This article presents an operational statistical methodology to explicitly integrate the informal economy into Morocco's Social Accounting Matrix (SAM). It represents the first national level attempt within the Moroccan official statistical system to coherently link microeconomic and macroeconomic data within an integrated statistical framework. The objective is to enhance the SAM's capacity to reflect the real economic structure and to support policy evaluation addressing the structural duality of the Moroccan economy. The approach combines a bottom-up method, using data from the informal sector survey and the employment satellite account, with a top-down procedure that disaggregates macroeconomic aggregates based on ratios derived from informal sector data, using the national accounts' supply and use tables. Results reveal strong structural interdependencies between the formal and informal sectors. Pronounced backward linkages highlight the dependence of informal units on formal sector inputs, while significant forward linkages emphasize their pivotal role in supplying goods and services to households.
Credit scoring models trained exclusively on accepted applicants suffer from sample selection bias, a problem amplified in peer-to-peer (P2P) lending where rejection rates exceed 90%. Reject inference methods attempt to correct this bias by incorporating information from rejected applicants, yet most approaches assume linear decision boundaries and rarely evaluate whether the underlying missing-data assumptions hold empirically. We propose LogisticCEMD2, a semi-supervised classification algorithm that integrates degree-2 polynomial feature expansion with ℓ1 regularization within the Classification Expectation-Maximization (CEM) framework. The polynomial extension captures pairwise feature interactions and quadratic effects while ℓ1 penalization performs automatic feature selection among the expanded feature set. We develop a mini-batch variant with proven convergence guarantees and O(Jd2(B + Inl)) computational complexity. We evaluate the method on LendingClub data comprising 1,266,782 accepted and 2,401,300 rejected loan applications (2007–2018). Empirical validation reveals that the selection mechanism is strongly non-random (classification AUC = 0.943), establishing that the Missing at Random assumption underlying CEM is substantially violated. All subsequent results therefore characterize CEM behavior under documented Missing Not at Random conditions. Across 25 independent runs comparing 12 classifiers, Friedman tests confirm significant performance differences (χ2 = 193.4, p < 0.001). With the standard hard classification step, LogisticCEMD2 achieves AUC 0.602 with particular strength in recall (0.656), but exhibits poor calibration (Brier = 0.304). An ablation study with Bonferronicorrected Wilcoxon signed-rank tests reveals that replacing the hard classification step with soft posterior weights (SoftEM) recovers both discrimination (AUC = 0.633, matching the supervised baseline) and calibration (Brier = 0.236; Hosmer- Lemeshow p = 0.083), demonstrating that the discrimination loss is caused by the binarization of posteriors rather than the CEM framework itself. Sensitivity analysis reveals performance degradation when unlabeled samples substantially exceed labeled samples. These results establish that polynomial CEM with soft assignment preserves both discrimination and calibration under MAR violation, while the hard C-step variant trades discrimination for sensitivity. Practical value depends on the label assignment strategy and the asymmetric cost structure typical of credit scoring.
This paper examines whether Keynesian demand management can secure full employment once the cumulative path of output and the accompanying price dynamics are taken into account. Building on the principle of effective demand, it develops a reduced-form framework in which public expenditure, the interest rate and taxation determine a policy-induced component of expected output growth, unemployment depends jointly on inflation and output, and inflation evolves through proportional growth factors. Iterating the framework yields a horizon-dependent expression for expected unemployment that separates cumulative output expansion from inflation. Each equation is confronted with quarterly United States data for 1955Q1–2026Q2 using unit-root, cointegration, stability and pseudo out-of-sample forecast tests. Government expenditure relative to GDP is positively associated with subsequent growth, but the policy instruments explain only about 3% of its variance, their association with growth has weakened since the mid-1970s, and they add no predictive content beyond the persistence of output. Unemployment is unrelated to the level of output but closely tied to the output gap (coefficient −0.62), whereas its association with inflation is small and unstable. Conditional projections imply that raising the expenditure ratio by one percentage point lowers unemployment by about 0.15 points after two years. Full employment thus requires demand to outpace productive capacity, not merely sustained growth. JEL Classification: E12 ⋅ E24 ⋅ E31 ⋅ E52 ⋅ E62
Credit scorecards trained on accepted applicants alone carry a sample selection bias and ignore the asymmetric cost of misclassifying defaulters. We propose RI-CSCEM, a Classification Expectation--Maximisation procedure that fits a cost-sensitive logistic model on both accepted and rejected applicants. Each rejected applicant receives either a hard pseudo-label, assigned by a cost-weighted balance constraint, or a soft responsibility that distributes the applicant across both classes under the same constraint. We prove monotone descent of the objective for the hard variant and consistency of the estimator under a missing-at-random rejection mechanism. On Lending Club data, across thirty stratified draws and twelve competing models, RI-CSCEM matches the best accepted-only baselines on misclassification cost, with significance confirmed by the Friedman--Nemenyi procedure. An ablation study traces this performance to two components: cost-sensitive weighting moves the model off the trivial all-non-default solution, and soft responsibilities recover the discrimination that hard pseudo-labelling destroys, raising AUC from 0.589 to 0.614.
This paper develops a statistical framework for assessing Pareto-improving resource reallocations under weighted allocation schemes. The central objective is to characterize the admissible set of allocation weights that can improve at least one participant's outcome without reducing the outcome of any other participant. The framework formalizes Pareto admissibility through non-deterioration and strict-improvement conditions and interprets these conditions as testable restrictions on weighted reallocations. To illustrate the empirical relevance of the approach, the paper applies the framework to real data from the RAND Health Insurance Experiment, comparing access-related outcomes across cost-sharing regimes and health-status groups. The empirical application shows how the proposed framework can distinguish average improvements from genuine Pareto-admissible reallocations. The paper contributes a methodological tool for welfare analysis, public policy evaluation, and resource-allocation studies where distributional non-deterioration is analytically important. JEL Classification: D60; D61; D63; I13; C18; C21.