Will Climate Change Increase Inequality in Wage Growth? -Evidence from China's 30 Provincial Capitals

Examining the impact of long-term temperature rise on wages is 7 not only beneﬁcial for gaining insight into the economic consequences of cli- 8 mate change, but also of great reference value to reduce income disparity and 9 alleviate relative poverty. Based on the panel data of 30 provincial capitals 10 from 1996-2018, the Wet Bulb Globe Temperature (WBGT) is adopted in 11 this paper to conduct an empirical analysis of temperature rises’ impact on 12 wage in various industries. It is found that, ﬁrstly, temperature rise will signif- 13 icantly reduce the growth of wage and show heterogeneity. Among them, the 14 growth of wage in the manufacturing industry is most prominently aﬀected 15 by rising annual average temperatures. Second, in terms of seasonal diﬀer- 16 ences, the negative impact of temperature rise on wage growth is mainly in 17 summer. Diﬀerent temperature swings and diﬀerent vulnerability producers 18 demonstrate lead to largely dissimilar marginal impacts of temperature rise 19 in diﬀerent regions. Wages in relatively vulnerable regions and regions with 20 relatively sharp temperature ﬂuctuations are more signiﬁcantly aﬀected by 21 the increase in average summer temperatures. Thirdly, in the long run, the 22 negative impact of annual average temperature increases on the wages of agri- 23 culture industries shows a notable cumulative eﬀect, which mainly comes from 24 the irreversible impact of temperature increase on labor productivity, and will 25 further widen the income gap between regions. Based on the above ﬁndings, 26 this paper proposes targeted strategies from two dimensions, “mitigation” and 27 “adaptation”, in order to narrow the regional income gap and achieve balanced 28 development, and to provide theoretical references for subsequent policies re- 29 sponding to climate change. 30

this paper proposes targeted strategies from two dimensions, "mitigation" and 27 "adaptation", in order to narrow the regional income gap and achieve balanced 28 development, and to provide theoretical references for subsequent policies re-29 sponding to climate change. 30 Keywords Climate change · Wet bulb globe temperature index (WBGT) · 31 Wage income growth · Relative poverty 32 1 Introduction 33 Today, climate change and its associated gradual slow-onset and extreme 34 sudden-onset weather events have become one of the greatest challenges to 35 human natural systems. As a sensitive region to global climate change, China 36 has been witnessing an annual average temperature rise of 0.24℃ per decade. 37 The rate is notably higher than the global average during the same period, 38 accompanying of which is a significant rise of the number of extreme high tem-39 peratures and heavy precipitation events (Zhang et al., 2020). According to the 40 IPCC assessment report, high temperature will be more frequent and stronger The main reason for this paper to select temperature rises as the research perspective is after a consideration of the difference between temperature change and climate change. While the latter represents a long-term transformation, temperature change, being shortterm, is the most direct and widest manifestation of climate change. Generally, the climate can be understood as the distribution function of the temperature. As climate itself is a distribution of multi-dimensional variables consisting of temperature, precipitation and wind, an accurate description of climate change must involve complex measurements on those dimensions, i.e., to measure the independent and joint distributions of multiple variables (such as those between wind and precipitation). However, in practice, it is often difficult for researchers to consider all the dimensions (mean, variance, covariance, extreme values, among others), therefore they tend to be able to investigate the effects caused by changes in only one (or several) meteorological factors to identify the impact of climate change. For example, looking at the economic effects of changes in average temperature (while controlling average precipitation) is only a rough description of (the multidimensional) climate change. As climate (change) economics develops, the controlled variables and statistical dimensions in econometric models are gradually increasing in numbers, thus the description of climate change is becoming more and more accurate.
2 Wet Bulb Globe Temperature (WBGT) index is the most commonly used index that comprehensively considers temperature and humidity to measure heat stress, and represents the heat intensity of body exposure to the environment.
to a certain extent on the mitigation and adaptation measures adopted by 96 the disaster-bearers (Brysse et al., 2013). Some literature, on the other hand, 97 has pointed out that no significant differences of productivity losses caused 98 by climate change have been seen between developed and backward regions 99 (Brysse et al., 2013), or that the economic consequences of climate change have 100 a greater impact on developed regions (Ramsey and Kwon, 1992). As research  To better clarify the impact of temperature rise on wages, this paper draws a 124 logic diagram for analyzing the differences in the damage caused by tempera-125 ture rise and its impact under different scenarios (Fig.1). Among them, (a) (b) 126 in Fig.1 presents the intrinsic linkage between the level of wages Q, the degree 127 of temperature fluctuation W , and the loss of wages D, respectively. When 128 the increase in temperature is at W 1 , the optimal level of wages (under the 129 condition of maximizing workers' benefits) is Q. Excessively cold or hot tem-130 peratures make it impossible for workers to concentrate on their work and may 131 damage their health, resulting in a decrease in marginal labor productivity (or 132 an increase in marginal cost of labor) and a decrease in work time. In particu-133 lar, the labor productivity of outdoor workers will be significantly reduced due 134 to loss of labor capacity under heat waves (such as fatigue and reduced muscle 135 endurance) (González-Alonso et al., 1999). At the same time, heat stress will 136 also lead to a loss of productivity due to reduced attention, cognitive abilities 137 and low-quality decision making among indoor workers (Orlov et al., 2020).

138
In this scenario, workers, as rational economic men, will adjust labor input 139 to maximize utility according to the current climate state. Therefore, in the 140 short run, with factor prices and labor supply and demand remain constant, 141 while temperature rise intensifies, i.e., when the temperature compared with 142 the historical average increases from W 1 to W 2 , the marginal output value of 143 labor factors keeps decreasing due to the negative temperature shock, while 144 the marginal cost curve of factors increases from M C 1 to M C 2 . At this point, 145 based on the objective to maximize utility, workers will inevitably further re-146 duce their labor input, and thus deviate from the optimal factor input W 1 .

147
As a result, affected by changes in the input of labor factors, the output will 148 decline, followed by a downward trend in wages. of temperature rise as in Fig. 1(b), its impact on wages usually shows a gradual 153 increase (Schlenker et al., 2006). Thus, for workers in different regions (with 154 different degrees of temperature rises), the temperature rise will likely cause 155 and even further the inter-regional wage gap. Second, in terms of the vulner- where Y denotes wages; L denotes work hours; A denotes labor productivity;
where ∆y it denotes the wage growth. According to the above equation, when 208 the temperature returns to the normal level, its average effect on work hours where T denotes the degree of temperature increase with a period lag K; θ i factor reducing workers' exposure and sensitivity, and helping adopt adaptive    The data used in this paper covers 30 provincial capitals. With reference to 332 existing studies, and due to the limited availability of data, the sample data in  Table.2.        rises on the growth of wage in agriculture will persist for a long time, and 481 the temperature rise will not only affect the growth of wage in the current 482 period, but also through the lagged effect. This may be due to the fact that 483 temperature increase not only affects workers in these industries, but also di-484 rectly affects their production (Roberts and Schlenker, 2013), both of which 485 will directly affect the wage growth in these sectors. Take crops as an example, 486 the rise in temperatures may increase the probability of occurrence and spread 487 of pests and diseases, shorten the fertility period, and lead to a decrease in 488 yield. In this case, producers, as rational economic men, will inevitably reduce 489 the input of each factor of production so as to maximize profits. When labor 490 supply remains unchanged, the decrease of labor hire demand will directly 491 affect the labor hire price, resulting in a decrease in wages, which will have a 492 lagged level effect. When the capital stock adjusts to a new steady state, the 493 permanent shock to productivity from this effect may in turn affect subsequent 494 capital accumulation, producing a lagged growth effect. As can be seen, there 495 is not only a level effect but also a growth effect of the negative effect of rising 496 temperatures, both of which may have an impact on the growth of wages in 497 agriculture. The difference is, when the temperature returns to normal, the 498 labor input time can return to normal, i.e., the level effect on wages will be re-499 versed. In contrast, however, the lagged growth effect on workers' productivity 500 is not reversed by reduced temperatures, and the existence of the effect means 501 that an increase in temperature rise in a period may lead to a prolonged low 502 wage level.

The cumulative effects of rising temperatures on wage growth -based
where T bin m it denotes the total number of days that the daily average temper- Standard errors in parentheses. * * * p<0.01 , * * p<0.05, * p< 0.1. fixed effects, and ε it denotes robustness standard errors of clustering to the 518 regional level. The standard for selecting the temperature reference group is: to try to select different temperature intervals as the reference group for regression analysis. When a certain temperature interval is used as the reference group, the estimated coefficients of the remaining intervals are all significantly negative, then the temperature interval is selected as the final reference group. At the same time, the range of the reference temperature selected in this paper is also similar to those used by Burke           and smart service technologies, should be introduced. This paper shows that 638 the "pro-poor" nature of the negative impacts of climate change may be at-639 tributed to the relatively weak coping capacity of workers in backward areas 640 and their difficulties in implementing climate adaptation strategies. Therefore, 641 while developing and diffusing technologies, we must make them more avail-642 able and acceptable. By introducing a reasonable technology subsidy policy, we 643 can transfer advanced technologies to backward areas by means of knowledge 644 popularization and technology promotion. This way, we may effectively make 645 workers in these areas more resilient to climate change and we may encourage 646 them to adopt proper adaptive measures to reduce negative impacts, so as to 647 prevent further widening of the income gap between regions and industries.

Conflict of interest 649
The authors declare that they have no known competing financial interests or 650 personal relationships that could have appeared to influence the work reported 651 in this paper.

677
The datasets generated during and/or analysed during the current study are 678 available from the corresponding author on reasonable request. 679