## E-commerce during Covid in Spain: One “Click” does not fit All

## Source details

**Canonical URL:** [E-commerce during Covid in Spain: One “Click” does not fit All](https://www.imf.org/-/media/files/publications/wp/2024/english/wpiea2024107-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2024/english/wpiea2024107-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2024/english/wpiea2024107-print-pdf.pdf.json)

---

### I. Introduction
- Aggregate e-commerce spending in Spain surged during Covid and reverted quickly toward (or below) the pre-Covid trend as mobility restrictions disappeared.
- Aggregate behavior hides important heterogeneity by gender, age, and rural vs. urban residence.
- Key behavioral observations:
  - All groups increased use of e-commerce as pandemic restrictions intensified.
  - Women, youth, and urban consumers used e-commerce proportionally more during the pandemic, especially for services.
  - Use of e-commerce reverted to pre-pandemic trends in most groups as restrictions waned.
- Paper organization (as described):
  - Section 2: literature review.
  - Section 3: dataset on e-commerce in Spain.
  - Section 4: aggregate trends.
  - Section 5: panel data analysis.
  - Section 6: conclusions, contrasting persistence of working from home vs. e-commerce.

### II. Data sources and measurement
- Source: comprehensive individual-level financial transaction database from Banco Bilbao Vizcaya Argentaria (BBVA).
- Sample:
  - Random sample of 1,000 active BBVA clients (active = at least ten transactions each quarter).
  - Panel period: January 2017 to December 2022 (client-level); aggregate trends extended to January 2024.
  - Excludes self-employed and firms; weighted to match Spanish adult population on gender, age and income.
- Payment instruments included:
  - a) BBVA credit and debit card transactions; b) direct debits paid through BBVA current accounts; c) direct transfers done by BBVA clients; d) cash withdrawals done with BBVA cards (imputed across COICOP categories).
- Classification and definitions:
  - Each card PoS transaction tagged as online vs. offline; definition of “e-commerce” = cardholder and card not physically present (internet, telephone, mail-order).
  - From BBVA PoS, 14 percent are online in the studied period.
  - Transactions classified by COICOP categories using MCC; direct debits/transfers classified via NACE.
- Sample transaction totals (Annex Table 1):
  - Sample accounts for 1,8 million transactions and 110 million euros from January 2017 to December 2022.
  - Payment-share breakdown (volume):
    - Card transactions: 40 percent (30 percent physical cards, 10 percent online).
    - Cash: 21 percent.
    - Transfers and direct debits: 39 percent.
  - Offline card spending shares: Restaurants and hotels: 29 percent; Transport: 24 percent; Recreation & culture: 11 percent.
  - Online card spending shares: Transport: 31 percent; Clothing: 27 percent.
  - Transfers/direct debits shares: Housing utilities: 33 percent; Communications: 15 percent.
- Consumption measured in nominal terms; total consumption excludes imputed rents and covers all means of payment.

### III. Aggregate stylized facts (2017–Jan 2024)
- Pre-Covid trend (beginning of 2017 to just before March 2020):
  - Online consumption ratio rose by approximately 3.5 percentage points, from 4.5 percent to about 7.9 percent.
- During and after lockdowns:
  - E-commerce ratio peaked at around 10 percent in June 2020 (an increase of nearly 2 percentage points in four months).
  - E-commerce shares showed spikes around new variants (EU-Delta March 2021, Omicron November 2021).
  - By beginning of 2024: share of online consumption at 10 percent (described as “a share slightly below the levels observed before the pandemic”).
  - E-commerce share by number of transactions: below trend by end of Q1 2024 (Annex Figure 3).

### IV. E-commerce by categories (2017–24)
- Categories with temporary increases during mobility restrictions:
  - Essential goods (Food & Beverages, Alcohol & Tobacco, Health products): notable online growth during restrictions, receded after restrictions lifted.
  - Home equipment and maintenance: temporary spike during lockdowns; higher e-commerce share thereafter (7.5 percent at the beginning of 2024 vs. 3.0percent pre-Covid).
- Categories with sustained increases:
  - Textiles and Footwear: increased from 10 percent in January 2020 to 35 percent by March 2024 (sustained, permanent increase).
- Other category patterns:
  - Housing and energy utilities: did not increase above trend during lockdowns; continued pre-Covid growth trend until mid 2023; high variation thereafter.
  - Transport: online spending dropped from 10.0 percent to 4.5 percent during initial lockdown; recovered in early 2021 to 16.5 percent by beginning of 2023.
  - Communication: positive secular trend; reached maximum share of 6.5% at the beginning of 2024.
  - Recreation & Culture: continued pre-Covid upward trend with temporary Covid-aligned increases.
  - Restaurants & Hotels: short-lived boost during lockdowns; returned to pre-Covid trend by beginning of 2024.
  - “Other” (Insurance and Financial payments): upward trend, reached 8.0% at the beginning of 2024.

### V. Demographic patterns (2018–24)
- Age cohorts:
  - <35 years: pre-Covid e-commerce ≈ 15 percent; grew during pandemic to near 23 percent by end of period; only cohort with e-commerce shares remaining above trend by beginning of 2024.
  - 35–45: accelerated during pandemic and aligned to pre-Covid trend thereafter.
  - Older cohorts (45–55, 55–65, >65): temporary increases but returned to below pre-Covid trend by beginning of 2024.
  - Conclusion: only <35 (and to a lesser extent 35–45) sustained higher post-Covid shares; youngest cohort not large enough to drive aggregate trends.
- Gender:
  - Women had higher e-commerce share than men by about 2 percentage points pre-Covid; gender gap widened post-Covid and further increased by beginning of 2024.
- Urban vs Rural:
  - Both accelerated during Covid but returned to pre-Covid trends thereafter; urban effects are nuanced after controlling for other characteristics.

### VI. Micro empirical methodology
- Main regression specification estimates individual-sector-month e-commerce share s_i,s,r,t as function of:
  - Male_i, Urban_i, Age_i (or Young dummy), pos_s,pre (pre-pandemic sector online share), Lockdown_r,t (pandemic or continuous Google residential mobility), fixed effects: sector (α_s), region (β_r), time (γ_t), individual (δ_i).
- Interaction specification includes interactions of Male, Urban, Age, pos_s,pre with Lockdown to capture heterogeneity in responsiveness.
- Sample and estimation details (selected):
  - Observations: 473,315; variations with 225,214 in post-pandemic subsets.
  - No. Clients: 1000; No. Categories: 19; No. Provinces: 51.
  - Periods reported: 2017m1 to 2022m12; some specifications 2020m2 to 2022m12.
  - Standard errors clustered at client; various specifications include client fixed effects.

### VII. Micro empirical findings — key magnitudes and heterogeneity
- Main demographic and sector patterns:
  - Women spend more online than men; younger individuals spend more online than older individuals.
  - Individuals in sectors with higher pre-pandemic online shares spend more online.
  - Urban individuals show higher online shares in some specifications; insignificance after full controls in others.
- Pandemic and mobility effects:
  - Pandemic period associated with higher online shares; effects increase with intensity of mobility restrictions (Google residential mobility).
  - Interaction highlights:
    - Responsiveness to restrictions increased for young, women, and those in sectors with higher pre-pandemic online shares.
- Representative magnitudes (reported based on Column 8 and other columns):
  - Men have 1.7 percentage points lower e-commerce shares on average compared to women.
  - With mobility restrictions at maximum, the gender difference widens to 4.7 percentage points (women higher).
  - Pre-pandemic sectoral shares: a one percentage point higher pre-pandemic online share associated with 5.5 percent higher e-commerce shares on average; effect increases to 7 percent at maximum mobility restrictions.
  - Urban individuals: on average 18 percent higher online shares relative to rural; increases to 21 percent at maximum pandemic restrictions.
- Selected regression coefficients and significance (preserved as reported):
  - Age coefficient examples: -0.0020***, -0.0022***, -0.0024***, -0.0025***.
  - Young dummy coefficients: 0.0481***, -0.0331, -0.006, -0.043 (various specifications).
  - Avg. sector online share pre-pandemic: 1.0526***, 1.1337***.
  - Pandemic = 1 (March 2020 onwards): 0.0185***, -0.002 (specifications vary).
  - Mobility Restriction (Residential) coefficients: 0.0018***, 0.0023***, 0.0021***, 0.0020***, 0.0019***.
  - Male*Mobility Restriction: -0.0010***, -0.0012***.
  - Young*Mobility Restriction: 0.0009**, 0.0012***.
  - Mobility * Avg. sector online share pre-pandemic: 0.0076***, 0.0070***.

### VIII. Goods versus services heterogeneity
- Younger individuals’ online purchases skew toward services rather than goods; triple interactions for increased responsiveness not significant for services.
- Post-pandemic e-commerce shares higher for services than goods (goods dummy interactions negative with post-pandemic and mobility).
- But goods sectors with higher pre-pandemic online shares experienced relatively higher online shares post-pandemic and under increased mobility restrictions (positive significant triple interactions).
- Selected goods-related coefficients (preserved):
  - Goods*Young: -0.055***, -0.055***, -0.0974**, -0.130***.
  - Pandemic*Goods: -0.0205***, -0.0212***.
  - Pandemic*Goods*Avg. sector online share pre-pandemic: 0.4192***, 0.4650***.

### IX. Extensive margin (learning to go online)
- Definition: extensive-margin clients = 0–1 transactions pre-pandemic who increased to >=3 transactions during/after pandemic.
- Findings:
  - 25 percent of clients with 0–1 transactions pre-pandemic increased to >=3 (Annex Table 4).
  - Extensive-margin clients exhibit much higher online shares post-pandemic (Figure 4), consistent with a learning channel and enduring change among new online customers.
  - Negative relationship between initial pre-Covid online share and increase in online spending during pandemic; stronger for extensive-margin clients (a one pp lower pre-Covid online share associated with higher subsequent online engagement for extensive-margin clients).
  - Reported statistic: 0.27 pp higher online share on average for these customers (reported in the source).

### X. Robustness checks
- Mobility constructed as continuous variable with pre-pandemic assumed zero; main findings robust (Annex Table 5).
- Horse-race between pandemic dummy and continuous mobility variable: e-commerce shares for men and younger individuals higher as restrictions increase, controlling for interactions with post-pandemic period (Annex Table 6).
- Additional robustness (Annex Table 7):
  - Replacing amount-share with transaction-count share: results robust.
  - Controlling for cash spending: cash not a significant determinant of e-commerce shares; interaction between cash usage and restrictions negative and marginally significant.
  - Redefining pandemic dummy to March 2020–December 2021: findings broadly similar.

### XI. Interpretation, conclusions, and open questions
- Main conclusions:
  - Covid-19 increased e-commerce share over total spending but the aggregate increase was largely transitory; large heterogeneity across demographic groups observed.
  - Evidence consistent with several mechanisms: fear/avoidance, compulsive buying/hoarding, and technology competency enabling online adoption.
  - Persistent change concentrated among specific groups: young consumers (especially <35) and new online adopters (extensive margin) show enduring higher online shares.
  - Contrast with working from home (WFH): WFH exhibited persistent post-Covid changes in other studies, whereas e-commerce did not for most groups. Suggested reasons:
    - Limited scope for discovery/learning in e-commerce because it was well established pre-Covid.
    - Groups with low pre-pandemic e-commerce increased adoption (consistent with discovery/coordination conjectures).
    - Coordination costs differ: WFH requires sustained organizational coordination (hysteresis), while e-commerce is easily reversible day-to-day.
- Open questions and suggested future research:
  - Study other behaviors beyond e-commerce and WFH to evaluate whether discovery and coordination predict permanent change.
  - Distinguish between two forms of e-commerce shifts:
    - (i) preference for different store/form of payment for the same good (e.g., buying a ticket online), and
    - (ii) preference for a different good or service (e.g., online movies replacing theater visits).
  - Distinguishing these cases requires other datasets and is crucial to assess welfare gains from online purchasing.

### XII. Annex highlights and selected summary statistics
- Annex Figure 1: Distribution of consumption categories by payment method using COICOP and MCC/NACE classifications.
- Annex Figure 2: Aggregate trends in e-commerce shares and Oxford Stringency index.
- Annex Figure 3: Aggregate e-commerce share measured as number of online transactions over total transactions (3-month moving average).
- Annex Figure 5: Relationship between pre-Covid e-commerce share and increase during pandemic; definition of extensive margin and plotted differences.
- Annex Table 1 (transactions, 2017–2022):
  - Total transactions: 1.83 (millions), 110.55 (millions, euros).
  - Card transactions: 1.28 (millions), 44.05 (millions, euros).
  - Offline cards: 1.06 (millions), 33.44 (millions, euros).
  - Online cards: 0.22 (millions), 10.61 (millions, euros).
  - Cash: 0.16 (millions), 23.16 (millions, euros).
  - Transfers and direct debits: 0.38 (millions), 43.34 (millions, euros).
- Annex Table 2 (selected summary statistics, full sample N = 473,315):
  - Online spending % (transactions) mean = 0.0820, sd = 0.242, min = 0, max = 1.
  - Online spending % (amount spent) mean = 0.0783, sd = 0.243.
  - Cash % (amount spent) mean = 0.182, sd = 0.217, max = 0.998.
  - Age mean = 55.51, sd = 14.73, min = 22, max = 103.
  - Extensive margin client mean = 0.0458.
  - Avg. online spending % in category pre-pandemic (amount) mean = 0.0689, sd = 0.0825, max = 0.262.
  - Hybrid Pandemic (Residential mobility-based) mean = 102.7, sd = 5.133, min = 96.03, max = 133.5.
- Post-pandemic sample N = 225,214 (selected):
  - Online spending % (transactions) mean = 0.0920, sd = 0.254.
  - Online spending % (amount spent) mean = 0.0880, sd = 0.256.
  - Cash % (amount spent) mean = 0.158, sd = 0.209.
  - Mobility Restriction (Residential) mean = 105.6, sd = 6.232, min = 96.03, max = 133.5.
  - Extensive margin client mean = 0.0485.

*Source — wpiea2024107-print-pdf (IMF Working Paper — Annex Figures and Tables extracted).*

### References  ............................................................................................................

### References

### I. Introduction
- The share of e-commerce spending boomed during Covid in Spain and then reverted quickly back (or below) to the pre-Covid trend as the Covid pandemic and attendant mobility restrictions disappeared.
- The aggregate behavior of e-consumption masks important heterogeneity across demographic groups by gender, age, rural vs. urban.
- Key behavioral observations:
  - All groups used more E-commerce progressively as pandemic restrictions increased.
  - Women, youth, and urban consumers used e-commerce proportionally more during the pandemic, especially for services.
  - The use of e-commerce reverted back to its pre-pandemic trend in all groups as the pandemic and attendant containment measures waned.
- Paper organization (as described):
  - Section 2: literature review on ecommerce and Covid theories.
  - Section 3: dataset on e-commerce in Spain.
  - Section 4: aggregate trends.
  - Section 5: panel data analysis.
  - Section 6: conclusions, comparing persistence of working from home vs. ecommerce.

### II. Literature
- Five related strands:
  - E-commerce and structural transformation before the pandemic and welfare implications.
  - Why some groups responded more or less to the pandemic.
  - Why some behavioral changes persist in e-commerce.
  - Long-term effects of Covid on behavior beyond e-commerce.
  - Use of private data to improve and complement survey-based economic measurement.
- Selected referenced findings and claims:
  - Dolfen et. al. (2020): e-commerce yielded consumers the equivalent of a 1 percent permanent boost to their consumption (United States transaction-level data).
  - Jo, Matsumura, and Weinstein (2022): impact of retail e-commerce on pricing behavior and welfare (Japanese data).
  - Couture et. al. (2020): randomized control trial and microdata from China to estimate impact of first nationwide e-commerce expansion on rural households.
  - Protection Motivation Theory and Theory of Planned Behavior proposed to explain pandemic-related shifts in shopping preferences.
  - Moon et al. (2021): evidence in Korea favors Protection Motivation Theory.
  - Keane and Neal (2021): document consumer panic or hoarding during the pandemic.
  - Alcedo et al. (2023): e-commerce share went back to pre-Covid trends in 47 countries as mobility restrictions were lifted.
  - Auer, Cornelli, and Frost (2023): find no persistence in online shares using different data sources.
  - Barrero, Bloom, and Davis (2023): Covid changed preference for working from home and argue the change is permanent.
  - Alipour et. al. (2022): use card transactions from Mastercard for German cities to evaluate geographical relocations of offline consumption after COVID-19.
  - Wang et al. (2022): literature review on COVID-19 effects on a variety of consumer behaviors.
- Methodological literature using private data for real-time measurement:
  - Cavallo (2013): online price data for inflation measurement.
  - Chetty et al. (2020) and Carvalho et al. (2021): track economic activity in real-time during COVID-19.
  - Aladangady et al. (2019): use credit and debit card transactions to create daily estimates of retail spending to approximate official Census retail surveys in the United States.
- Contribution of the paper:
  - Examines dynamics of e-commerce during Covid across demographic groups (not only aggregate level).
  - Shows contrast between temporal persistence of working from home (which stuck) and e-commerce (which did not).

### III. Data Sources and Description
- Data source:
  - Comprehensive financial transaction database at the individual level from Banco Bilbao Vizcaya Argentaria (BBVA).
- Sample design:
  - Random sample of 1,000 clients of BBVA drawn from a larger sample of active clients (active clients defined as those that make at least ten transactions each quarter).
  - Panel period: January 2017 to December 2022.
  - Sample includes only Spanish consumers excluding self-employed and firms.
  - Sample is weighted to correct for biases with the Spanish adult population in terms of gender, age and income, following Buda et al. (2022).
- Payment instruments and classification:
  - Payments included:
    - a) transactions with credit and debit cards issued by BBVA;
    - b) direct debits paid by BBVA clients through their current accounts;
    - c) direct transfers done by BBVA clients;
    - d) cash withdrawals done with BBVA cards at ATMs or over the counter (assumed used for consumption and distributed across COICOP categories according to observed COICOP distribution for physical debit and credit cards).
  - Each card transaction is recorded at a point of sale (PoS) and tagged as online vs. offline.
  - From total number of BBVA PoS, 14 percent are online in the studied time period.
  - Each transaction is classified according to COICOP categories from merchant category codes (MCC); direct debits and transfers classified via NACE.
- Sample transaction totals and distribution:
  - The sample of 1000 individuals accounts for 1,8 million transactions, which represents 110 million euros from January 2017 to December 2022.
  - Annex Table 1 (described) distribution across payment methods:
    - 40 percent comes from card transactions according to the volume (30 percent physical cards and 10 percent online).
    - Cash accounts for 21 percent.
    - Transfers and direct debits for 39 percent.
  - Offline (in-person) card spending distribution highlights:
    - Restaurants and hotels: 29 percent of total.
    - Transport: 24 percent.
    - Recreation & culture: 11 percent.
  - Online card (e-ecommerce) spending distribution highlights:
    - Transport: 31 percent.
    - Clothing: 27 percent.
  - For other means of payment (direct debits and transfers):
    - Housing utilities: 33 percent.
    - Communications: 15 percent.

*Source: wpiea2024107-print-pdf - References*

### Annex Figure 1).

### wpiea2024107-print-pdf - Annex Figure 1)

### Data and measurement
- Dataset includes clients’ sociodemographic characteristics: age, gender, geographical location (rural or urban) and province.
- Consumption is measured in nominal terms.
- Total consumption record excludes imputed rents and covers all means of payment: cash, online and offline card transactions, direct debits and money transfers (Buda et al.,2022).
- Definition: "e-commerce" spending = transactions where the cardholder and the card are not physically present (internet, telephone, mail-order). Transactions not meeting this are labeled "Offline".
- Individual e-commerce share s_i,t:
  - s_i,t = Total Online spending_i,t / Total Consumption_i,t
- Category-specific e-commerce share s_i,j,t:
  - s_i,j,t = Total Online Spending by category_i,j,t / Total Consumption by category_i,j,t

### Background on Covid and Spain
- First Covid case in Spain: January 31st 2020 (Canary Islands); mainland cases in late February 2020; significant outbreak in early March 2020 in Madrid, Catalonia, Basque Country.
- Spain declared a national state of emergency and enforced stringent lockdowns; residents allowed out only for essential activities or work when remote work not feasible.
- Government introduced “Plan for the Transition to a New Normality” with phased easing by region.
- National state of emergency concluded on June 21, 2020.
- Fall 2020–early 2021: regional restrictions reimposed, including curfews, perimetral lockdowns, and limits on gathering sizes.

### Aggregate stylized facts on e-commerce (2017–Jan 2024)
- Pre-Covid (beginning of 2017 to just before March 2020): online consumption ratio rose by approximately 3.5 percentage points, from 4.5 percent to about 7.9 percent.
- After lockdowns: e-commerce ratio peaked at around 10 percent in June 2020 (an increase of nearly 2 percentage points in four months).
- During the pandemic: e-commerce shares fluctuated with spikes around new variants (EU-Delta March 2021, Omicron November 2021).
- By beginning of 2024: share of online consumption at 10 percent (described as “a share slightly below the levels observed before the pandemic”).
- E-commerce share by number of transactions: below trend by end of Q1 2024 (Annex Figure 3).
- Client-level data availability: until December 2022; aggregate trends extended to January 2024.

### E-commerce by categories (2017–24)
- Categories with temporary increases during mobility restrictions:
  - Essential goods (Food & Beverages, Alcohol & Tobacco, Health products) experienced notable online sales growth during restrictions and receded after restrictions lifted.
  - Home equipment and maintenance: temporary spike during lockdowns, leading to higher e-commerce share thereafter (7.5 percent at the beginning of 2024 compared to 3.0percent pre-Covid).
- Categories with sustained increases:
  - Textiles and Footwear: increased from 10 percent in January 2020 to 35 percent by March 2024 (sustained, permanent increase).
- Housing and energy utilities:
  - Did not increase above trend during lockdowns; continued pre-Covid growth trend until mid 2023 driven largely by direct debit payments; high variation thereafter.
- Transport:
  - Online spending dropped from 10.0 percent to 4.5 percent during initial lockdown; began recovery in early 2021 reaching 16.5 percent (slightly above pre-Covid trend) by beginning of 2023.
- Communication sector:
  - Positive secular trend throughout period, reaching a maximum share of 6.5% at the beginning of 2024.
- Recreation and Culture: continued pre-Covid upward trend with temporary Covid-aligned increases.
- Restaurants and Hotels:
  - Initially low e-commerce share, significant short-lived boost during lockdowns; returned to pre-Covid trend by beginning of 2024.
- "Other" category (Insurance and Financial payments): upward trend, reached 8.0% at the beginning of 2024.

### Demographic patterns (2018–24)
- Age groups:
  - <35 years: pre-Covid e-commerce around 15 percent; grew during pandemic to near 23 percent at the end of the period; only cohort with e-commerce shares remaining above trend by beginning of 2024.
  - 35–45: acceleration in e-commerce consumption and aligned to pre-Covid trend.
  - Older cohorts (45–55, 55–65, >65): temporary post-Covid increases but returned to shares below pre-Covid trend by beginning of 2024.
  - Conclusion: only <35 (and to a lesser extent 35–45) sustained higher post-Covid shares; youngest cohort not large enough to drive aggregate trends.
- Gender:
  - Women had an initial higher e-commerce share than men by about 2 percentage points pre-Covid.
  - Gender gap widened post-Covid and further increased by beginning of 2024.
- Urban vs Rural:
  - Similar growth trends; both accelerated during Covid but returned to pre-Covid trends thereafter.

### Micro empirical methodology
- Main specification (Equation (1)) estimates e-commerce share s_i,s,r,t as a function of:
  - Male_i (dummy = 1 if male)
  - Urban_i (dummy = 1 if urban)
  - Age_i (or alternative young dummy = bottom quartile)
  - pos_s,pre (pre-pandemic average of online spending in sector s)
  - Lockdown_r,t (measure of lockdown restrictions; alternative pandemic dummy from March 2020 or continuous Google residential mobility measure)
  - Fixed effects: sector (α_s), region (β_r), time (γ_t), individual (δ_i)
- Interaction specification (Equation (2)) includes interactions of Male, Urban, Age, pos_s,pre with Lockdown to capture changing elasticities during restrictions.
- Summary statistics and socioeconomic characteristics reported in Online Annex (Annex Table 2, Annex Figure 4).

### Micro evidence: key empirical findings
- Main patterns (Table 1, Columns 1–5):
  - Women spend more online than men.
  - Younger individuals spend more online than older individuals.
  - Individuals in sectors with greater pre-pandemic online shares continued to spend more online.
  - Urban individuals: not significantly different from rural after controlling for other characteristics (in specifications without individual fixed effects).
  - Pandemic associated with a rise in online shares relative to pre-pandemic; higher e-commerce shares post-pandemic and increasing with intensity of pandemic restrictions (Google mobility). Effect robust to individual fixed effects (Column 5).
- Interaction results (Table 1, Columns 6–9):
  - Responsiveness to pandemic restrictions increased for young, women, and those spending in sectors with higher pre-pandemic online shares.
- Magnitudes (based on Column 8):
  - On average, men have 1.7 percentage points lower e-commerce shares compared to women.
  - With mobility restrictions at maximum, the gender difference widens to 4.7 percentage points (women higher).
  - Pre-pandemic sectoral shares: on average, a one percentage point higher pre-pandemic online share is associated with 5.5 percent higher e-commerce shares; effect increases to 7 percent at maximum mobility restrictions.
  - Urban individuals: on average 18 percent higher online shares relative to rural; increases to 21 percent at maximum pandemic restrictions.
- Goods versus Services (Annex Table 3):
  - Younger individuals’ online purchases skewed towards services rather than goods; responsiveness did not significantly increase post-pandemic or with higher restrictions (triple interactions not significant).
  - Post-pandemic e-commerce shares higher for services than goods (negative interaction coefficients for goods dummy with post-pandemic and mobility restrictions).
  - Spending in goods’ sectors with higher pre-pandemic online shares reported relatively higher online shares during post-pandemic and with increased mobility restrictions (positive, significant triple interaction).

### Micro evidence: extensive margin (learning to go online)
- Definition: “extensive margin” clients = 0–1 transactions pre-pandemic who increased to >=3 transactions during/after pandemic (Annex Table 4).
- Finding: 25 percent of clients who did 0–1 transactions pre-pandemic increased to >=3.
- Extensive margin clients exhibit much higher online shares post-pandemic (Figure 4), consistent with learning and enduring change among new online customers.
- Additional finding: negative relationship between initial pre-Covid online share and increase in online spending during pandemic; stronger for extensive margin clients (a one pp lower pre-Covid online share associated with higher subsequent online engagement for extensive margin clients).

### Interpretation and discussion
- Evidence can be reconciled with multiple theories of consumer behavior around shocks:
  - Fear: avoidance of going out leading to higher online purchases for some groups.
  - Compulsive buying/hoarding: purchases to build home reserves in certain categories.
  - Technology competency: necessary enabling factor—higher spikes in sectors with higher pre-pandemic online presence suggest prior technological adoption mattered.
- Heterogeneity matters: women and young appear less fearful and more compulsive; technological competency amplified uptake where pre-existing online intensity was higher.
- Comparison with work-from-home (WFH):
  - WFH shows enduring effects post-Covid in other studies (e.g., Barrero, Bloom, Davis 2021).
  - E-commerce shows transitory aggregate effect with persistent heterogeneity: new online adopters (extensive margin) show enduring higher online shares, suggesting a learning channel for a subset of consumers.

*Italic: Source — wpiea2024107-print-pdf - Annex Figure 1).*

### 0.27 pp higher online share on average for these customers. This could be suggestive evidence for learning by

### E-commerce during Covid in Spain: One “Click” does not fit All

### Robustness checks and methods
- Constructed a continuous mobility variable for pre- and post-pandemic samples by assuming mobility restrictions take a value of zero in the pre-pandemic variable; main findings remain robust (Annex Table 5).
- Conducted a "horse race" between the pandemic dummy and continuous mobility restriction variable: Ecommerce shares for men and younger individuals are higher as pandemic restrictions increase controlling for interactions with the post-pandemic period (Annex Table 6).
- Additional robustness tests (Annex Table 7):
  - Replaced share of online amount with share of number of transactions conducted online; results robust.
  - Controlled for spending in cash: cash is not a significant determinant of e-commerce shares; interaction between cash usage and pandemic restrictions is negative and marginally significant (as mobility restrictions went up, lower usage of cash was associated with higher online shares).
  - Redefined the pandemic dummy to take value one for March 2020 to December 2021 (instead of the post-pandemic dummy in Table 1); findings remain broadly similar.

### Main empirical findings (as reported in Table 1 and related tables)
- Demographic and sector patterns:
  - Male = 1 coefficients (selected): -0.0079*, -0.0100**, -0.0074, -0.0094*, -0.0126***, 0.0882***, -0.004, -0.004, -0.005, -0.005, -0.004, -0.033.
  - Urban = 1 coefficients (selected): -0.0049, -0.0007, -0.0081, -0.0039, 0.0012, 0.0277, -0.005, -0.006, -0.006, -0.006, -0.006, -0.04.
  - Age coefficient: -0.0020***, -0.0022***, -0.0024***, -0.0025***, 0000.
  - Young (Dummy if Age < percentile 25) = 1 coefficients: 0.0481***, -0.0331, -0.006, -0.043.
  - Avg. sector online share pre-pandemic: 1.0526***, 1.1337***, -0.032, -0.035.
- Pandemic and mobility effects:
  - Pandemic = 1 (March 2020 onwards): 0.0185***, -0.002.
  - Mobility Restriction (Residential): 0.0018***, 0.0023***, 0.0021***, 0.0020***, 0.0019***, 0, -0.001, -0.001, -0.001, -0.001.
  - Pandemic*Young interaction: 0.0098**, 0.0104** (-0.004 standard errors reported).
  - Pandemic* Avg. sector online share pre-pandemic: 0.1519***, 0.1582*** (-0.024 standard errors reported).
  - Male*Mobility Restriction (Residential): -0.0010***, -0.0012***.
  - Young*Mobility Restriction (Residential): 0.0009**, 0.0012***.
  - Mobility* Avg. sector online share pre-pandemic: 0.0076***, 0.0070*** (-0.002 standard errors reported).
- Sample, estimation, and fit details (selected):
  - Observations: 473,315; 225,214 (various specifications).
  - R-squared examples: 0.15, 0.162, 0.161, 0.176, 0.262, 0.154, 0.237, 0.166, 0.263.
  - Periods reported include: 2017m1 to 2022m12; 2020m2 to 2022m12.
  - No. Clients: 1000.
  - No. Categories: 19.
  - No. Months: 72; 33 in post-pandemic subsamples.
  - Average Dep. Var. examples: 0.0783, 0.088.
  - Average Pandemic. Var.: 0.493; 105.6 (reported in table entries).
  - Average Male: 0.526. Average Urban: 0.203, 0.204. Average Age: 55.51, 55.39.
  - Average Cash % of amount: 0.182; 0.158 (post-pandemic).

### Goods vs services heterogeneity (Annex Table 3 highlights)
- Goods-related interactions (selected coefficients):
  - Goods = 1: 0.00010, 0.0018 (-0.002 standard errors).
  - Goods*Male: 0.0105**, 0.0066, -0.0073, 0.0024.
  - Goods*Urban: 0.0101, 0.0091, 0.0446, 0.007.
  - Goods*Young: -0.055***, -0.055***, -0.0974**, -0.130***.
  - Pandemic*Goods: -0.0205***, -0.0212***.
  - Pandemic*Goods*Avg. sector online share pre-pandemic: 0.4192***, 0.4650*** (-0.075, -0.072).
  - Goods* Res. Mobility * Avg. sector online share pre-pandemic: 0.2727***, 0.2593*** (-0.014, -0.013).
- R-squared examples in goods vs services table: 0.15, 0.161, 0.155, 0.238, 0.173, 0.269.

### Conclusions and interpretation
- The Covid-19 pandemic increased e-commerce share over total spending; the paper documents large heterogeneity across demographic groups and a surprising return to normalcy for most groups after the pandemic waned.
- Several theories (compulsive behavior, hoarding, fear) are consistent with observed heterogeneity; these theories imply transitory effects that dissipate once the pandemic wanes. The paper’s evidence supports this transitory interpretation.
- Contrast with working from home (WFH): Covid had a permanent effect on WFH (as surveyed by Barrero, Bloom, and Davis (2023)), but not on e-commerce for most groups. Reasons emphasized:
  - Limited “learning by locking” for e-commerce because e-commerce was already well established before Covid; scope for discovery and learning was limited.
  - Groups that used less e-commerce pre-pandemic show increases post-pandemic (consistent with BBD conjecture about discovery and coordination), whereas the most experienced users did not learn much by being locked.
  - Coordination costs: WFH requires coordination between employers and employees and organizational change that creates hysteresis; e-commerce requires less coordination and can be more easily reversed day-to-day.
- Open questions and suggestions for future research:
  - Study other behaviors beyond e-commerce and WFH to evaluate whether discovery and coordination are the primary predictors of permanent change.
  - Distinguish between two forms of e-commerce shifts: (i) preference for different store/form of payment for same good (e.g., buying a ticket online in advance) versus (ii) preference for a different good or service (e.g., online movies replacing theater visits). Distinguishing these cases requires other datasets and is important for assessing welfare gains from the ability to purchase online.

### Annex highlights and key summary statistics
- Annex Figure 1: Distribution of consumption categories by payment method using COICOP and MCC/NACE classifications; "Online cards" are transactions where cardholder and card are not physically present (internet, telephone, mail-order).
- Annex Figure 2: Aggregate trends in E-commerce shares and Oxford Stringency index; e-commerce shares defined as economy-wide online expenditures divided by total consumption expenditure through all means of payments.
- Annex Figure 3: Aggregate E-commerce share in Spain (2017–24) measured as number of online transactions over total transactions (3-month moving average).
- Annex Figure 5: Relationship between pre-Covid e-commerce share and increase during the pandemic; defines extensive margin = new online clients (0-1 transactions pre-pandemic and >=3 post), with difference between pre- (2017m1-2020m2) and post- (2020m3-2022m12) average e-commerce shares plotted.
- Annex Table 1 (financial transaction data by means of payments):
  - Total transactions: 1.83 (millions), 110.55 (millions, euros).
  - Card transactions: 1.28 (millions), 44.05 (millions, euros).
  - Offline cards: 1.06 (millions), 33.44 (millions, euros).
  - Online cards: 0.22 (millions), 10.61 (millions, euros).
  - Cash: 0.16 (millions), 23.16 (millions, euros).
  - Transfers and direct debits: 0.38 (millions), 43.34 (millions, euros).
- Annex Table 2 (selected summary statistics):
  - Full sample N = 473,315:
    - Online spending % (transactions) mean = 0.0820, sd = 0.242, min = 0, max = 1.
    - Online spending % (amount spent) mean = 0.0783, sd = 0.243.
    - Cash % (amount spent) mean = 0.182, sd = 0.217, max = 0.998.
    - Age mean = 55.51, sd = 14.73, min = 22, max = 103.
    - Extensive margin client mean = 0.0458.
    - Avg. online spending % in the category pre-pandemic (amount) mean = 0.0689, sd = 0.0825, max = 0.262.
    - Hybrid Pandemic (100 before onset, 100+ afterwards based on Residential Mobility) mean = 102.7, sd = 5.133, min = 96.03, max = 133.5.
  - Post-pandemic sample N = 225,214:
    - Online spending % (transactions) mean = 0.0920, sd = 0.254.
    - Online spending % (amount spent) mean = 0.0880, sd = 0.256.
    - Cash % (amount spent) mean = 0.158, sd = 0.209.
    - Age mean = 55.39, sd = 14.71.
    - Mobility Restriction (Residential) mean = 105.6, sd = 6.232, min = 96.03, max = 133.5.
    - Extensive margin client mean = 0.0485.
    - Avg. online spending % in the category pre-pandemic (amount) mean = 0.0689, sd = 0.0828.
    - Hybrid Pandemic mean = 105.6, sd = 6.232.

*Source: IMF Working Paper — E-commerce during Covid in Spain: One “Click” does not fit All (printed PDF content provided).*

### Annex Table 4. Extensive margin Clients

### Annex Table 4. Extensive margin Clients

### Transactions (2017 - 2019)
- Frequency
  - 0: 209
  - 1: 63
  - Total: 272

### Transactions (2020 - 2022)
- Frequency, Percent, Extensive Margin
  - 0: 144, 52.94, No
  - 1: 47, 17.28, No
  - 2: 19, 6.99, No
  - 3: 10, 3.68, Yes
  - 4: 6, 2.21, Yes
  - 5: 3, 1.10, Yes
  - 6: 4, 1.47, Yes
  - 7: 4, 1.47, Yes
  - 8: 1, 0.37, Yes
  - 9: 2, 0.74, Yes
  - 10: 4, 1.47, Yes
  - 11: 3, 1.10, Yes
  - 12: 1, 0.37, Yes
  - 13: 1, 0.37, Yes
  - 15: 3, 1.10, Yes
  - 16: 2, 0.74, Yes
  - 17: 5, 1.84, Yes
  - 21: 1, 0.37, Yes
  - 22: 2, 0.74, Yes
  - 25: 1, 0.37, Yes
  - 26: 2, 0.74, Yes
  - 27: 1, 0.37, Yes
  - 28: 1, 0.37, Yes
  - 31: 1, 0.37, Yes
  - 32: 1, 0.37, Yes
  - 33: 1, 0.37, Yes
  - 39: 1, 0.37, Yes
  - 55: 1, 0.37, Yes
  - Total: 272, 100

### Annex Table 5. Online spending shares, macroeconomic characteristics: Mobility extended pre-pandemic

- Dependent variable: Online spending % (amount)
- Key coefficients (columns labeled (1) to (5); standard errors in parentheses)
  - Male = 1:
    - (1): -0.0077* (0.004)
    - (2): -0.0100** (0.004)
    - (4): 0.0553* (0.028)
  - Urban = 1:
    - (1): -0.0049 (0.005)
    - (2): -0.0007 (0.006)
    - (4): 0.0536 (0.033)
  - Age:
    - (1): -0.0020*** (0.000)
    - (2): -0.0022*** (0.000)
  - Young (Dummy if Age < percentile 25) = 1:
    - (2): -0.0637* (0.038)
  - Avg. sector online share pre-pandemic:
    - (1): 1.0527*** (0.032)
  - Mobility Restriction (Residential):
    - (1): 0.0022*** (0.000)
    - (2): 0.0010* (0.001)
    - (3): 0.0009 (0.001)
    - (4): 0.0002 (0.001)
    - (5): 0.0000 (0.001)
  - Male*Residential mobility:
    - (2): -0.0007** (0.000)
    - (3): -0.0008*** (0.000)
  - Urban*Residential mobility:
    - (2): -0.0005 (0.000)
    - (3): -0.0002 (0.000)
  - Young*Residential mobility:
    - (1): 0.0011*** (0.000)
    - (2): 0.0013*** (0.000)
  - Residential mobility * Avg. sector online share pre-pandemic:
    - (2): 0.0137*** (0.002)
    - (3): 0.0135*** (0.002)

- Sample, fit, and fixed effects
  - Observations: 473,315 (for all columns)
  - R-squared: 0.151, 0.162, 0.236, 0.154, 0.237 (columns (1) to (5))
  - Sector FE: No, Yes, Yes, Yes, Yes
  - Province FE: No, Yes, Yes, Yes, Yes
  - Time FE: No, Yes, Yes, Yes, Yes
  - Client FE: No, No, Yes, No, Yes
  - Standard Errors: Cluster at client
  - Period: 2017m1 to 2022m12
  - No. Clients: 1000 (for each column)
  - No. Provinces: 51
  - No. Categories: 19
  - No. Months: 72
  - Average Dep. Var.: 0.0783
  - Average Pandemic. Var.: 102.7
  - Average Male: 0.526
  - Average Urban: 0.203
  - Average Age: 55.51
  - Average Cash % of amount: 0.182

### Annex Table 6. Online spending shares, macroeconomic characteristics: (Pandemic dummy and restrictions)

- Dependent variable: Online spending % (amount)
- Key coefficients (columns (1) to (3); standard errors in parentheses)
  - Male = 1:
    - (1): -0.0078* (0.004)
    - (2): 0.0789** (0.032)
  - Urban = 1:
    - (1): -0.0049 (0.005)
    - (2): 0.0417 (0.039)
  - Age:
    - (1): -0.0020*** (0.000)
  - Young (Dummy if Age < percentile 25) = 1:
    - (1): -0.0387 (0.042)
  - Avg. online spending % pre-pandemic (amount):
    - (1): 1.0527*** (0.032)
  - Pandemic = 1 (March 2020 onwards):
    - (1): 0.0089*** (0.002)
  - Mobility Restriction (Residential):
    - (1): 0.0018*** (0.000)
    - (2): -0.0043 (0.003)
    - (3): -0.0029 (0.003)
  - Pandemic*Male:
    - (2): 0.0049 (0.004)
    - (3): 0.0069** (0.003)
  - Pandemic*Urban:
    - (2): -0.0025 (0.004)
    - (3): -0.0035 (0.004)
  - Pandemic*Young:
    - (2): 0.0052 (0.005)
    - (3): 0.0045 (0.005)
  - Pandemic* Avg. sector online share pre-pandemic:
    - (2): 0.1052*** (0.029)
    - (3): 0.1151*** (0.029)
  - Male*Mobility Restriction (Residential):
    - (2): -0.0009*** (0.000)
    - (3): -0.0012*** (0.000)
  - Urban*Mobility Restriction (Residential):
    - (2): -0.0004 (0.000)
    - (3): -0.0001 (0.000)
  - Young*Mobility Restriction (Residential):
    - (2): 0.0009** (0.000)
    - (3): 0.0011*** (0.000)
  - Res Mobility* Avg. sector online share pre-pandemic:
    - (2): 0.0084*** (0.002)
    - (3): 0.0077*** (0.002)
  - Pandemic*Mobility Restriction (Residential):
    - (2): 0.0051* (0.003)
    - (3): 0.0036 (0.003)

- Sample, fit, and fixed effects
  - Observations: 473,315 (for all columns)
  - R-squared: 0.151, 0.154, 0.237
  - Sector FE: No, Yes, Yes
  - Province FE: No, Yes, Yes
  - Time FE: No, Yes, Yes
  - Client FE: No, No, Yes
  - Standard Errors: Cluster at client
  - Period: 2017m1 to 2022m12
  - No. Clients: 1000
  - No. Provinces: 51
  - No. Categories: 19
  - No. Months: 72
  - Average Dep. Var.: 0.0783
  - Average Pandemic. Var.: 102.7
  - Average Male: 0.526
  - Average Urban: 0.203
  - Average Age: 55.51
  - Average Cash % of amount: 0.182

### Annex Table 7. Online spending shares, macroeconomic characteristics: Additional robustness

- Variables and selected coefficients across columns (1) to (10) (starred significance preserved; omitted cells reflect table layout)
  - Mobility Restriction (Residential):
    - (1): 0.0021*** 
    - (2): 0.0020*** 
    - (3): 0.0020*** 
    - (4): 0.0020***
    - (5) through (10): -0.001, -0.001, -0.001, -0.001 (as shown)
  - Male = 1:
    - (1): -0.0101**
    - (2): 0.0916***
    - (3): -0.0127***
    - (4): 0.0877***
    - (5): -0.0095**
    - (6) to (10): -0.004, -0.033, -0.004, -0.033, -0.004 (as shown)
  - Urban = 1:
    - (1): 0.0051
    - (2): 0.024
    - (3): 0.0011
    - (4): 0.0283
    - (5): 0.0039
    - (6) to (10): -0.006, -0.041, -0.006, -0.04, -0.006 (as shown)
  - Young (Dummy if Age < percentile 25) = 1:
    - (1): 0.0470***
    - (2): -0.0471
    - (3): 0.0480***
    - (4): -0.0394
    - (5): 0.0473***
    - (6) to (10): -0.006, -0.043, -0.006, -0.043, -0.006 (as shown)
  - Male*Mobility Restriction (Residential):
    - (1): -0.0010***
    - (2): -0.0012***
    - (3): -0.0009***
    - (4): -0.0012***
  - Urban*Mobility Restriction (Residential):
    - (1): -0.0002
    - (2): 0.0001
    - (3): -0.0003
    - (4): 0.0001
  - Young*Mobility Restriction (Residential):
    - (1): 0.0010**
    - (2): 0.0014***
    - (3): 0.0009**
    - (4): 0.0011***
  - Mobility * Avg. sector online share pre-pandemic (# transactions):
    - (1): 0.0069***
    - (2): 0.0063***
    - (5) to (10): -0.002, -0.002 (as shown)
  - Pandemic*Male, Pandemic*Urban, Pandemic*Young:
    - Pandemic*Young shows positive and sometimes significant effects in several columns, e.g., (1): 0.0077*; (2): 0.0083*; (3): 0.0080*; (4): 0.0098**; (5): 0.0110***; (6): 0.0114***
  - Pandemic * Avg. sector online share pre-pandemic:
    - (1): 0.1715***
    - (2): 0.1761***
    - (3): 0.1642***
    - (4): 0.1642***
    - (5) to (8): -0.023, -0.022, -0.018, -0.018 (as shown)
  - Cash % (amount spent) and interactions:
    - Cash % (amount spent) coefficients vary across columns (e.g., (1): 0.0036; (2): 0.0008; (3): -0.0083; (4): 0.0781)
    - Pandemic*Share of cash in payments: (1): -0.0406***; (2): -0.0138**
  - Res mob*Sector online share pre-pandemic:
    - (1): 0.0077***
    - (2): 0.0070***
  - Res mob*Share of cash:
    - (2): -0.0008*

- Sample sizes, fit, and fixed effects (columns (1) to (10))
  - Observations: 473,315; 473,315; 225,214; 225,214; 473,315; 473,315; 225,214; 225,214; 473,315; 473,315
  - R-squared: 0.192, 0.278, 0.21, 0.308, 0.155, 0.237, 0.167, 0.263, 0.192, 0.277
  - Sector FE: Yes (all columns)
  - Province FE: Yes (all columns)
  - Time FE: Yes (all columns)
  - Client FE: No, Yes, No, Yes, No, Yes, No, Yes, No, Yes (alternating as shown)
  - Standard Errors: Cluster at client
  - Periods:
    - Columns (1), (2), (5), (6), (9), (10): 2017m1 to 2022m12
    - Columns (3), (4), (7), (8): 2020m2 to 2022m12
  - No. Clients: 1000 (all columns)
  - No. Provinces: 51 (all columns)
  - No. Categories: 19 (all columns)
  - No. Months: 72 for columns covering 2017m1 to 2022m12; 33 for columns covering 2020m2 to 2022m12
  - Average Dep. Var.: values shown per column (e.g., 0.082, 0.082, 0.092, 0.092, 0.0783, 0.0783, 0.088, 0.088, 0.082, 0.082)
  - Average Pandemic. Var.: values shown per column (e.g., 0.493, 0.493, 105.6, 105.6, 0.493, 0.493, 105.6, 105.6, 0.314, 0.314)
  - Average Male: 0.526 (all columns)
  - Average Urban: 0.203 or 0.204 depending on column (as shown)
  - Average Age: 55.51 or 55.39 depending on column (as shown)
  - Average Cash % of txs: 0.0908, 0.0908, 0.0687, 0.0687, 0.0908, 0.0908 (as shown)
  - Average Cash % of amount: 0.182, 0.182, 0.158, 0.158 (as shown)

*IMF WORKING PAPERS — E-commerce during Covid in Spain: One “Click” does not fit All — Annex Tables 4–7 (extracted tables and regression results contained in the source PDF).*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2024/english/wpiea2024107-print-pdf.pdf_