foundation. Although H1–H3 are clearly written, the transition to them sometimes comes after a long, unfocused overview. For example, H3 was inserted after a very long paragraph on banking evidence and financial policies, which could have been shortened to be more direct.
4. Lack of critical discussion of the literature. The section presents the findings of the literature in a descriptive manner (for example, some studies prove this, others indicate that), but it does not offer a critical position that clarifies the contradictions or gaps that the current research fills. It is important for the researcher to highlight: What is new in their hypotheses compared to the literature? Where does the research gap lie?
5. Excessive use of recent references at the expense of classics. While updating the references is good, the heavy reliance on very recent papers (2023–2025) may raise questions about the solidity of the classical foundations, especially since the topic is linked to theories dating back to the 1980s and 1990s. It is preferable to strike a balance between classic references (Myers & Majluf, 1984; Shyam-Sunder & Myers, 1999) and recent references.
6. Lack of a French dimension in the theoretical framework. Although the study applies to France, the theoretical section discusses family businesses globally in general terms without specificity. It would have been better to include literature or evidence specifically related to the French context, as this strengthens the logic of linking the theory to the studied environment.

Methodology

1. Sample Selection: The final number of firms after applying inclusion and exclusion criteria, nor the total sample size (N), is not stated. This makes it difficult for the reader to assess the strength of the sample. The time period 2003–2024 is quite long (22 years), but neither the homogeneity of the data nor the risks of structural changes in the French economy during this period are discussed. The exclusion of "financials" and "utilities" is logical, but the potential impact of this exclusion on the generalizability of the results is not explained.
2. Variable Construction: The Financing Deficit (DEF) equation is somewhat complex, and the economic rationale behind including each component is not explained. This may confuse readers who are not experts in this field.
3. It is not explained how family firms were identified in practice (was it through family names? Ownership ratios? Additional data from registers?). This detail is very important in research on "family firms."
4. The use of lagged controls is good, but the issue of dynamic bias is not discussed, nor is there a need to estimate dynamic models such as System GMM.
5. The possibility of endogeneity between DEF and change in debt is not discussed, although they briefly refer to 2SLS in the "Robustness" section. This point should have been addressed more clearly in the methodology rather than relegated to a footnote.
6. Relying solely on firm-level clustering may be insufficient, especially with long panel data (2003–2024), and the study may also need two-way clustering (firm × year).
7. 2SLS tests are mentioned, but without clarifying exactly which tools are used (only lags? How many periods?). The lack of detail here reduces the reader's confidence in the processing power.
8. There is no discussion of multicollinearity or a deeper examination of the data properties (correlation matrix, descriptive statistics).

Results

1. The sharp differences between family and non-family firms: The difference in the DEF coefficient (0.524 for family firms versus -0.117 for non-family firms) is very large compared to previous studies. This may indicate the presence of unobserved family-related factors (such as governance or banking relationships) that were not included.
2. The post-COVID effect: The results show an amplification of the effect for family firms after the pandemic, but this interpretation relies on the assumption of policy channels (subsidized loans, guarantees, moratoria).

Discussion

1. The discussion focuses more on family firms and neglects to analyze "contradictory" results in some sectors (e.g., transportation).
2. There is no link between the descriptive statistics and the time shock (COVID), which is discussed later in the basic models.
3. The narrative could have been strengthened by analyzing economic significance, not just statistical significance.

Conclusions

1. Despite mentioning measurement limitations and relying only on data from French firms, the text presents the results as if they are strongly causal, although the absence of bank-firm matched data or naturalistic instruments makes the interpretation vulnerable to bias.
2. The discussion of the robustness of the results across measurement alternatives (net-debt, long-term debt, family ownership thresholds) suggests strong robustness, but does not adequately address the risks of mechanistic analysis or association with unobserved factors.
3. The policy implications assume that credit programs have directly supported family firms, but long-term negative effects such as over-leverage or underinvestment in innovation are not examined.

Author Response

Point-by-Point Response to Comments (Reviewer #1)

Manuscript: Capital Structure in French Family Firms After COVID‑19: A Pecking Order Reassessment

Author: Faten Chibani, Jamel Eddin henchiri

This letter lists each comment, our response, the location of the revision, and the exact sentence(s) inserted or revised.

Thank you, Reviewer #1. We are grateful for your careful reading and constructive guidance. Your comments led us to (i) refocus the Introduction on the problem, the French context, the research gap, and objectives; (ii) streamline the theoretical framework and link it directly to H1–H3; (iii) clarify sample construction (reporting N and inclusion/exclusion logic) and add the economic intuition behind DEF; (iv) move and expand the endogeneity/IV discussion into Methods with full diagnostics; (v) report two‑way clustered standard errors (firm × year) as a robustness check in Supplementary Table S3—results are unchanged—while the main tables retain firm‑clustered SEs for comparability; and (vi) strengthen the discussion by connecting descriptives to the COVID‑time shock, adding economic magnitudes, and balancing policy implications with long‑run risks. Following your advice, we also revised the Abstract to be concise, non‑technical, and free of causal language. These changes substantially improved the clarity, rigor, and narrative flow of the paper.

Comment 1: The introduction is too long and includes methods/preliminary results that belong in “Methodology”/“Results”.

Response: We shortened the Introduction and removed all methodological and results content. It now focuses on the problem, France as a setting, the research gap, and objectives.

Where: Section 1. Introduction [p. 2–3].

Exact change(s):

• “France is an informative environment. Family ownership is pervasive among unlisted firms, and the financial system is bank‑based, making incumbent lender relationships central to day‑to‑day access to external finance.”
• “Despite this relevance, evidence for French private firms … remains limited. This scarcity defines the main research gap that this study addresses.”
• “Our contribution is threefold.”
• “The next section develops the theoretical background and hypotheses …”

 

Comment 2: Confusion with an executive summary; the intro details design (2SLS, triadic interaction, event study) and early findings.

Response: All design specifics (2SLS, interactions, event-time) and any results were moved to Sections 3–4; the Introduction now contains no coefficients or tests.

Where: Section 1 (no methods/results) [p. 2–3]; methods in 3.3 and results in 4.2–4.5.

Exact change(s):

• “The next section develops the theoretical background and hypotheses … Sections 3 and 4 present the data, variable construction, and empirical design, while Section 5 concludes.”

 

Comment 3: Repetition in the literature (e.g., multiple citations to Myers & Majluf; Frank & Goyal; Anderson & Reeb).

Response: We consolidated duplicative citations (one canonical source per claim) and removed overlaps.

Where: 2.1 Pecking order and family control; References.

Exact change(s):

• “In the canonical pecking‑order view, information frictions and issuance costs induce a financing hierarchy: firms use internal funds first, then debt, and resort to outside equity only as a last step (Myers & Majluf, 1984; Shyam‑Sunder & Myers, 1999; Frank & Goyal, 2003, 2009).”

 

Comment 4: Insufficient emphasis on why France matters.

Response: We added a dedicated paragraph tying France’s bank‑based system, prevalence of family control, and COVID credit backstops to our question.

Where: Section 1, paragraph beginning “France is an informative environment …”.

Exact change(s):

• “France is an informative environment. Family ownership is pervasive among unlisted firms, and the financial system is bank‑based … the COVID‑19 shock adds a clear credit‑supply lever: public guarantees, moratoria, and supervisory forbearance temporarily expanded banks’ lending capacity and lowered the relative cost of debt for viable borrowers …”

 

Comment 5: Tone too technical for broad readers.

Response: We simplified the prose, defined key terms before use, and deferred technicalities to Methods.

Where: Section 1 (entire), especially opening paragraphs; 2.1 for key definitions.

Exact change(s):

• “By ‘shadow cost of equity’ we mean the extra, non‑price cost families attach to issuing outside equity because it dilutes control.”

 

Comment 6: Lack of an explicit research gap.

Response: We state the gap explicitly and tie it to France and the COVID period.

Where: Section 1, paragraph beginning ‘Despite this relevance …’.

Exact change(s):

• “Despite this relevance, evidence for French private firms that directly ties financing deficits to debt adjustments across ownership types — and explicitly contrasts the pre‑ and post‑pandemic periods — remains limited. This scarcity defines the main research gap that this study addresses.”

 

Comment 7: Weak sub‑structure in the intro (problem → family firms → COVID shock → gap/objective).

Response: We re‑ordered the Introduction accordingly and added sign‑posting.

Where: Section 1, final sign‑posting sentence.

Exact change(s):

• “The next section develops the theoretical background and hypotheses derived from these motivations. Sections 3 and 4 present the data, variable construction, and empirical design, while Section 5 concludes.”

 

Comment 8: Theory section overly long/detailed (esp. 2.1 and 2.3).

Response: We shortened §2 by pruning non‑essential detail while preserving the argument.

Where: 2.1–2.3.

Exact change(s):

• Condensed single‑paragraph statement of the pecking‑order logic and family‑control mechanisms replaces prior multi‑paragraph digressions.

 

Comment 9: Confusion between theory and literature review.

Response: We structurally separated (A) foundations (2.1), (B) empirical regularities (2.2), and (C) macro‑shock logic and hypotheses (2.3).

Where: Headings 2.1, 2.2, 2.3.

Exact change(s):

• New subsection headers: “2.1. Pecking order and family control”, “2.2. Evidence on family firms”, “2.3. Macro shocks, credit supply, and the post‑pandemic environment.”

 

Comment 10: Weak linkage from foundations to H1–H3 (H3 arrives late).

Response: We tightened transitions and placed H1 at the end of 2.1 and H2–H3 at the end of 2.3 with one‑paragraph derivations.

Where: 2.1 (end), 2.3 (end).

Exact change(s):

• “H1. For a given financing deficit, the sensitivity of the change in the leverage ratio to the financing deficit is positive and larger in family‑controlled firms than in otherwise comparable non‑family firms.”
• “H2. After COVID‑19, family firms increased their reliance on debt to finance deficits more than non‑family firms.”
• “H3. The post‑COVID increase in debt‑financing sensitivity among family firms is stronger in credit‑intensive industries and for firms with tighter relationship banking.”

 

Comment 11: Lack of critical discussion of the literature.

Response: We added a short critical synthesis highlighting contextual contradictions and why France is informative.

Where: 2.2, closing lines.

Exact change(s):

• “Overall, the empirical evidence remains fragmented across contexts and definitions … This reinforces the need for a country‑specific test in France, where family prevalence and bank intermediation coexist.”

 

Comment 12: Too many recent references vs. classics.

Response: We re‑balanced by foregrounding the classical foundations while retaining selective recent work.

Where: 2.1 Foundations; References.

Exact change(s):

• Canonical citations now appear once per claim (Myers & Majluf, 1984; Shyam‑Sunder & Myers, 1999; Frank & Goyal, 2003, 2009; Fama & Jensen, 1983).

 

Comment 13: Add a French dimension in the theoretical framework.

Response: We wove the bank‑based, relationship‑lending features of France into the theory and hypothesis logic.

Where: 2.2 (France paragraph) and 2.3 (policy transmission).

Exact change(s):

• “France combines a dense base of non‑listed, family‑controlled firms with a bank‑based financial system where relationship lending is central.”

 

Comment 14: Sample size (N) not stated; long window 2003–2024; effect of excluding financials/utilities.

Response: We report N, discuss the horizon, and motivate exclusions; we also note FE structure to address macro changes.

Where: 3.1 Sample selection.

Exact change(s):

• “The final analysis sample contains 4,004 firm‑year observations — 2,200 family and 1,804 non‑family — drawn from roughly 1.3k distinct firms; by construction, the median firm contributes ≥ 3 years.”
• “We exclude financials (NACE K) and regulated public utilities … All continuous variables are winsorized at the 1st/99th percentiles by year.”

 

Comment 15: DEF equation complex; add economic rationale for each term.

Response: We added intuitive explanations for DIV, CAPEX, ΔNWC, CPLTD, and CFO, in addition to the formal equations.

Where: 3.2 Variable construction.

Exact change(s):

• “Why each term is in the numerator (uses of cash): DIV (Dividends): paying owners uses cash; if operating cash is insufficient, dividends must be financed externally or reduced.”
• “CAPEX (Capital expenditures): investment outlays that create assets but require funding when internal cash falls short.”
• “ΔNWC … We define ΔNWC as the change in inventories/receivables net of payables.”
• “CPLTD (Current portion of long‑term debt due within 12 months): scheduled amortization is a cash outflow … if internal cash cannot cover repayments, the firm must refinance or raise equity.”
• “CFO (Operating cash flow after interest and taxes): cash generated by operations; it reduces the need for external finance.”

 

Comment 16: How are family firms identified in practice?

Response: We detailed the DIANE/Orbis‑based procedure (UBO tracing, aggregation of kinship-linked stakes, handling ambiguous cases) and alternative thresholds.

Where: 3.1, identification paragraphs; 4.4.1/Table 6 for alternative definitions.

Exact change(s):

• “We use the DIANE/Orbis shareholding module … to retrieve named natural‑person blockholders, their percent stakes, and holder type (individual vs. corporate). When the top shareholder is a holding company, we trace through to the ultimate owner(s).”
• “A firm‑year is coded family‑controlled (FAM=1) if one or more identifiable family members are the largest controlling block … We aggregate family members’ stakes … Ambiguous cases remain uncoded until control can be established from filings.”

 

Comment 17: Lagged controls introduce potential dynamic bias; consider System‑GMM or justify not using it.

Response: We clarified our choice: baseline FE + 2SLS with predetermined lags as instruments; System‑GMM is not adopted as baseline given panel structure and instrument weakness risk.

Exact change(s):

• “Financing needs and debt adjustments may be jointly determined within the year. To address potential simultaneity and measurement error in DEF, we instrument the contemporaneous deficit with its first and second lags (L1.DEF, L2.DEF). We deliberately restrict the instrument set to two lags … We report Kleibergen–Paap … and Hansen J.”

 

Comment 18: Endogeneity between DEF and ΔDebt should be addressed in Methods (not only in robustness).

Response: We moved and expanded the IV discussion to §3.3 and report first‑stage strength and over‑ID tests.

Where: 3.3 (IV discussion) and Table 9.

Exact change(s):

• “We instrument the contemporaneous deficit with its first and second lags (L1.DEF, L2.DEF) … We report Kleibergen–Paap rk Wald F and Hansen J; instruments are strong by conventional thresholds and are not rejected by over‑identification tests.”

 

Comment 19: Firm‑only clustering may be insufficient; consider two‑way (firm × year).

Response: We have clarified the inference structure throughout the manuscript. The main tables continue to report one-way clustering at the firm level, which is standard in long firm-panel settings. To ensure robustness, we re-estimated all models with two-way clustering (firm × year) and report the results in the Supplementary Materials (Table S3). The estimates and significance levels remain virtually unchanged, confirming that the inference is not sensitive to the clustering choice.

Where: Supplementary table S3.

 

Comment 20: 2SLS details unclear (which instruments/how many lags).

Response: We now specify L1.DEF and L2.DEF (L3 tested in robustness) with a rationale for parsimony and provide full first‑stage diagnostics.

Where: 3.3 and Table 9.

Exact change(s):

• “We deliberately restrict the instrument set to two lags … Using L1 only or L1–L3 yields similar coefficients.”

 

Comment 21: No discussion of multicollinearity or basic data properties.

Response: We added descriptive statistics, a correlation matrix, and VIFs (≈1.2).

Where: 4.1 Descriptives and correlations; Tables 2–3.

Exact change(s):

• “The mean variance‑inflation factor (VIF) is ~1.2, indicating no material multicollinearity.”

 

Comment 22: Large split in DEF coefficients suggests omitted family‑specific channels.

Response: We acknowledge the magnitude and discuss governance/relationship‑banking channels; heterogeneity tests probe these mechanisms.

Where: 4.2 Baseline results; 4.3 Mechanisms.

Exact change(s):

• “Beyond statistical precision, these patterns matter on corporate balance sheets … In family firms, shortfalls of internal funds are primarily bridged with bank debt rather than outside equity or asset disposals … For otherwise similar nonfamily firms, financing gaps are more often closed through nondebt margins …”

 

Comment 23: Post‑COVID effect relies on assumed policy channels without direct evidence.

Response: We use cautious language (“consistent with”), add event‑time dynamics, industry×year saturation, and IV with lagged DEF; amplification concentrates where passthrough should be strongest.

Where: 4.3 and 4.4; Table 7 (Ind×Year FE), Table 11 (event‑time), Table 9 (IV).

Exact change(s):

• “We interpret the post‑COVID‑19 increase in the family‑specific slope (δ₃>0) as consistent with a credit‑supply channel … We therefore use cautious language (‘consistent with’) …”

 

Comment 24: Discussion under‑analyzes contradictory sectors (e.g., transportation).

Response: We added sectoral commentary linking asset tangibility, working‑capital needs, and credit intensity to apparent descriptive ‘contradictions.’

Where: 4.1 (end) and 4.3 (mechanism narrative).

Exact change(s):

• “Some industries display patterns that depart from the aggregate picture, notably transportation … These deviations concern levels of DEF, not the marginal responses estimated in the regressions … the transportation sector is also one of the most credit‑intensive and relationship‑dependent …”

 

Comment 25: No link between descriptives and the COVID time shock later modeled.

Response: We now bridge levels to dynamics and make that link explicit.

Where: 4.1, closing paragraph.

Exact change(s):

• “The descriptive contrasts in Tables 1–2 … provide the starting point for the time‑series analysis.”

 

Comment 26: Emphasize economic (not just statistical) significance.

Response: We translate interactions into marginal effects in percentage‑point terms.

Where: 4.2 (interpretation of Table 4, Panel B).

Exact change(s):

• “Economically, a 10‑pp increase in DEF is associated with ≈ +5.07 pp in ΔDT for family firms pre‑COVID and ≈ +7.36 pp post‑COVID, versus −1.17 pp and −1.03 pp for non‑family firms.”

 

Comment 27: Conclusion reads as strongly causal despite data/ID limits.

Response: We toned down language, presenting findings as descriptive and noting identification limits (no bank–firm matches; no firm‑level policy take‑up).

Where: 5. Conclusion (opening and policy paragraphs).

Exact change(s):

• “We present these findings as descriptive rather than structural causal effects …”

 

Comment 28: Robustness doesn’t fully address mechanistic risks/unobserved factors.

Response: We expanded robustness (alternative family definitions; leverage constructs; industry×year FE; winsorization; IV) and explicitly caution against mechanistic interpretation.

Where: 4.4.1–4.4.4; Section 5 (caveats).

Exact change(s):

• “Robustness checks indicate that the core patterns do not hinge on measurement choices … At the same time, we acknowledge the risk of mechanistic interpretation and therefore avoid strong causal claims.”

 

Comment 29: Policy implications assume direct support; discuss long‑run downsides (over‑leverage, innovation).

Response: We added a balanced policy discussion (stabilization vs. debt‑overhang and potential crowd‑out of innovation) and propose guardrails for deleveraging.

Where: 5. Conclusion, Policy implications.

Exact change(s):

• “Policy implications should be read with caution … the same forces can create debt‑overhang risk if recovery is slow and may crowd out investment in innovation … Program design that pairs access with guardrails for timely deleveraging … could temper these risks.”

 

Comment A (from existing points):

#2 “Confusion with an executive summary; the intro details design (2SLS, triadic interaction, event study) and early findings.”

#5 “Tone too technical for broad readers.”

Response A: We revised the Abstract to remove design specifics (e.g., 2SLS, triple interactions, event‑time), coefficients, and table mentions, and to adopt accessible language that foregrounds question, setting (France), and descriptive findings with an explicit non‑causal caveat.

Where: Abstract (first paragraph) in the revised manuscript.

Exact change (excerpts):

Removed (original abstract): “…we replicate the original approach and extend it along three dimensions: (i)… (ii) an explicit difference‑in‑differences specification with a triple interaction… (iii) heterogeneity tests… We estimate a two‑stage model à la Shyam‑Sunder & Myers…”

Added (revised abstract): “We examine how firms finance deficits when cash is tight… We find that family firms consistently use debt to bridge shortfalls… The amplification is stronger in credit‑intensive sectors and for firms with deeper bank ties. The results, presented without strong causal claims, connect control preservation and intermediation to marginal financing choices… ”

 

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

This study findings are theoretically consistent, empirically sound, and offer valuable insights into how credit policy and control considerations jointly shape corporate financing behavior in the post-pandemic era. However, there are some minor editorial and typographical issues that require correction. Please refer to the attached file for detailed comments.

     

Comments for author File: Comments.pdf

Author Response

Comment 1. The research model is indented inconsistently compared to other paragraphs; please adjust the indentation for alignment.

Response 1. Done. We unified the paragraph and equation styles: the displayed models use the same left indent as the body text and the same “Equation” style throughout.
Where: §3.3 Regression model and control variables, Equations (1) and (2).

Comment 2. All variable names (symbols) should be written in italics.

Response 2. Implemented. We created a Word character style for variables and applied it consistently to inline symbols and table stubs (e.g., ΔDT, DEF, DT, size, profit, tang, growth).
Where: Throughout (§§3.2–3.3; table headers and notes).

Comment 3. Several minor errors and inconsistencies in the tables:

(3.1) In Table 2, Panel B, the final row value “4004” should be deleted.
Response 3.1. Deleted. Panel B now contains only the Non‑family counts; All‑sample N is no longer shown in that panel.
Where: Table 2, Panel B.

(3.2) In Table 3, the notation “N = 4004” as a separate row appears awkward; please move it next to the title (“Table 3 (N = 4004)”).
Response 3.2. Revised. We removed the “N = 4004” row from the body and placed the sample size in the table title: “Table 3 — Pearson correlation matrix (N = 4,004).”
Where: Table 3 title.

(3.3) In Table 4, Panel A1, “N” and “Observations” are duplicated—please remove one.
Response 3.3. Fixed. Only “Observations” remains at the bottom of Panel A1; the redundant “N” line was removed. In the current version, Panel A1 lists “Observations: 2000” for family and “1640” for non‑family, with no duplicate “N.”
Where: Table 4, Panel A1.

(3.4) Where possible, please report Adjusted R‑squared instead of R‑squared in the regression tables.
Response 3.4. Implemented. We now report Adjusted‑R‑squared in regression tables (and keep the unadjusted value only where the template requires). For example, Table 4 Panel A1 shows both R‑squared and Adjusted‑R‑squared (0.460 and 0.028 in the two columns), and Table 4 Panel B includes Adjusted‑R‑squared in the panel footer.
Where: Table 4 (Panels A1 & B) and other regression tables where applicable.

(3.5) In Table 4, Panel C, ensure consistent reporting of decimal places for Wald test values (e.g., three or four digits throughout).
Response 3.5. Standardized. Wald F‑statistics and p‑values in Panel C now use a uniform precision (three decimals for F; three or four for p as needed). For instance: 495.220; 14.085; 0.081; 45.676; 251.481 with correspondingly rounded p‑values.
Where: Table 4, Panel C.

(3.6) In Tables 8, 10, and 11, “Observations” are listed as “2200.000”; please correct this to “2200.”
Response 3.6. Corrected. Observation counts are integers everywhere (e.g., 2000, 1640, 1046, 954 etc., depending on the split). See, for example, Table 8 (2000 and 1640) and Table 11 (2000 and 1640).
Where: Tables 6–11 (checked globally; examples: Table 8 & Table 11).

(3.7) The variable “DEF” is uppercase in the model/appendix but lowercase in results tables—please standardize the notation.
Response 3.7. Harmonized. We now use uppercase “DEF” everywhere (text, equations, table stubs, and appendix). See the variable list “DEF{i,t}” in Appendix A1 and the DEF row labels in Table 4.
Where: §3.2–3.3; Table 4; Appendix A1 (variable construction).

Author Response File: Author Response.pdf