Moral attitudes towards effort and efficiency: a comparison between American and Chinese history

The study investigates how societies assign moral value to effort versus efficiency, focusing on whether effort is morally praised even when it is inefficient. Building on evidence that people infer moral worth from displays of effort irrespective of outcomes, the authors explore historical trajectories of these moral attitudes in two distinct cultural contexts: the United States (influenced by the Protestant Work Ethic) and the People’s Republic of China (influenced by Confucian values). The purpose is to trace how moral attitudes toward effort and efficiency evolved over time, how they relate to major socio-political and economic shifts, and how cultural values such as individualism/collectivism and cultural tightness/looseness may shape or be shaped by these attitudes. The study leverages large historical corpora and computational linguistics to overcome limitations of traditional surveys/experiments in uncovering long-term, society-level evolution of moral norms.

Prior research shows that effort is often moralized across contexts, with harder-working individuals perceived as more moral and deserving, even when outcomes are identical (e.g., Amos et al., Bigman & Tamir, Celniker et al.). Societal discourse also reflects this norm (e.g., wealthy Americans valorizing hard work). However, the consequences may diverge at individual (e.g., reduced learning, meaningless work) versus societal levels (e.g., persistence, innovation, growth). Cultural-religious traditions emphasize diligence (Protestant Work Ethic in the West; Confucian industriousness in East Asia). Cultures differ in valuing efficiency and equality (e.g., Confucian contexts prioritizing equality and cooperation; PWE linking discipline, efficiency, and achievement). Historical psychology has used corpora and NLP (frequency analyses, co-occurrence, embeddings) to reveal long-term cultural trends (e.g., individualism rise, cultural loosening, changing stereotypes). Much of this work is WEIRD/Eurocentric; fewer studies use Chinese or other non-English corpora. This study fills gaps by comparing U.S. and China, which differ in institutions, values (individualism/collectivism; tightness/looseness), and work ethics (PWE vs. Confucian), and by developing tools in both English and Chinese.

Design: Comparative, historical text analysis using NLP on two political corpora: U.S. Congressional speeches (1873–2011) and China’s People’s Daily (1950–2021). The study develops dictionaries for effort and efficiency in English and Chinese, measures their moral associations over time via word embeddings, and links trajectories to cultural variables and GDP using time-series models and Granger causality. Dictionary development: Seed words for effort and efficiency were gathered from WordNet (English) and Chinese equivalents. Using pre-trained embeddings (English: Google News word2vec, Wikipedia fastText, Twitter GloVe; Chinese: Renmin/people.cn, Wikipedia, Weibo), the mean vector of seed words was computed and the 50 nearest neighbors in each model were extracted. Two native English and two native Chinese raters evaluated conceptual relevance (1–5); words with mean rating >3 were retained, with further exclusions for duplication, low frequency, multi-token forms, or ambiguity. Inter-rater reliability (ICC) was good for effort (EN 0.88; ZH 0.76) and moderate for efficiency (EN 0.67; ZH 0.51). Final dictionaries: English effort (10 words), English efficiency (10); Chinese effort (13), Chinese efficiency (10). Examples: effort/toil/perseverance; efficiency/productivity/effective/economical/profitable; Chinese 努力/力求/尽力 and 效益/效率/效用. Moral lexicons: Moral Foundations Dictionary 2.0 (MFD 2.0) used for virtue/vice terms; Chinese version adapted from English and validated in prior work. Chosen for compatibility and positive/negative valence separation. Corpora and embeddings: Year-sliced corpora were trained into word2vec (gensim, skip-gram) with vector size 100, window 5, min_count 5, negative sampling 5, epochs 5. Token counts per year: U.S. mean ~13.4M (range ~0.87–29.25M), China mean ~12.3M (range ~5.29–31.04M). Operationalization of moral attitudes: For each year, compute average cosine similarity between concept dictionary (effort or efficiency) and moral virtue words, and between concept dictionary and moral vice words. Moral attitude = virtue similarity − vice similarity for each concept. Inefficient Effort = (Effort–Virtue similarity) − (Efficiency–Virtue similarity), indexing the positive moral bias toward effort beyond efficiency. Cultural covariates and controls: Yearly individualism/collectivism and tightness/looseness indices derived from Google Ngram frequency-based dictionaries (with reference checks against COHA for U.S. robustness). GDP per capita from Maddison Project (log-transformed in models). Analyses: (1) Descriptive trajectories. (2) Null model baselines via 10,000 random word-set simulations per year matched in dictionary size to derive 95% CIs. (3) Bayesian Change Point Detection (R bcp) to locate potential shifts and compare with historical events. (4) ARIMA models (R forecast auto.arima) relating moral attitude time series to cultural variables, with and without GDP control. (5) Granger causality tests to assess directionality between cultural variables and moral attitudes. Robustness checks included party-specific U.S. embeddings (Democrats vs Republicans), pruning dictionaries for words missing >50% of years, and rerunning ARIMA/Granger with pruned dictionaries and alternative cultural sources (COHA).

Descriptive and change-point patterns:

• United States (Congressional speeches, 1873–2011): Prior to the late 1940s, moral attitudes toward effort and efficiency were similar and often overlapped with null-model bounds. From the 1950s through the 1970s, both effort and efficiency gained positive moral value, with effort surpassing efficiency; moralization of inefficient effort increased during this period. Party-specific analyses showed largely overlapping trajectories for Democrats and Republicans (post-ARIMA correlations: effort r=0.31; efficiency r=0.46; inefficient effort r=0.46; all p<0.001).
• China (People’s Daily, 1950–2021): Effort consistently showed positive moral value above null-model confidence intervals across decades. Efficiency was generally within null-model intervals except during the 1990s–2000s. Inefficient effort remained positively valued overall but showed a notable change point around 1959; it declined after 1978 and more steeply after 1992, paralleling market-oriented reforms and increasing emphasis on efficiency. Time-series associations (ARIMA):
• U.S.: Effort positively associated with cultural looseness (b=0.03, p<0.001), with and without GDP control. Efficiency negatively associated with collectivism before GDP control (b≈−0.01, p<0.001), but not after; individualism became a negative predictor after GDP control (b≈−0.01, p=0.027). Inefficient Effort negatively associated with looseness after GDP control (b≈−0.06, p<0.001).
• China: Effort positively associated with collectivism after GDP control (b≈0.04, p<0.001). Efficiency positively associated with looseness before (b≈0.03, p<0.001) and after GDP control (b≈0.05, p=0.040). Inefficient Effort: collectivism positive before GDP control (b≈0.06, p=0.005); looseness negative after GDP control (b≈−0.15, p<0.001). Granger causality:
• U.S.: Individualism Granger-caused Effort (F(9,220)=3.05, p=0.002) and Efficiency (F(9,220)=2.56, p=0.008); Collectivism Granger-caused Effort (F(10,214)=2.89, p=0.002) and Efficiency (F(10,214)=3.12, p<0.001). Inefficient Effort Granger-caused Individualism (F≈1.93, p=0.049), Collectivism (F=3.30, p<0.001), and Tightness (F=2.45, p=0.009); reverse directions were generally not significant.
• China: Individualism (F(9,82)=2.07, p=0.042) and Collectivism (F(9,82)=2.97, p=0.004) Granger-caused Effort; Effort Granger-caused Tightness (F(9,82)=2.01, p=0.048). Efficiency Granger-caused Individualism (F(9,82)=2.33, p=0.021). Inefficient Effort and Collectivism showed bidirectional causality (F(9,82)=2.94, p=0.004; reverse F=2.37, p=0.020). Looseness preceded Inefficient Effort (F(2,124)=3.73, p=0.027) without reverse causality. Supplementary insights:
• Difficulty-as-improvement mindset related to moralization of effort in China (r=0.557, p<0.001) but not U.S. (r=0.092, p=0.285) after ARIMA detrending. Sentiment largely independent of moralization, except positive sentiment correlated with effort moralization in Chinese corpus. Pruned dictionaries and alternative cultural sources produced qualitatively consistent results.

The findings reveal culturally distinct trajectories and mechanisms in the moralization of effort and efficiency. In the U.S., consistent with PWE traditions, effort and efficiency both gained moral salience post-1950s, with effort surpassing efficiency, and cultural looseness associated with greater moral valuation of effort but lower valuation of inefficient effort when accounting for economic performance. In China, effort remained robustly moralized across decades, while efficiency showed limited moral salience except during market reform eras. The positive association between collectivism and effort (and bidirectional links with inefficient effort) suggests that Confucian-influenced norms valorize diligence and obligations independent of outcomes, though increasing looseness and market reforms corresponded with declining moralization of inefficient effort. Across both contexts, cultural looseness was negatively related to moralization of inefficient effort, implying that societies with more flexible norms may place less moral value on effort divorced from outcomes. Granger analyses indicate that shifts in individualism/collectivism tend to precede changes in moral valuation of effort (and efficiency), while inefficient effort can, in turn, forecast changes in broader cultural norms in some cases, highlighting dynamic interplay rather than a one-way causal story. The results underscore that the moral meaning of work is historically and culturally contingent and can shift with socio-political and economic transformations.

This study develops bilingual (English/Chinese) computational tools to track moral attitudes toward effort and efficiency and applies them to century-scale political corpora from the U.S. and China. It documents distinct historical trajectories: rising postwar U.S. moralization of effort (and efficiency) with effort surpassing efficiency, and consistently high Chinese moralization of effort alongside limited and period-specific moralization of efficiency, with declining moralization of inefficient effort during reform eras. Time-series models and Granger tests suggest that cultural values—especially individualism/collectivism and looseness/tightness—are associated with and in some cases precede changes in these moral attitudes, while inefficient effort can forecast shifts in cultural norms. The work provides evidence on the socio-cultural roots and historical contingencies of moralizing effort, offering a foundation for historical psychology of work ethics. Future research should broaden corpora beyond political texts, incorporate additional languages and genres, examine earlier historical periods (e.g., Reformation, Industrial Revolution; China’s late imperial to republican transitions), compare alternative moral lexicons, and explore how accelerating AI-driven automation may reshape the moral valuation of human effort and efficiency.

• Corpus scope: Analyses rely on formal political texts (U.S. Congressional Record; People’s Daily), which may reflect elite discourse and not the full society; results may not generalize to other genres or everyday language.
• Time frame: U.S. (1873–2011) and China (1950–2021) omit earlier eras that likely shaped work ethics; longer historical coverage could reveal additional dynamics.
• Moral lexicon choice: Dependence on MFD 2.0 and its Chinese adaptation; other moral dictionaries (e.g., eMFD) might yield different nuances; moral foundations demarcations vary across frameworks.
• Dictionary construction: Moderate ICC for Chinese efficiency terms (ICC≈0.51) warrants caution interpreting Chinese efficiency results; some dictionary terms missing in certain years (addressed via pruning robustness checks).
• Cultural variables: Google Ngram indices are imperfect proxies and subject to corpus composition biases; though COHA checks support U.S. robustness, Chinese alternatives remain limited.
• Statistical approach: ARIMA and Granger tests are exploratory/descriptive and do not establish definitive causal mechanisms; unmeasured confounds (political/economic shocks) may influence associations.