{"id":"0e1472ff-767e-424a-bdc9-a1c38860befa","arxiv_id":"2411.17165","paper_version":3,"verdict":"REJECT","confidence":"MODERATE","novelty_score":4.0,"correctness_risk":"high","formal_verification":"none","parameter_count":7,"one_line_summary":"The paper reports calibrated persistence values of 0.8 for the COVID demand shock and 0.9 or 0.95 for the vaccination supply shock in a behavioral DSGE model for India.","lead":"Using a behavioral New Keynesian DSGE model, the paper calibrates India's post-COVID output gap and inflation and claims behavioral expectations fit better than rational expectations. It also quantifies vaccination as a positive supply shock that lasted longer than the pandemic's negative demand shock.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The vaccination-persistence result is underidentified: three shock parameters are calibrated from only two moment targets, so the reported ordering rho_eta > rho_epsilon is not identified.","rationale":"The paper's central claim has two components: behavioral expectations fit better than rational expectations, and vaccination persistence exceeds demand shock persistence. The second component depends entirely on the calibrated values of rho_epsilon and rho_eta. These values are identified only through two mean targets, so the grid-search 'optimum' is not unique. The reader flagged the broader identification assumption (shock structure and assignment of the Q1 2020 dip); our concern sharpens this to underidentification from two moments. If the concrete test shows that many triples, including ones with rho_eta <= rho_epsilon, fit essentially equally well, the paper's vaccination conclusion is unsupported. The rational-expectations baseline is also not calibrated, supporting REJECT, but the underidentification alone is sufficient to reject the headline claim. Therefore we agree with the reader's REJECT verdict and see no reason to change it.","tokens_in":90,"tokens_out":5262,"duration_ms":68162,"concrete_test":"Recompute the Mahalanobis distance over the same grid for all (eta1, rho_epsilon, rho_eta) and record the set of triples whose distance is within 10% of the reported minimum. If this set contains triples with rho_eta <= rho_epsilon or with eta1 near zero, the ordering is an artifact of the grid. Also rerun the calibration with demand/supply durations changed to (9,5) and (11,7) quarters; if the best triple's ordering flips, the persistence comparison is not robust to the assumed shock horizons.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Section 4 calibrates the vector (eta1, rho_epsilon, rho_eta) by minimizing the Mahalanobis distance between simulated and actual post-COVID means of the output gap and inflation. With only two moments and three free parameters, the model is underidentified; a continuum of parameter triples can match the two means equally well. The grid search returns a single triple, but the coarse grid (increments of 0.01 for eta1 and 0.05 for rho_epsilon and rho_eta) does not resolve this non-uniqueness. Consequently, the headline conclusion that the vaccination shock's persistence (rho_eta) exceeds the COVID demand shock's persistence (rho_epsilon) is not an identified empirical finding. This also undermines the behavioral-versus-rational comparison, because the behavioral parameters themselves are not pinned down. The problem is compounded by the different assumed shock durations (10 quarters for demand, 6 for supply), which makes the persistence parameters non-comparable even if identified. The reader's weakest_assumption correctly identifies the shock structure as the soft spot; our concern sharpens it to the specific underidentification from having only two moment targets for three parameters.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper extends a behavioral New Keynesian DSGE model, following De Grauwe and Ji, to the Indian economy during the COVID-19 pandemic. It models the pandemic as a deterministic negative aggregate demand shock and the vaccination program as a positive aggregate supply shock, calibrating the initial supply shock magnitude and both persistence parameters by minimizing a Mahalanobis distance between simulated and actual post-COVID means of the output gap and inflation rate. The authors report that the behavioral-expectations model matches the first moments better than a rational-expectations model and that the calibrated persistence of the vaccination supply shock exceeds that of the COVID demand shock, which they interpret as evidence of the vaccination program's efficacy.","tokens_in":11960,"tokens_out":4331,"duration_ms":39318,"significance":"If the identification and comparison were valid, the paper would offer a useful quantitative application of behavioral DSGE modeling to an emerging economy and a novel quantification of vaccination effects within a microfounded framework. The authors are transparent about data sources, provide both HP and Kalman filter output gap estimates, and include a structural break test. However, the central empirical claims rest on an underidentified calibration and an uncontrolled comparison between behavioral and rational expectations, so the paper's headline findings are not currently established. The distributional comparison in Section 4.1 also shows substantial variance mismatches that are not addressed.","major_comments":[{"comment":"The calibration is underidentified: three parameters (eta_1, rho_epsilon, rho_eta) are chosen to match only two moment targets (mean output gap and mean inflation). A continuum of parameter triples can match the two means equally well, and the coarse grid search (increments of 0.01 for eta_1 and 0.05 for the persistence parameters) does not resolve this non-uniqueness. Consequently, the reported ordering rho_eta > rho_epsilon is not an identified empirical finding, and the conclusion about vaccination efficacy stated in the abstract and Section 5 is not supported by the calibration.","section":"Section 4, Eqs. (4)-(5) and Tables II-III"},{"comment":"The Mahalanobis distance is defined using Sigma = cov(sdata, ssim), which is not a standard covariance matrix for the moment vector. A proper Mahalanobis distance requires the covariance matrix of the moment estimator (e.g., the sampling covariance of sdata - ssim), not the cross-covariance between the data and simulated vectors. As written, the criterion is not justified and its properties are unclear, which affects the validity of the calibration result.","section":"Section 4, Mahalanobis distance definition"},{"comment":"The behavioral-vs-rational comparison is not controlled. The rational expectation model is assigned arbitrary boundary values (rho_epsilon = 0.0, rho_eta = 1.0, eta_1 = 1.0) rather than being calibrated by the same distance-minimization procedure. The behavioral model is optimized over a grid, so the lower Mahalanobis distance for the behavioral model is expected by construction and does not demonstrate that behavioral expectations are empirically superior.","section":"Section 4, Tables II and III, rational expectations row"},{"comment":"The simulated output gap variance (0.0609 for the HP-based model and 0.0500 for the Kalman-based model) is more than an order of magnitude larger than the corresponding actual variances (0.0028 and 0.0017, respectively). The paper claims the simulated data 'reasonably approximate the empirical distribution' based on Jarque-Bera p-values, but those p-values only test normality within each series and do not compare the simulated and actual distributions. The variance mismatch directly contradicts the claim of matching distributional characteristics and is not discussed.","section":"Section 4.1, Tables IV and V"},{"comment":"The identification of the vaccination effect rests on the assumption that the entire pandemic is captured by a deterministic negative demand shock lasting exactly 10 quarters and a positive supply shock lasting exactly 6 quarters, with epsilon_1 set equal to the Q1 2020 output gap dip. If other shocks, supply-side effects of COVID-19, policy responses, or alternative shock durations contributed to the 16 post-COVID quarterly means, the calibrated persistence comparison is not identified. The paper provides no robustness analysis with respect to these assumptions.","section":"Section 4, Eqs. (4)-(5) and shock structure"}],"minor_comments":[{"comment":"The notation for Sigma is undefined beyond 'cov(sdata, ssim)'; the paper should specify the dimension and the exact estimator used, and ideally provide a reference for this version of the Mahalanobis distance.","section":"Section 4, Mahalanobis distance formula"},{"comment":"The text says 'Tables III and IV present the Likelihood Ratio test results' but the relevant tables in the appendix are numbered VII and VIII; the cross-references are incorrect.","section":"Section 6.3, Tables VII and VIII"},{"comment":"The reference to 'Goyel and Arora (2016)' appears twice and should be 'Goyal and Arora (2016)'.","section":"Appendix, references"},{"comment":"The paper states that gamma and rho are taken from De Grauwe and Ji (2019) for the US; a sensitivity analysis on these parameters for India would strengthen the results, though this is not central to the main critique.","section":"Section 3, parameter table"},{"comment":"The paper reports running 88.2 million simulations but does not state the random seed or the exact method for generating the behavioral expectation draws; providing reproducible code or a precise algorithm description would be helpful.","section":"Section 4, simulation details"}],"recommendation":"reject","confidential_remarks":"The underidentification of the three shock parameters from two moment targets is a fundamental identification failure that undermines the paper's central conclusion about vaccination persistence. Additionally, the comparison of behavioral versus rational expectations is invalid because the rational model is not estimated under the same criterion. These issues cannot be fixed with minor revisions; the paper would need a substantially different identification strategy or additional moment conditions, which goes beyond the current scope. The distributional claims are also weakened by the large variance mismatch in Section 4.1. I therefore recommend rejection."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"You should know two things about this paper before reading it. First, it is a clean, straightforward application of De Grauwe's behavioral New Keynesian model to India's post-COVID output gap and inflation, and that country application is genuinely new. Second, the central claim—that the vaccination supply shock is more persistent than the COVID demand shock—is not identified by the calibration. Three parameters (eta_1, rho_epsilon, rho_eta) are grid-searched to match two moment targets (mean output gap and mean inflation), so a continuum of parameter triples fits equally well. The reported ordering rho_eta > rho_epsilon is therefore not an empirical finding; it is an artifact of the grid search and the arbitrary shock durations (10 quarters for demand, 6 for supply).\n\nWhat the paper does well: it carefully estimates India's output gap with both HP and Kalman filters, runs a proper structural break test for Q1 2020, and is honest about borrowing the learning intensity and memory parameters from De Grauwe and Ji. The shock structure follows Dasgupta and Rajeev's static model, and the DSGE extension is a legitimate contribution to the India-focused macro literature. The simulation methodology is transparent, even if the code itself is only available on request.\n\nThe soft spots are serious. The rational expectations baseline is not calibrated at all: the paper simply sets rho_epsilon=0, rho_eta=1, and the initial supply shock to 1, then declares the behavioral model superior because it gets closer to the means. That is not a controlled comparison. The first-moment fit is circular because the calibration targets exactly the moments later presented as successful matches. The higher-order moments do not match well: simulated output gap variance is 0.06 versus actual 0.003, roughly twenty times too large. And the Mahalanobis distance is specified as cov(sdata, ssim), which is not a standard covariance distance—with only two points in each vector, that matrix is degenerate. The paper seems to use it as a scalar normalization, but the description is confusing.\n\nThe appendix robustness exercise using the full 20-year sample is a different simulation with stochastic shocks and does not rescue the COVID calibration. The paper's central comparison is therefore not supported as stated.\n\nWho is this for? Readers interested in India-specific DSGE applications or in teaching calibration pitfalls will get something out of it. It deserves a serious referee only if the authors are willing to fix the identification problem—calibrate the rational baseline, use more moments or a likelihood, and report the fit honestly. In current form, I would not cite it for the vaccination result. My verdict: give it a chance in revision, but the load-bearing flaw must be addressed.","headline":"A readable but underidentified calibration exercise: the behavioral model fits India's post-COVID means only by construction, and the vaccination-persistence claim rests on three parameters fit to two moments.","tokens_in":757,"tokens_out":1001,"would_cite":false,"duration_ms":25688,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"Behavioral expectations beat rational models for India's COVID recovery","keywords":["behavioral expectations","New Keynesian DSGE","output gap","inflation","COVID-19","vaccination","Mahalanobis distance","India"],"falsifier":"Re-run the same Mahalanobis calibration on India's post-COVID output gap and inflation with an extended shock specification—for example, adding a separate negative supply shock during lockdown quarters, fiscal transfer shocks, or allowing the demand shock to last longer than ten quarters—and check whether the calibrated vaccination persistence ρη still exceeds the demand persistence ρε. If the ranking reverses or the behavioral model loses its fit, the paper's central claim about vaccination efficacy and behavioral expectations would be contradicted.","tokens_in":11437,"feed_emoji":"💉","tokens_out":2374,"duration_ms":24181,"temperature":0.7,"pith_summary":"This paper claims that a New Keynesian DSGE model with behavioral expectations—where agents switch between simple forecasting rules based on past success—matches India's post-pandemic output gap and inflation moments better than the same model with rational expectations. Using quarterly Indian data filtered by both Hodrick-Prescott and Kalman methods, the authors model COVID-19 as a ten-quarter negative demand shock and vaccination as a six-quarter positive supply shock, then calibrate the shock sizes and persistence by minimizing Mahalanobis distance. The calibrated behavioral model reproduces the actual post-COVID means of output gap and inflation, while the rational expectations model produces means far from the data. The paper reads the calibrated persistence parameters as evidence that the vaccination-triggered positive supply shock was more persistent than the COVID-induced demand shock, quantitatively extending an earlier static Keynesian analysis of the same episode.","feed_headline":"Behavioral forecasts fit India's post-COVID economy","feed_subtitle":"A DSGE model with adaptive forecasters matches India's output and inflation means, and puts the vaccine shock's persistence above the…","key_machinery":"The machinery is a three-equation behavioral New Keynesian DSGE model in the style of De Grauwe: an aggregate demand equation, a Phillips curve, and a Taylor rule, plus a behavioral expectation formation mechanism in which agents choose between fundamentalist and extrapolative forecasts using discrete-choice probabilities based on discounted mean squared forecast errors. The pandemic and vaccination enter as deterministic autoregressive shocks, ϵt = ρε^(t−1) ϵ1 for ten quarters and ηt = ρη^(t−1) η1 for six quarters, with ϵ1 taken from the observed output-gap drop. Calibration searches a grid over η1, ρε, and ρη and selects the triple minimizing the Mahalanobis distance between simulated and actual two-dimensional moment vectors (mean output gap, mean inflation), a scale-invariant distance that accounts for covariance between simulated and actual moments.","core_discovery":"The paper's central discovery is that replacing rational expectations with behavioral expectations in a small New Keynesian DSGE model lets the model match the first moments of India's post-pandemic output gap and inflation rate, whereas rational expectations fails even on the mean. After fixing the initial COVID demand shock to the observed Q1 2020 output gap dip, the grid-search calibration over the initial supply shock and the two persistence parameters yields (η1, ρε, ρη) = (0.64, 0.8, 0.9) for the HP-filter-based output gap and (0.57, 0.8, 0.95) for the Kalman-filter-based output gap. In both cases the behavioral model's simulated mean output gap and mean inflation nearly equal the actual post-COVID averages, with smaller Mahalanobis distance than the rational expectations model. The calibrated positive supply shock persistence ρη exceeding the negative demand shock persistence ρε is presented as evidence that India's vaccination program generated a sustained favorable supply-side impulse that outlasted the pandemic's demand drag.","pith_inferences":["One editorial extension is that the persistence comparison ρη > ρε may conflate vaccination with other post-2021 recovery forces—fiscal transfers, pent-up demand, or looser policy—since the model assigns all positive supply-side recovery to the vaccination shock.","Another extension is that the same behavioral DSGE calibration strategy could be applied to other emerging economies with observable output-gap dips and vaccination timelines, giving a cross-country measure of vaccination-driven supply persistence.","A testable implication not pursued in the paper is that the behavioral model should also match the sign and timing of forecast disagreement during the recovery; survey-based expectation dispersion data could be used to check the switching proportions estimated by the model."],"forward_implications":["If behavioral expectations are the right description, policy analysis for India should not rely on rational-expectations models that miss even the post-crisis mean of output and inflation.","The calibrated persistence ranking implies that vaccination programs can be quantified as measurable positive supply shocks within a standard structural model, not just as public-health events.","The model provides a dynamic complement to the static Keynesian analysis of the same episode, allowing shock persistence and magnitudes to be separated empirically.","The behavioral model's closer fit on higher-order moments (variance, skewness, kurtosis, and normality tests) suggests it can serve as a simulation tool for distributions, not just averages.","Successful replication on the full 20-year Indian sample supports the claim that behavioral expectations, not only extreme-shock episodes, better capture Indian business-cycle moments."],"supporting_citations":[{"why":"Supplies the behavioral macroeconomic model with fundamentalist/extrapolator expectation switching that is the paper's core departure from rational expectations.","marker":"[15]"},{"why":"Provides the behavioral expectation formation framework and simulation parameters (memory parameter ρ and learning intensity γ) adopted for India.","marker":"[16]"},{"why":"Gives the expression for the Phillips curve coefficient κ and the structural interpretation used to parameterize the supply side.","marker":"[17]"},{"why":"The Dasgupta and Rajeev static Keynesian analysis of COVID demand shocks and vaccination supply shocks that this paper extends into DSGE form.","marker":"[14]"},{"why":"Earlier Dasgupta and Rajeev work establishing the paradox of supply-constrained Keynesian equilibrium, grounding the shock interpretation for the pandemic.","marker":"[13]"},{"why":"Calvo staggered pricing underlies the Phillips curve derivation in the New Keynesian model.","marker":"[9]"},{"why":"Binder-Pesaran solution technique is used to simulate the rational and behavioral expectation DSGE models.","marker":"[5]"},{"why":"Supplies the Calvo price rigidity parameter θ used in the Indian calibration.","marker":"[22]"}],"fun_headline_variants":["Behavioral expectations beat rational in India DSGE","India's vaccine shock outlasts demand drag in DSGE","Adaptive forecasts match India's COVID output and inflation","DSGE with behavioral forecasts fits India's recovery","Vaccine supply shock persistence exceeds demand drag in India"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The paper assumes the pandemic is fully captured by a deterministic ten-quarter negative demand shock and vaccination by a six-quarter positive supply shock, with the entire Q1 2020 output-gap drop assigned to the demand shock; if other shocks, supply-side COVID effects, or policy responses contributed to the post-COVID quarterly averages, the calibrated persistence comparison is not identified.","fun_headline_variants_meta":{"raw":{"variants":["Behavioral expectations beat rational in India DSGE","India's vaccine shock outlasts demand drag in DSGE","Adaptive forecasts match India's COVID output and inflation","DSGE with behavioral forecasts fits India's recovery","Vaccine supply shock persistence exceeds demand drag in India"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000659,"raw_usage":{"total_tokens":3056,"prompt_tokens":1028,"completion_tokens":2028,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":644,"completion_tokens_details":{"reasoning_tokens":1951}},"tokens_in":644,"tokens_out":2028,"duration_ms":13066,"temperature":1.0,"reasoning_tokens":1951,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T12:26:37.108460+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Re-run the same Mahalanobis calibration on India's post-COVID output gap and inflation with an extended shock specification—for example, adding a separate negative supply shock during lockdown quarters, fiscal transfer shocks, or allowing the demand shock to last longer than ten quarters—and check whether the calibrated vaccination persistence ρη still exceeds the demand persistence ρε. If the ranking reverses or the behavioral model loses its fit, the paper's central claim about vaccination efficacy and behavioral expectations would be contradicted.","supporting_citations":[{"cited_title":"(2019), Behavioural Macroeconomics: Theory and Policy , Oxford University Press, UK","cited_arxiv_id":null,"evidence_quote":"Provides the behavioral expectation formation framework and simulation parameters (memory parameter ρ and learning intensity γ) adopted for India."},{"cited_title":"(2023), Covidonomics or the Curious Case of a Supply Constrained Keynesian Equilibrium, In: Raychaudhuri, A., Ghose, A","cited_arxiv_id":null,"evidence_quote":"The Dasgupta and Rajeev static Keynesian analysis of COVID demand shocks and vaccination supply shocks that this paper extends into DSGE form."},{"cited_title":"(2020), The paradox of a supply- constrained Keynesian equilibrium, Economic and Political Weekly , 55(22)","cited_arxiv_id":null,"evidence_quote":"Earlier Dasgupta and Rajeev work establishing the paradox of supply-constrained Keynesian equilibrium, grounding the shock interpretation for the pandemic."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Calvo staggered pricing underlies the Phillips curve derivation in the New Keynesian model."},{"cited_title":"(2000), Solution of finite-horizon multivariate linear ra- tional expectations models and sparse linear systems, Journal of Economic Dynamics and Control, 24(3), 325–346","cited_arxiv_id":null,"evidence_quote":"Binder-Pesaran solution technique is used to simulate the rational and behavioral expectation DSGE models."},{"cited_title":"(2023), A basic two-sector new keynesian dsge model of the indian economy, Theoretical and Practical Research in the Economic Fields , 14(1), 36","cited_arxiv_id":null,"evidence_quote":"Supplies the Calvo price rigidity parameter θ used in the Indian calibration."}],"review_version":1}