1# RECORD a34939ebd1d20ba654297f426501ad47056b65a42# TITLE: Adjusting to the New Normal: River Flood Risk and the Real Estate Market3# YEAR: 2025.0 VENUE: Environmental and Resource Economics DOI: 10.1007/s10640-025-00995-x4# AUTHORS: Henri Gruhl; Johannes Brehm; Arijit Ghosh; P. Breidenbach; Nils Aus Dem Moore5# PAGES: 26 CANDIDATE TABLE PAGES (1-based): [10, 13, 15, 21]6# ABSTRACT: Do individuals learn from salient river floods and update their risk perceptions accordingly? We exploit the Western European flood in 2021 to estimate changes in property values across German flood risk regions. Flood-prone areas in proximity to the flood or with higher climate change beliefs experience housing price declines, while we observe no statistically significant effects in other flood risk zones. Our results imply that flood risk discounts in the housing market will not adjust uniformly, which may result in suboptimal capital allocation in areas at risk of future flooding.789===== PAGE 1 (score 6, numeric-paren 0, table-words 0, var-words 6) =====1011Accepted: 7 May 2025 / Published online: 18 July 202512© The Author(s) 202513 14 Henri Gruhl15henri.gruhl@rwi-essen.de16 17 Johannes Brehm18johannes.brehm@rwi-essen.de191 20RWI - Leibniz Institute for Economic Research, Vrije Universiteit Amsterdam, Berlin, 21Germany222 23RWI - Leibniz Institute for Economic Research, Hertie School, Berlin, Germany243 25RWI - Leibniz Institute for Economic Research, Essen, Germany264 27Hertie School, Berlin, Germany28Adjusting to the New Normal: River Flood Risk and the Real 29Estate Market30Henri Gruhl131 · Johannes Brehm232 · Arijit Ghosh3 · Philipp Breidenbach3 · 33Nils aus dem Moore434Environmental and Resource Economics (2025) 88:2331–235635https://doi.org/10.1007/s10640-025-00995-x36Abstract37Do individuals learn from salient river floods and update their risk perceptions accord38ingly? We exploit the Western European flood in 2021 to estimate changes in property 39values across German flood risk regions. Flood-prone areas in proximity to the flood or 40with higher climate change beliefs experience housing price declines, while we observe no 41statistically significant effects in other flood risk zones. Our results imply that flood risk 42discounts in the housing market will not adjust uniformly, which may result in suboptimal 43capital allocation in areas at risk of future flooding.44Keywords Flood risk · House prices · Risk updating45JEL Codes Q54 · Q51 · D81 · R31461 Introduction47If the global mean temperature increases by 1.5◦C compared to pre-industrial temperatures, 48global human fatalities from river flooding are estimated to rise by 70–83% and flood dam49ages by 160–240% without further adaptation (Dottori et al. 2018). According to the most 50optimistic IPCC pathway, it is highly likely that average global warming will cross the 51threshold of 1.5◦C, with temperatures likely exceeding this level by 2100 (IPCC Working 521 3535455===== PAGE 10 (score 28, numeric-paren 8, table-words 3, var-words 9) =====5657H. Gruhl et al.58cellar, and the conditions of the structure (categorical)11. We include separate indicators for 59characteristics with missing information. Zip code fixed effects capture all unobserved zip 60code-specific characteristics that are invariant over time. The average zip code in Germany 61covers 43.2 km2. Following Abbott and Klaiber (2011), in addition to the zip code fixed 62effects, we include several different spatial fixed effects in our main estimations at the larger 63county12 level or the smaller 1 km2 level. While fine-scale fixed effects effectively control 64for unobserved spatial heterogeneity, they may discard valuable between-variation, lead65ing to incomplete estimates of broader capitalization effects (Abbott and Klaiber 2011). In 66contrast, coarse-scale fixed effects retain between-variation but risk omitted variable bias by 67failing to account for fine-scale heterogeneity. The inclusion of varying spatial fixed effects 68allows us to assess the sensitivity of the results, yielding information about capitalization of 69public goods and amenities at varying spacial scales. The year-month fixed effects control 70per-square-meter values than larger ones, especially in urban settings.7111 The categories for the condition of the house range from “simple” over “normal” and “sophisticated” to 72“deluxe”.7312 Counties in Germany correspond to the NUTS-3 level and are called “Landkreise” or “Kreisfreie Städte”. 74There are 400 counties in Germany.75Table 1 Sample statistics76Inside flood risk zones77Outside flood risk zones78∆in−out79mean80sd81min82max83mean84sd85min86max87ln(Price per m2)887.404890.624905.487918.865927.483930.603945.486958.86596−0.079∗∗∗97(0.017)98Price (thousand €)99331,7100238,210120,51022,759103348,1104244,5105211062,780107−16,4∗∗∗108(5,518)109Living space (m2)110178.0311190.2811255.67113767114172.7411580.9711655.741177681185.29∗∗∗119(1.44)120Construction year1211961122561231500124202012519721264612715001282020129−11∗∗∗130(1)131Number of rooms1325.4581331.7221341135101365.4221371.6411381139101400.036141(0.024)142Cellar1430.3341440.472145014611470.3351480.47214901501151−0.000152(0.009)153Property facilities:154”simple” 0.0311550.173156015711580.0251590.156160016111620.006∗∗∗163(0.001)164”normal” 0.1371650.344166016711680.1221690.327170017111720.015∗∗∗173(0.003)174”sophisticated” 0.2941750.456176017711780.3371790.47318001811182−0.044∗∗∗183(0.007)184”deluxe” 0.0151850.122186018711880.0171890.13119001911192−0.002∗∗193(0.001)194Number of obs.195232,3901961,234,7701971,467,160198Notes: This table summarizes the characteristics of the houses in the final sample, i.e. houses inside flood-199risk zones (treatment group) and within 0.5–3 km of flood-risk zones (control group). The columns show 200the mean, standard deviation, minimum value, and maximum value respectively for houses within and 201outside of flood risk zones. The last column reports the difference between the means of the two groups, 202clustered at the zip code level. * p < 0.1, ** p < 0.05, *** p < 0.012031 32042340205206207===== PAGE 11 (score 7, numeric-paren 0, table-words 0, var-words 7) =====208209Adjusting to the New Normal: River Flood Risk and the Real Estate Market210for all unobserved time-specific effects in a specific month and year, which are the same 211for all included regions in Germany, such as a general upward or downward shift in the 212German housing market. Furthermore, we also show specifications in which we control for 213county-specific time trends or relax the temporal fixed effects by only including year and 214month fixed effects, capturing temporal general changes between years and general seasonal 215changes between months. Additionally, we present results when we include county-specific 216quadratic trends, cubic trends, or county-by-year fixed effects. Standard errors are clustered 217at the zip code level.218The identifying assumption in the canonical difference-in-differences framework is the 219parallel trends assumption. By comparing changes in outcomes over time between a treat220ment group (houses in flood risk zones) and a control group (houses outside flood risk 221zones but in close proximity), DiD designs assume that, in the absence of treatment, both 222groups would have developed similarly over time (parallel trends). Under this assumption, 223the effect is identified even if shocks affect the potential outcome as long as they are not 224correlated with the flood risk zones. Although it is impossible to test this assumption con225clusively, it is common practice to perform an event study to show that there is no statisti226cally significant effect on property prices between the groups in the period before the flood 227occurred. Methodologically, this concept can be expressed as follows:228 229log Yi,z,t =230t−1231∑232s = t233αs(flood riski × Posts) +234¯t235∑236s = t237αs(flood riski × Posts)238 239+β flood riski + γ Xi + λz + ϕt + εi,z,t,240(2)241where t is the first pre-treatment period for which we want to rule out anticipation effects, 242while ¯t represents the last post-treatment period for which we expect adjustment effects. We 243take α−1 as the reference category. Therefore, the period-specific effects in the first sum244mation (anticipatory effects) and the second summation (reactive effects) are interpreted 245relative to the period t −1. This approach enables us to assess whether there are any statisti246cally significant changes in property prices between the treatment and control groups in the 247periods leading up to and following the flood event, and to quantify the magnitude of these 248effects over time.249Figure 3 displays event studies based on Eq. (2). We do not detect any statistically sig250nificant pre-trends in the quarters leading up to the flood. We also do not find longer-term 251systematic pre-trends in the years before the flooding. The event studies support the parallel 252trends assumption, indicating that the potential outcomes for the treated and control groups 253follow parallel trends when conditioned on the specified model. This further strengthens our 254analysis, as it suggests that any changes in the outcomes after the flood event can be attrib255uted to the flood and not to other external factors.256For identification, we rely on intertemporal variation in prices. However, as noted by 257Kuminoff and Pope (2014), capitalization may not precisely reflect households’ willingness 258to pay if the gradient of the equilibrium price function shifts following the flood. Specifi259cally, if the flood changes residential sorting in flood-risk areas, altering the hedonic gradi260ent, panel variation may yield biased estimates of risk valuation updating. Therefore, our 2611 32622341263264265===== PAGE 13 (score 25, numeric-paren 5, table-words 5, var-words 9) =====266267Adjusting to the New Normal: River Flood Risk and the Real Estate Market268approach assumes a stable hedonic gradient over time. Since this assumption is difficult to 269test, the estimated effects should be interpreted with caution in terms of welfare.2704 Results2714.1 Main Results272Our analysis shows that after the flood, risk perceptions are updated in some regions of 273Germany while remaining unchanged in others. Specifically, we observe localized effects 274in regions near the flooding that are statistically significant at the one percent level, which 275diminish with distance, and significant effects at the five percent level in regions with higher 276baseline beliefs in climate change.277Table 2 presents the average effect of the flood on house prices in flood risk regions that 278were not directly affected. We do not detect any statistically significant effects at the one, 279five or ten percent level across model specifications with varying restrictiveness. The result 280indicates that, on average, the flood did not lead to a statistically significant change in flood 281risk valuation across German flood risk zones. We run our model using five specifications. 282Column 1 estimates the effect according to Eq. (1). In Column 2, we loosen the restrictive283ness of the model by removing the year-month fixed effect and only include year fixed 284effects for unobserved differences between years and month fixed effects to control for sea285sonality across years. Similarly, Column 3 allows for more spatial variation by expanding 286the spatial fixed effect to the county level. In Column 4, we restrict the model by including 287a 1 km2 fixed effect, capturing more unobserved spatial differences, but in turn reducing 288variation in the estimation of the effect. Finally, Column 5 includes linear time trends at the 289county level to account for different trends between regions.290Table 2 Average effect291Dependent variable: log (price per m2)292(1)293(2)294(3)295(4)296(5)297Treatment effect298-0.0042299-0.00423000.0035301-0.0066302-0.0071303(0.0098)304(0.0098)305(0.0104)306(0.0083)307(0.0100)308Treatment group dummy309✓310✓311✓312✓313✓314Post-treatment dummy315✓316✓317✓318✓319✓320Property characteristics321✓322✓323✓324✓325✓326County×year linear trends327✓328Year fixed effects329✓330✓331✓332✓333✓334Month fixed effects335✓336✓337✓338✓339✓340Year-Month fixed effects341✓342✓343✓344✓345Zipcode fixed effects346✓347✓348✓349County fixed effects350✓3511km2 fixed effects352✓353Number of observations3541,466,5093551,466,5093561,468,5673571,457,4763581,466,509359Notes: This table presents the average results following Eq. (1). Property characteristics include the 360construction year of the house (categorical), the size of the living space, the number of floors, the number 361of rooms, the presence of a cellar, and the conditions of the structure. Standard errors are clustered at the 362zip code level and shown in parentheses. * p < 0.1, ** p < 0.05, *** p < 0.013631 33642343365366367===== PAGE 14 (score 11, numeric-paren 0, table-words 2, var-words 7) =====368369H. Gruhl et al.370While the average effect is a null result, statistically and economically significant effects 371may be averaged out across heterogeneous groups. For example, economic theory and 372empirical evidence from the behavioral economics literature suggest that the distance to 373an event may play a role in how economic agents react (Gallagher 2014; Alok et al. 2020). 374Moreover, empirical evidence – primarily from the United States – suggests that climate 375change belief heterogeneities could play a role in how flood risk information, e.g., induced 376by a natural disaster, is processed (Barrage and Furst 2019; Bernstein et al. 2019; Baldauf et 377al. 2020; Bakkensen and Barrage 2022).378To assess the role of proximity, we interact the treatment effect with a continuous mea379sure of the distance of a property to the closest area that was flooded in July 2021. Our find380ings, presented in Panel A of Table 3, show an economically meaningful negative effect of 381the flood on nearby houses across various specifications that is statistically significant at the 382one percent level (Column 5). The statistically significant (one percent level) positive inter383action term indicates that the discount on house prices decreases as houses are located far384ther away from the flood. Specifically, the square meter price of at-risk houses immediately 385next to the flood decreases by about 5.9% (Column 5) compared to houses outside flood 386risk zones. The effect then decreases by approximately 0.03% points with each kilometer 387distance to the flood. While the direct comparison of effect sizes from previous studies is 388challenging due to differences in research settings, designs, and flooding types, our findings 389align most closely with Ortega and Taṣpınar (2018). The authors document an 8% price 390penalty for flood zone properties not directly damaged by Hurricane Sandy. Hallstrom and 391Smith (2005) find that Hurricane Andrew conveyed new risk information, causing property 392prices in high-risk storm zones to decline by at least 19%. Gibson and Mullins (2020) report 393an 11% price reduction for properties newly designated as being in a flood zone after nar394rowly escaping damage from Hurricane Sandy, attributing this reduction to belief updating. 395These findings are consistent with our proposed learning mechanism increasing perceived 396flood risk.397Since distance can be categorized in various ways with varying assumptions about the 398functional form, we employ various alternative measures to test the robustness of our main 399finding. Figure B.1 illustrates the distribution of distances between properties listed for sale 400and the flooded area. We observe a large concentration of properties near the flooded areas, 401which gradually declines with distance, a pattern driven by the high population and property 402density in the state of North-Rhine Westphalia. To avoid relying on a linear functional form 403assumption, we employ alternative distance bins rather than a continuous measure. This 404approach effectively captures potential non-linearities and thresholds in the relationship. 405Table B.1 displays the results relying on the median split, which creates two distance bins, 406allowing us to estimate effects independently for properties above and below the median 407distance to the flooded areas, without imposing any functional form. The result holds: we 408find a statistically significant negative effect for properties geographically close to the flood 409at the one percent level, but no significant effect for properties located above the median 410distance (with α = 0.05). Further, Fig. 4 displays the event studies for the median split-411based distance heterogeneity, which show no systematic pre-trends. Figure B.2 displays 412the event studies for 50 kilometer distance bins. Due to decreasing precision from smaller 413sample sizes in each distance bin, we refrain from interpreting the precise location where the 414distance-based effect reaches zero. Nonetheless, Fig. B.2 displays a pattern where the effect 415size is largest for those properties listed for sale closest to the flood and then tapering off. 4161 34172344418419420===== PAGE 15 (score 82, numeric-paren 51, table-words 6, var-words 13) =====421422Adjusting to the New Normal: River Flood Risk and the Real Estate Market423Smaller bins generally introduce more noise due to fewer observations, this is particularly 424the case for the distance bins far from the flooded area (see Figs. B.1 and B.2). Conse425quently, we exclude regions that are more than 400 kilometers away from the flooded areas 426and re-estimate the main distance result. Table B.2 shows that the results remain statistically 427significant at the one percent level and are slightly smaller than the main estimates. Lastly, 428we take inspiration from Callaway et al. (2024) and treat the distance-based effect as a 429continuous treatment. Comparing houses in flood risk areas relatively close to the flood to 430flood risk areas far from the flood does not display significant pre-trends (with α = 0.05) 431and identifies an imprecise effect of around ten percent (Fig. B.3).432These distance-based results may be attributed to greater amount of information about 433the flood in regions closer to the inundated areas, and cognitive biases. Geographically 434closer regions likely experienced more media coverage of the event through local news 435Table 3 Heterogeneous effects: Distance to flooded areas and belief in climate change436Dependent variable: log (price per m2)437(1)438(2)439(3)440(4)441(5)442A: Distance heterogeneity443Treatment effect444−0.0592***445−0.0589***446−0.0361**447−0.0421***448−0.0585***449(0.0165)450(0.0165)451(0.0179)452(0.0127)453(0.0166)454Treatment effect × distance to 455flooded area (km)4560.0003***457(0.0001)4580.0003***459(0.0001)4600.0002***461(0.0001)4620.0002***463(0.0001)4640.0003***465(0.0001)466Distance to flooded area (km)467−0.0006**468−0.0006**469−0.0029***470−0.0012471−0.0030***472(0.0002)473(0.0002)474(0.0005)475(0.0045)476(0.0007)477B: Climate change belief 478heterogeneity479Treatment effect4800.01344810.01304820.01654830.00294840.0088485(0.0122)486(0.0122)487(0.0131)488(0.0109)489(0.0129)490Treatment effect × high climate 491change beliefs492−0.0422**493(0.0178)494−0.0410**495(0.0179)496−0.0313*497(0.0187)498−0.0218499(0.0149)500−0.0377**501(0.0180)502High climate change beliefs5030.0252*5040.0254*5050.0915***506−0.01255070.0227508(0.0143)509(0.0144)510(0.0204)511(0.0148)512(0.0150)513Treatment group dummy514✓515✓516✓517✓518✓519Post-treatment dummy520✓521✓522✓523✓524✓525Property characteristics526✓527✓528✓529✓530✓531County×year linear trends532✓533Year fixed effects534✓535✓536✓537✓538✓539Month fixed effects540✓541✓542✓543✓544✓545Year-Month fixed effects546✓547✓548✓549✓550Zipcode fixed effects551✓552✓553✓554County fixed effects555✓5561km2 fixed effects557✓558Number of observations5591,466,5095601,466,5095611,468,5675621,457,4765631,466,509564Notes: This table presents the heterogeneous effects of the flood. Panel A interacts the treatment effect 565with the continuous distance of a house to the nearest potentially flooded area. Panel B interacts the 566treatment effect with an indicator for counties that have an above-median share of the population that 567believes in human-made climate change. Property characteristics include the construction year of the 568house (categorical), the size of the living space, the number of floors, the number of rooms, the presence of 569a cellar, and the conditions of the structure. Standard errors are clustered at the zip code level and shown 570in parentheses. * p < 0.1, ** p < 0.05, *** p < 0.015711 35722345573574575===== PAGE 16 (score 8, numeric-paren 0, table-words 2, var-words 4) =====576577H. Gruhl et al.578Fig. 4 Event studies of the heterogeneous effects Notes: the figure presents event study coefficients and 57995% confidence intervals using Eq. (2) and specification (5) from Table 3 that show how flood risk af580fects the house prices above and below the median of the distance to the flooded area (a) and the regional 581believe in climate change (b) before and after the Western European flood in July 2021. The coefficient 582in the quarter prior to the flood is normalized to zero. Standard errors are clustered at the zip code level583 5841 35852346586587588===== PAGE 21 (score 82, numeric-paren 58, table-words 3, var-words 10) =====589590Adjusting to the New Normal: River Flood Risk and the Real Estate Market591Table 4 Robustness checks592Dependent variable: log (price per m2)593Spillovers594Control Groups595Exclusions of flooded areas596Spatial trends5971 km buffer5981.5 km buffer Distance to 599nearest treated 600house ≤3km601Distance to 602flood risk 603area ≤2km604Only certainly 605flooded area606Poten607tially flooded 608counties609Placebo 610flood July 6112019612Placebo 613flood July 6142020615(1)616(2)617(3)618(4)619(5)620(6)621(7)622(8)623Average effect624Treatment effect625−0.0063626−0.01566270.0026628−0.0021629−0.0053630−0.00806310.00476320.0049633(0.0103)634(0.0106)635(0.0103)636(0.0102)637(0.0107)638(0.0097)639(0.0071)640(0.0074)641Distance heterogeneity642Treatment effect643−0.0629***644−0.0729***645−0.0428***646−0.0503***647−0.0690***648−0.0734***649−0.0070650−0.0092651(0.0168)652(0.0156)653(0.0166)654(0.0165)655(0.0197)656(0.0163)657(0.0108)658(0.0112)659Treatment effect × distance to flooded area (km)6600.0003***6610.0004***6620.0003***6630.0003***6640.0004***6650.0003***6660.00016670.0001668(0.0001)669(0.0001)670(0.0001)671(0.0001)672(0.0001)673(0.0001)674(0.0001)675(0.0001)676Climate change belief heterogeneity677Treatment effect6780.01086790.00646800.01526810.01256820.01416830.01416840.00296850.0068686(0.0129)687(0.0134)688(0.0131)689(0.0132)690(0.0137)691(0.0122)692(0.0096)693(0.0101)694Treatment effect × high climate change beliefs695−0.0401**696−0.0490***697−0.0297*698−0.0341*699−0.0460**700−0.0503***7010.0042702−0.0044703(0.0179)704(0.0172)705(0.0179)706(0.0181)707(0.0192)708(0.0172)709(0.0123)710(0.0127)711Full set of controls and FEs712✓713✓714✓715✓716✓717✓718✓719✓720N7211,050,981722831,3957231,024,5957241,125,2857251,263,9857261,559,1327271,466,5097281,466,509729Notes: This table presents the robustness checks for the results in Tables 2 and 3. All estimations include a treatment group dummy, a post-treatment dummy, property 730characteristics, annual spatial trends, year-month fixed effects, and zip-code fixed effects. Regressions for distance to the flood heterogeneity include the uninteracted distance 731to the flooded area. Regressions for the belief in climate change heterogeneity include the uninteracted indicator for high climate change belief. Property characteristics 732include the construction year of the house (categorical), the size of the living space, number of floors, number of rooms, presence of a cellar, and conditions of the structure. 733Spatial×year linear trends are at the county level, except for Column 5, where we instead include state-level trends for computational reasons. Standard errors are clustered 734at the zip code level and are shown in parentheses. * p < 0.1, ** p < 0.05, *** p < 0.017351 37362351737738739===== PAGE 22 (score 5, numeric-paren 0, table-words 0, var-words 5) =====740741H. Gruhl et al.742The average effect of the flood remains consistent with our main findings, and the heteroge743neous effects by distance are also robust. Specifically, we estimate a smaller but statistically 744significant (at the five percent level) negative effect of approximately 2.6% for houses closer 745to the flood, which decreases, statistically significant at the ten percent level, for houses 746farther away. However, the climate change belief-based heterogeneous effects are not robust 747in the repeat-listings approach, as we find no statistically significant effects or interactions 748(α = 0.05). These results indicate that differences related to climate change beliefs warrant 749further investigation across various contexts and models to validate the indicative evidence.750Finally, we aim to ensure that our results are not sensitive to different time trends across 751regions in Germany. Although we control for county-specific time trends in the regressions, 752we confirm that we do not pick up any remaining regional differences by conducting pla753cebo tests in time. In Columns (7) and (8), we estimate the effects of hypothetical floods that 754occurred one and two years before the actual flood, respectively. If our estimation strategy is 755biased by differences in time trends across space, we would observe (heterogeneous) treat756ment effects in these tests. However, we find no effects in either placebo exercise, indicating 757that our main results are not picking up on general regional developments.7585 Conclusion759This study investigates the impact of a significant fluvial flooding event on flood risk valua760tion in the German real estate market. Specifically, we focus on the Western European flood 761in July 2021. Using a hedonic pricing model and a difference-in-differences approach, we 762explore the changes in flood risk valuation across different flood risk areas in Germany.763By excluding physical flood damages using satellite imagery, we disentangle changes in 764flood risk valuation from direct flood damages and find that, on average, flood risk valuation 765across flood-prone areas in Germany did not change. However, the average effect across 766the entire country masks substantial regional heterogeneity. First, house prices declined by 767approximately 5.9% in flood risk areas near inundated areas, with the effect diminishing as 768distance from the flooding increases. Second, house prices decrease by approximately 3% 769in flood-prone regions, with a relatively higher share of people believing in climate change. 770Regions with below-median belief in climate change show no statistically significant change 771in flood risk valuation. Our findings highlight the importance of considering regional differ772ences in the flood risk valuation updating following fluvial floods. These findings suggest 773that flood risk undervaluation may persist despite increasing fluvial flood frequency, par774ticularly in regions distant from past floods or with lower belief in human-induced climate 775change. This heterogeneity in flood risk perceptions may lead to suboptimal capital alloca776tion in regions at risk of future flooding (Harrison and Kreps 1978; Scheinkman and Xiong 7772003; Bakkensen and Barrage 2022).778The lack of change in flood risk valuation in distant regions or areas with below-median 779climate change beliefs may necessitate government intervention. The policy implications 780hinge on diverging perspectives of who is responsible to adapt to future climate change: the 781government or the market. The government-centric perspective argues that governments 782should act by implementing protective measures against fluvial flash floods, enforcing land-783use and zoning restrictions to discourage development in flood-prone areas, or mandat784ing elementary insurance for at-risk properties. In contrast, the market-centric perspective 7851 37862352787