Economic and Agricultural Impacts of Building a Dam—Evidence from Natural Experience of the Three-Gorges Dam

Abstract

As the largest dam in the world, the impacts of the Three-Gorges Dam (TGD) on economy and agriculture in the counties along the Yangtze River in China have been subject to debates for a long time, but no conclusions have been made. This paper employs panel data with a wide variety of economic and agricultural variables for 751 counties over the period from 1997 to 2010, which covers the whole building period of the TGD. By calculating the distance of these counties to the dam site and applying the differences-in-differences (DID) method, the results generally imply that the TGD negatively affected economic growth in the downstream counties along the mainstream river. Conversely, counties located closer to the dam site in the upstream received economic benefits. Regarding its effects on agricultural productions, cotton yields of the upstream were negatively affected by building the TGD, while oil production planted in the upstream region were stimulated to grow since the functioning of the dam in 2003. This study proves both the economic and agricultural advantages/disadvantages of the dam in different construction periods for different locations of counties, and sheds light on policy implications for compensating the counties from economic and agricultural lenses due to the TGD construction.

1. Introduction

Dams are regarded as wise inventions. For instance, dams are capable of controlling the yield of water and generating electricity. The existing research, see McCully [1] for example, mentioned that a dam is an enormous, long-term and irreversible project which is also without benchmarks to follow. This standpoint indicates research on dams and particular the consequences of building a dam is not obvious, but it should have long-lasting interest and importance.

The functions of dams have been widely studied in the literature. On the one hand, dams provide great opportunities to meet the needs of human well-being [2], such as flood control, power generation, waterway dredging, and agricultural irrigation [3,4], and it even contributes to social sustainability by generating electricity instead of incinerating the solid fuel [5]. Indeed, dams have been proven to benefit the environment by reducing CO₂ emissions via providing hydropower in China [6]. Still, we cannot ignore the negative effects of building a dam, such as accelerating soil erosion, reducing biodiversity, producing toxic wastes [7,8], and even lead to economic losses [9].

This article focuses exactly on the construction period of the TGD, which is across the longest river, the Yangtze River, in Asia, and hence, it has gained tremendous attention around the world. However, there exist various aspects of debates of the TGD even before its construction, which include environmental, social, biological and ecological concerns [10,11,12,13,14], respectively.

In fact, building the TGD could have both economic and agricultural effects. First, it could contribute to environmental and economic losses of the downstream areas along the dammed river. As far as known, the dam cuts water and prevents it from flooding during the wet seasons, and hence the downstream lakes are more likely to be dried up especially during seasons with rare rains, such as the phenomena of the Poyang Lake [15] and Dongting Lake. Similarly, Shankman and Liang [16] and Shankman et al. [17] researched the flood disasters of Poyang Lake since the construction of the TGD. Their articles suggested that the TGD had adverse effects towards the downstream areas. Second, building a dam could also lead to a negative effect towards certain types of agriculture yields. As a large number of fertile lands are drowned during the period of rising water, crops pay the price, especially in those counties that lose the most lands, which, in turn, could also lead to local economic shrinkage. However, this is not always the case, as Duflo and Pande [18] found that dams benefit the downstream areas. From the year 2003, the TGD began to generate electric power, which could also contribute to the booming of the local economy [12,19]. Additionally, Zhu et al. [20] found that the construction of dams in Yunnan province enriched the diversity of the local plants. There is also with evidence that construction of the TGD created favorable natural conditions for agricultural output located in the downstream regions [21]. From the above analysis, both the positive and negative economic and agricultural effects could exist due to the building of a dam along the damming river. However, few studies investigated this issue from both the economy and agriculture lenses.

In order to further study the potential economic and agricultural effects caused by dams, this study takes the TGD as a case study, and empirically explores the following research questions: Are there any economic impacts caused by the TGD during the building period towards different regions along the damming river? Are there any heterogeneous impacts on different types of crops along the dammed river? In this article, the damming process is regarded as a shock towards the counties along the Yangtze River, and especially to those counties closer to the dam site, and hence, this paper compares counties in different locations and tries to explore both the economic and agricultural impacts caused by the TGD. In particular, the dam-building event mainly has direct impacts on 90 counties, which were exactly along/across the Yangtze River from the headstream region down to the sea entrance across 14 years.

From the results, the economic effect during the construction period turned out to be negative towards the downstream counties, and also damaged those along the mainstream of the river by shrinking the GDP per capita by 12 and 6.5%, respectively. However, counties 100 km closer to the dam site from the upstream region benefited an additional 2% GDP per capita growth during the building of the TGD. When the dataset was limited to counties within 500 km, those counties neighboring the dam site in the downstream were observed to obtain benefits by being protected from floods. Regarding the agricultural aspect, during the construction of the TGD, cotton yield in the upstream regions was negatively affected by building the dam. Interestingly, since 2003, dam building showed a positive impact on oil production in the upstream regions.

This study contributes to the existing literature in the following aspects: (i) it simultaneously explores the economic and agricultural impacts and evaluates the benefits and losses of building the TGD for a relatively longer period of time; (ii) it treats the TGD building as a natural experiment and calculates the distance between counties around/across the Yangtze River and the dam site, and then uses the ’Treatment for Treated’ method to compare the counties in the treatment group and control group before and after damming; (iii) the results offer important political implications in examining the damaged areas, and hence, there are opportunities to redistribute welfare and resources to compensate the ’victims’.

The structure of the remainder of this paper is as follows. Section 2 describes the background and summarizes the relevant literature on building the TGD. Section 3 shows the data and methodology. In Section 4, the main results and robustness checks are reported. This paper finishes with concluding remarks in Section 5.

2. Background

This section reports the functions and consequences of building dams, and specifically describes the background of the TGD. Then, the literature on economic and agricultural impacts of dams is listed.

Dams are well known for the main functions of hydropower generation, flood control, irrigation and navigation [18,22]. However, it is impossible to receive most of the advantages from a dam simultaneously. For example, when water is stored in a reservoir, it is an efficient way to reduce floods, but this action dries up the downstream regions, which is likely to break out into a terrible drought. Moreover, when water is released in order to make the reservoir empty and prepare for the next period of the rainy season, it is good for the whole biological cycle, but it is cost-inefficient as the dam fails to save the water for generating hydropower. When the disadvantages outweigh the advantages, dams are removed after construction. See for example, Loomis [23] and Gelfenbaum et al. [24]. In general, some functions of the dams are mutually reinforcing, while others are conflicting; they are the vivid embodiment of the direct and indirect impact of water resource projects on the social-economic and environmental system [21].

There are some relevant studies that explored the economic and agricultural impacts of various dams with different results. Duflo and Pande [18] found that agriculture in downstream regions of Indian dams receive benefits; however the dam site is sacrificed. They also demonstrated that big dams turn out to be inefficient by considering both the costs and benefits. Bao [22] found that large-scale Chinese dams cause the decrease in revenue for the upstream regions, but the dam site counties earn benefits. Bohlen and Lewis [25] drew attention to the benefits of two dams in the U.S. by comparing the habitats along the river with those further away. According to the investigation of Borimnejad and Salimian [26] on the Taleghan Dam, economic benefits appeared along the river areas, but the environment was observably damaged. Some other papers further mentioned the impacts of dams on agricultural production. As far as known, agriculture is closely associated with the natural environment. As an important part of the ecological barrier in the upper Yangtze River, the TGD has an important function in the ecological security of the Yangtze Basin [13,27].

The TGD project spent more than 500 billion RMB and 17 years to finish from 1997 to 2009, and our data recorded the whole construction period from the year 1997 to 2010. Before 1997, the project was mainly the preparatory step for construction. In 1997, the Yangtze River was successfully dammed at the location of Yichang, which was regarded as the end of the first period. The second period of building the TGD is from 1998 to 2003 for the main building process, which elevated the water level to 135 m deep, and it began to function. In 2009, the project was finished and it became the largest dam in the world. The last seven years, from 2003 to 2009, were classified into the third period.

The topics of the TGD and its impacts on economics and agriculture are very popular and debated in the literature. In the research of Jackson and Sleigh [11], they investigated the socio-economic effects of the TGD, which turned out to boost the local economy, but the resettlement of a large population appeared to have a negative effect. Xu and Milliman [28] explored that the TGD corroded the Yangtze River, especially in the downstream regions. Moreover, the research of Müller et al. [29] found that the downstream area of the TGD was threatened. Regarding the agricultural aspect, on the one hand, some papers showed that the TGD has damaged the biodiversity [14], reduced cultivated land resources [27,30], and caused agricultural pollution [27] in the reservoir area. Similarly, Nakayama and Shankman [31] found that the TGD enhances the risk of flood in early summer near the area of Poyang Lake. Tian et al. [32] suggested that the Three Gorges Reservoir Area has become one of the most ecologically sensitive areas in China. Liang et al. [33] and Yang et al. [13] pointed out that the ecosystem along the Yangtze River has gradually improved, represented by the transition from the traditional grain system to the fruit forest ecosystem in the mountainous agricultural areas of the Three Gorges Reservoir Area. Moreover, a couple of other studies demonstrated that the construction of the TGD had an adverse effect on agricultural environment, including the agricultural water and the land resources [14,27,30]. However, on the other hand, its effective flood control function is able to reduce the scale and frequency of flooding. Consequently, it created favorable conditions for agricultural development [21].

Besides the literature on the TGD, it is also worth exploring the methodologies that aim to solve the causes and effects of the natural experience of the TGD. In the previous research, Bao [22] investigated Chinese dams by applying the DID. Qian [34] used an exogenous event to estimate income changes by observing the phenomenon of sex imbalances. Some other research used the regression discontinuity (RD) design to evaluate some projects, such as Meng [35] and Chen et al. [36]. Deng et al. [37] employed the average-effect model and alternative dynamic model in order to determine how the natural disaster of the Wenchuan earthquake impacted living habits on heights. In this article, we take advantage of the Treatment Effect for Treated, a type of DID method, in order to compare the treatment group and the control group in examining the economic and agricultural effects of building the TGD.

From the analysis above, it is not difficult to recognize that the TGD has significant and lasting effects on the river areas in a wide variety of aspects; however, there is little empirical research that examine the impacts of building the TGD both on GDP growth and agricultural productions by applying the DID method.

3. Data and Methodology

3.1. Data

Data of this research are obtained from the National Bureau of Statistics of the People’s Republic of China (Further information can be found from the main page at http://www.stats.gov.cn (accessed on 21 May 2022). It is an official organization that records and publishes authoritative annual statistics of China in economic, demographic, environmental, and social aspects, etc. This data include 2084 counties across 31 provinces of China from the year 1997 to 2010, and it has information on economic factors (such as GDP, income), agricultural productions (such as cotton production and oil production), demography and some other characteristics of each county.

Regarding the sample selection, regions across and/or along the mainstream of the Yangtze River are regarded as the treatment group in this study, and the counties belong to the Yangtze River regions (Yangtze River regions include the provinces of Tibet, Qinghai, Sichuan, Yunnan, Chongqing, Hubei, Hunan, Jiangxi, Anhui, Jiangsu and Shanghai.) but not across the river are the control group. Counties along the main stream are selected as they experienced the exact same event, dammed and affected by the TGD since the year 1997, but they have various distances to the dam site. Notably, the control group is not directly affected by the TGD, but they are located at the neighborhood of the treatment group. On this basis, there exist 90 samples in the treatment group, and 661 counties in the control group. Figure 1 shows the distribution of the selected sample along the Yangtze River. The grey cutoff suggests the location of the dam site in Hubei province. The treatment group is exactly across or along the main stream.

In addition to county selection, distance to the dam site is another key variable of interest, but unfortunately, there are no existing data which provide the distance of each county to the dam site. Hence, it was necessary to collect this information personally. Specifically, the distances were calculated by using software to record how far away of each selected county is to the dam site. One issue of concern here is that distance to the dam site for the treatment group is not a simple straight line; instead, it is calculated exactly from the nearest boundary of a county to the dam site along the shape of the river, which should be longer than a straight line as the river is curved. For counties in the control group, as the distance according to the above description does not exist, the distances for each of them are obtained by calculating the average values of the treatment group of the same province. Samples are classified into three types according to their geographical locations: upstream, dam site (there is only one county just located at the dam site, which has zero distance to the dam), and downstream counties.

Descriptions of the main variables are given in

Table 1. Descriptions of the main variables.

Variable	Variable Description
Economic characteristics
GDP	The gross domestic product (Nominal GDP) of each county for a whole year by considering Consumer Price Index (in 10,000 Yuan)
GDP growth rate	The difference of two years’ GDP divided by the latter year’s GDP (in %)
GDP per capita	The gross domestic product for each person in a county (in Yuan)
$\Delta log{y}_{it}$	The logarithm term of GDP per capita minus the logarithm term of GDP per capita in the year of 1997
Lgdp_ pre	The logarithm term of GDP per capita in the previous year
Lgagri	The logarithm term of agriculture output per capita
Agricultural characteristics
lgcotton	The logarithm term of cotton production of a county (in tons)
Lgoil	The logarithm term of oil production of a county (in tons)
Area/time controls
d	The ‘curve distance’ From the nearest border on the main stream river to the location of the TGD site (in km)
Population	The population of a county (in 10,000)
Area	The area of a county (10 km2) divided by ‘Population’
Power spread	The political authority is/was spread in a county (‘1’ yes; ‘0’ no)
Direct administration	A county is directly administered by its province since 1983 (‘1’ yes; ‘0’ no)
U	Upstream of the dam site (‘1’ yes; ‘0’ no)
Down	Downstream of the dam site (‘1’ yes; ‘0’ no)
Year	The year variables range from 1994 to 2010
Year2	The square term of the variable ‘Year’
yearj	The dummy variables for year j (j can be the year from 1998 to 2010, and the year 1997 is regarded as the basic year in estimations)
Year03	The dummy variable for period. (‘1’ if the year equals or after 2003; ‘0’ if the year is before 2003)
River	The county is located along/across the mainstream. (‘1’ yes; ‘0’ no)
Agri_power	Total power of agricultural machinery of a county (in 10,000 kilowatts)
Population_agri	The rural population of a county (in 10,000)
employ_agri	The number of people working in the countryside of a county (in 10,000)
employ_nlmy	The number of persons engaged in farming, forestry, animal husbandry and fisheries in a county (in 10,000)

, while

Table 2. Descriptive statistics of the main variables (treatment group).

Variable	Mean	Std. Dev.	Min	Max	Observations
Economic characteristics
GDP	757,362.8	1,464,534	4819.66	1.62 × 107	1206
GDP per capita	11,040.14	16,443.53	798.11	142,028.2	1193
$\Delta log{y}_{it}$	0.528	0.512	−0.606	2.423	1157
Agricultural characteristics
Lgcotton	7.296	2.443	0	10.84	574
Lgoil	9.047	1.888	2.079	11.62	933
Area/regional controls
d	1235.777	804.672	0	3814.5	1253
Population	62.143	36.654	2	147	1226
Area	5526.65	15,712.84	330.9	123,463	1238
Power spread	0.151	0.358	0	1	1253
Direct administration	0.081	0.272	0	1	1253
River	1	0	1	1	1253
Agri_power	22.25	16.35	0	119	1191
Population_agri	50.40	30.18	1	127	1226
employ_agri	27.02	16.88	0.6	86	1226
employ_nlmy	14.37	9.296	0.409	49.33	966

,

Table 3. Descriptive statistics of the main variables (control group).

Variable	Mean	Std. Dev.	Min	Max	Observations
Economic characteristics
GDP	320,702.9	550,052.8	1561.451	1.70 × 107	8904
GDP per capita	6430.351	7342.187	708.82	238,922.5	8843
$\Delta log{y}_{it}$	0.577	0.526	−3.320	3.530	8178
Agricultural characteristics
Lgcotton	5.476	2.834	0	11.51	3177
Lgoil	8.392	1.947	0	12.43	6889
Area/regional controls
d	1399.363	813.995	349.6	3400.2	9263
Population	47.322	38.836	0.8	220	9003
Area	4378.515	18,002.45	1	1,400,000	9052
Power spread	0.118	0.323	0	1	9263
Direct administration	0.019	0.136	0	1	9263
River	0	0	0	0	9263
Agri_power	28.15	652.3	0	60958	8744
Population_agri	40.13	33.89	0	199	9037
employ_agri	21.45	18.04	0	110	9030
employ_nlmy	13.48	10.98	0.003	65.74	7201

,

Table 4. More details of the treatment group (Average data values of the upstream, dam site and downstream).

Variable	Upstream	Dam site	Downstream
Economic characteristics
GDP	231,018.3	842,443.9	1,149,659
GDP growth rate	0.105	0.028	0.091
GDP per capita	6064.69	15,505.44	14,771.15
$\Delta log{y}_{it}$	0.575	0.229	0.500
Agricultural characteristics
Lgcotton	2.907	3.430	7.966
Lgoil	7.567	9.978	10.133
Area/regional controls
Distance	1509.57	0	1054.228
Population	44.718	53.729	75.585
Area	10,964.98	3531.714	1468.44
Power spread	0.118	0.0286	0.173
Direct administration	0.083	0	0.081
Agri_power	11.46	17.81	30.13
Population_agri	39.16	42.30	59.18
employ_agri	21.52	20.16	31.32
employ_nlmy	14.31	11.75	14.42
Observations	532	14	707

and

Table 5. More details of the treatment group (Average data values of different distances away from the dam site).

Variable	D ≤ 1000	D > 1000	D ≤ 1500	D > 1500	D ≤ 2000	D > 2000
Economic characteristics
GDP	426,722.2	994,745.8	472,832.7	1,359,508	853,069	83,591.45
GDP growth rate	0.083	0.106	0.090	0.111	0.095	0.107
GDP per capita	6063.48	14,606.15	7541.47	18,326.77	11,472.68	8032.53
$\Delta log{y}_{it}$	0.432	0.598	0.475	0.638	0.531	0.506
Agricultural characteristics
Lgcotton	7.962	6.835	7.567	6.388	7.309	0
Lgoil	9.940	8.373	9.500	7.993	9.439	5.231
Area/regional controls
Population	72.386	54.871	66.403	53.247	69.732	8.516
Area	2037.617	7987.235	1913.013	13,125.25	1090.741	31,202.84
Power spread	0.295	0.049	0.200	0.047	0.162	0.071
Direct administration	0.195	0	0.119	0	0.092	0
Agri_power	24.13	20.85	23.30	19.90	24.40	4.212
Population_agri	57.95	45.04	53.96	42.95	56.56	6.828
employ_agri	29.04	25.58	28.07	24.82	30.32	3.720
employ_nlmy	16.78	12.66	16.03	10.93	15.92	3.425

report statistical analysis of both the treatment group and control group. According to GDP growth, those tables indicate that counties along the Yangtze River were generally developing faster. However, the upstream areas are mostly located in the west of China, which were much less developed than the east, but the western counties are larger in area and faster in economic growth rate (recorded as 10.5% on average). Regarding agricultural records, it can be seen from Table 2, Table 3, Table 4 and Table 5 that the average yields of cotton and oil crops of the counties along the Yangtze River were on average higher than those of the neighboring counties. For the treatment group, the average yields of cotton and oil crops in the downstream regions of the Yangtze River were generally higher than those in the upstream regions.

There is only one county which belongs to the dam site region, namely Yichang, with zero distance. In practice, the dam site has only one sample and it is very hard to run regressions; therefore, this observation is classified into the upstream group, as it has direct effects from the water falling down and accumulating at the reservoir. Additionally, the county Geermu in Qinghai province with the distance of 3958 km is the farthest sample away from the dam site, and it is also the headstream of the Yangtze River.

Figure 2 provides insights about the possible economic effects of the dam along the Yangtze River within 500 km/1000 km/1500 km/2000 km away from the dam site. By just focusing on the treatment group, samples closer to the dam site in the downstream area is with slower GDP comparing to the upstream counties. In the year 2005, both of the downstream and upstream regions have a declining trend especially for the counties within the distance of 500 km. However, this is not always the case; the GDP growth of the downstream region tends to exceed the upstream group after 2009, when the TGD project was completed.

Figure 3 and Figure 4 illustrate the effects of the dam on cotton yield growth and oil yield growth, respectively. It can be seen from the two figures that the growth rate of the cotton production by different distances between the county and the dam site was generally faster in the downstream regions comparing to the upstream ones. However, the growth of oil yields was faster in the upstream regions. Due to the different situations for different crops, the agricultural impacts of building the TGD could also turn out to be different for different locations.

3.2. Methodology

This article selects the counties along the Yangtze River as the treatment group, and the Yangtze River areas are the control group (see the distribution in Figure 1). The natural experience of the TGD cut the Yangtze River and brought enormous changes along the river. Before damming, the GDP, agricultural production and the county-level characteristics of the selected counties are well-recorded. After damming, counties along the river are directly affected by the shock, while those in the control group with similar characteristics were not affected by the TGD.

The methodology in use is Treatment Effect for Treated. Where Y_i1(0) is the GDP per capita/agricultural production of county i in the treatment group in the basic year: 1997. Y_i1t(1) represents the same issue of county i in the treatment group in year t (where t is from 1998 to 2010 according to the dataset). Similarly, we can obtain Y_i0(0) and Y_i0t(1) for the control group before and after the damming process. The causal effect is measured by calculating the different outcomes between the two groups, which is expressed as follows:

$$d{Y}_{it}=\left[Y_{i1t}\left(1\right)-Y_{i1}\left(0\right)\right]-\left[Y_{i0t}\left(1\right)-Y_{i0}\left(0\right)\right]$$

(1)

There are three variables of interest: the ‘river distance’, a dummy variable for whether the county is along the river, and a dummy variable for upstream location. Counties along the river are supposed to be directly affected by the TGD, and counties closer to the dam site naturally receive more direct impacts from the dam than those further away, and the upstream and downstream regions have some differences as well according to the biological/geographical characteristics of the dam system. Based on the previous analysis, the initial model is written as follows:

$$d{Y}_{it}=f\left(d_i,U_i,R_i,Z_{it},T_t\right)$$

(2)

In Equation (2), dY_it is the difference of GDP per capita of year j and the year 1997. Similarly, in order to examine the agricultural effect of the TGD, dY_it also serves as the records of a certain sort of agricultural production in year j. On the right-hand side, d_i represents for ‘distance’. U_i is a dummy variable and equals to 1 if a sample located in the upstream of the dam site. R_i equals to 1 if the county is along the river. Z_it represent some other control variables, for instance the provincial fixed effects and county-level characteristics. T_t is the year effect. This equation can be rewritten as Equation (3):

$$log{y}_{it}^{}={\alpha }_t+{\beta }_1{d}_i\ast \omega +{\beta }_2{U}_i\ast \omega +{\beta }_3{d}_i\ast U_i\ast \omega +{\beta }_4{R}_i\ast \omega +{\beta }_5{R}_i\ast d_i\ast \omega +{\beta }_6{Z}_{it}+{\beta }_7{T}_t+{\epsilon }_i^t$$

(3)

$\mathrm{where} \omega =1 \mathrm{if} t>1997, \mathrm{and} \omega =0 \mathrm{if} t=1997$

$log{y}_{it}^{}$ is the natural logarithm term of Y_it for the county i at year t. Subtracting terms with $\omega =1$ and $\omega =0$, we get Equation (4):

$$\Delta log{y}_{it}=log{y}_i^t-log{y}_i^{1997}=\Delta {\alpha }_t+{\beta }_1{d}_i+{\beta }_2{U}_i+{\beta }_3{d}_i\ast U_i+{\beta }_4{R}_i\ast \omega +{\beta }_5{R}_i\ast d_i\ast \omega +{\epsilon }_i^t-{\epsilon }_i^{1997}$$

(4)

where $‘{\epsilon }_i^t-{\epsilon }_i^{1997}’$ is the error term, which mainly expains the differences in time and some other unobserved factors. $\Delta {\alpha }_t$ symbolizes the intercept term. ${\beta }_l$(where l = 1,2,3,4) are parameters of interest, which suggest the underlying effects of the TGD project during the process of construction. In our model, the ’Hausman test’ turns out to reject the random effect at the p-value of 0.078. This suggests the fixed effect is more appropriate.

This basic model is extended in mainly two steps: first, instead of adding the simple year term and the square term of the year, we multiple the dummy variables of each year from 1998 to 2010 with the term of ‘$d_i\ast U_i$’ in Equation (4), and a series of intersection terms are obtained in the form of ‘$Yea{r}_j\ast d_i\ast U_i$’. This term gives more information on how the establishing periods and distance of each county dynamically contribute to GDP per capita, cotton yield and oil yield, respectively (see Equation (5)).

$$\Delta log{y}_{it}={\alpha }_t^{\prime }+{\beta }_1{d}_i+{\beta }_2{d}_i^2+{\beta }_3{U}_i+{\beta }_4{d}_i\ast U_i+{\beta }_4{R}_i\ast \omega +{\beta }_5{R}_i\ast d_i\ast \omega +{\beta }_{6j}{\displaystyle \sum }_{j=1}^{13}Yea{r}_j\ast d_i\ast U_i+{\beta }_7{Z}_{it}+v_i^t$$

(5)

By classifying the years into four periods according to the official processes during the TGD construction, dummy variables of the periods are added in the model to clarify the effects of the TGD in various construction processes. The first period is the year 1997, and the next period is from 1998 to 2003, which begins with the damming year and ends with the official operation year. The third period is from 2004 to 2009. Finally, in the year 2010, the dam is well-finished and this period is regarded as the last step in our analysis.

Furthermore, by separating the samples a certain distance away from the dam (e.g., farther than 1000 km and less than 1000 km), it is a robust check to explore the effects of building the TGD. Equation (6) is accordingly given as below and it just includes the treatment group:

$$\Delta log{y}_{i1t}={\alpha }_t+{\beta }_1{d}_i\ast \omega +{\beta }_2{U}_i\ast \omega +{\beta }_3{d}_i\ast U_i\ast \omega +{\beta }_4{R}_i\ast \omega +{\beta }_5{R}_i\ast d_i\ast \omega +{\beta }_6{Z}_{it}+{\beta }_7{T}_t+{\epsilon }_i^t$$

(6)

The above analysis only focuses on the damming shock in 1997, but this not the only shock within the 14 years. Another issue happened in 2003, which suggests the end of damming and the start of operation. Hence, the following model is accordingly established shown in Equation (7):

$$\begin{array}{cc}\hfill \Delta log{y}_{it}={\alpha }_t^{\prime }& +{\beta }_1{d}_i+{\beta }_2{d}_i^2+{\beta }_3{U}_i+{\beta }_4{d}_i\ast U_i+{\beta }_{5}Yea{r}_{03}+{\beta }_{6}Yea{r}_{03}\ast U_i+{\beta }_{7}Yea{r}_{03}\ast di\hfill \\ & + {\beta }_{8}Yea{r}_{03}\ast d_i\ast U_i+{\beta }_9{R}_i+{\beta }_{10}{d}_i\ast R_i+{\beta }_{11}Yea{r}_{03}\ast R_i+{\beta }_{12}Yea{r}_{03}\ast d_i\ast R_i\hfill \\ & +{\beta }_{13}{Z}_{it}+v_i^t\hfill \end{array}$$

(7)

where the dummy variable Year₀₃ suggests the shock took place in the year 2003, and it takes the value of 1 if the year equals to 2003 or afterwards.

4. Empirical Results

4.1. The Baseline Results

This part shows the baseline results based on Equation (4).

Table 6. Baseline results—the economic impact of building the TGD.

Variable
U	0.140 ***	0.120 ***
	(3.24)	(0.043)
d	0.0001	0.0001
	(0.00007)	(0.0001)
d*U	−0.0002 ***	−0.0002 ***
	(0.00007)	(0.00007)
River	−0.045 **	−0.065 ***
	(0.021)	(0.021)
d*River	0.00006 ***	0.00008 ***
	(0.00001)	(0.00001)
Year	0.051 ***	0.050 ***
	(0.003)	(0.003)
Year2	0.004 ***	0.005 ***
	(0.0002)	(0.0002)
Power spread		−0.11 ***
		(0.011)
Area		−2.28 × 10−6 ***
		(1.07 × 10−6)
Constant	−0.030	0.0004
	(0.055)	(0.056)
Observations	9378	9290
No. of group	11	11
R-squared	0.62	0.62

Notes: Figures in the parentheses are standard errors; *** denotes significance at the 1% level; ** denotes significance at the 5% level; * denotes significance at the 10% level.

reports the estimated results for the basic model. The dependent variable $\Delta log{y}_{it}$ represents the logarithm term of GDP per capita. It is clear that upstream counties averagely experienced significant positive effects on economic output comparing with downstream counties; the results show that the per capita GDP of the upstream counties increased by 14%. Building the TGD reduced the GDP per capita of the counties along the Yangtze River by 4.5%. By just observing the term ‘d*River’, generally, counties along the Yangtze River an additional 1 km further away from the dam site showed 0.006% less damage by the TGD. By exploring the effect of the interaction term ‘d*U’, the empirical results show that with 1 km distance away from the dam site, counties in the upstream region had a 0.02% increase in per capita GDP growth comparing to those in the downstream counties. Time effects are always positive as expected, which suggests the economy is growing with time.

Table 7. Baseline results—the agricultural impact of building the TGD.

	(1)	(2)	(3)	(4)
	Lg_cotton	Lg_cotton	Lg_oil	Lg_oil
U	−0.774 ***	−0.891 ***	0.160 *	0.145
	(0.268)	(0.268)	(0.097)	(0.097)
d	−0.000	−0.000	−0.000	−0.000
	(0.000)	(0.000)	(0.000)	(0.000)
d*U	0.001 **	0.001 **	−0.000	0.000
	(0.001)	(0.001)	(0.000)	(0.000)
River	0.292 *	0.298 *	0.168 ***	0.156 ***
	(0.170)	(0.171)	(0.049)	(0.049)
d*River	0.000	0.000	−0.000 *	−0.000
	(0.000)	(0.000)	(0.000)	(0.000)
Year	−31.460 ***	−31.241 ***	−11.775 ***	−12.810 ***
	(7.230)	(7.213)	(2.757)	(2.776)
Year2	0.008 ***	0.008 ***	0.003 ***	0.003 ***
	(0.002)	(0.002)	(0.001)	(0.001)
Agri_power		−0.001		−0.002 ***
		(0.001)		(0.000)
Population_agri		0.005 **		0.007 ***
		(0.003)		(0.001)
employ_agri		−0.000 *		−0.000 ***
		(0.000)		(0.000)
employ_nlmy		−		0.002
		(0.005)		(0.002)
Power_spread		0.121 **		0.036
		(0.054)		(0.023)
Area		0.002 ***		0.000***
		(0.001)		(0.000)
_cons	31,576.190 ***	31,359.599 ***	11,783.867 ***	12,818.263 ***
	(7247.295)	(7230.341)	(2763.840)	(2783.085)
Observations	3610.000	3587.000	7702.000	7484.000
R-squared	0.039	0.047	0.021	0.032

Notes: Figures in the parentheses are standard errors; *** denotes significance at the 1% level; ** denotes significance at the 5% level; * denotes significance at the 10% level.

reports the estimated results based on the basic model of Equation (4). As can be seen from Table 7, cotton production of the upstream counties was negatively affected by the damming. Additionally, the adverse effect is smaller if a county from the upstream is located further away from the dam site. In Column (3), oil crops along the upstream of the Yangtze River had a positive effect on its yield with 10% level of significance. Meanwhile, the yields of cotton and oil in the counties along the Yangtze River increased by 29.2% and 16.8% on average, respectively.

The reason for the negative effect of building the TGD on cotton yield for the upstream region could be that: first, the upstream regions of the Yangtze River are not the main cotton planting areas, and hence the endowments for planting cotton are worse comparing with the downstream regions, such as the Jianghan plain and the Jianghuai Plain. The damming shock was likely to further damage the cultivated land in the upstream reservoir area, which contributed to the adverse effect on cotton yield. In contrast, the natural and socio-economic conditions of agricultural development in the downstream region are advantageous and suitable for large-scale cotton cultivation. The results also confirmed that the construction of the TGD accumulated more favorable conditions for cotton cultivation in downstream areas. Indeed, as the conditions for growing crops along the Yangtze River are favorable, the results confirmed that the cotton and oil crops in the counties along the Yangtze River were positively affected by building the dam.

According to Equation (5),

Table 8. Results estimating the extended model (dependent variable is $\Delta log{y}_{it}$).

Variable
U	0.137 **	0.226 ***
	(0.065)	(0.061)
d	0.0001	0.0002 **
	(0.0001)	(0.0001)
d*U	−0.0005 ***	−0.0005 ***
	(0.0001)	(0.0001)
River	−0.060 *	−0.022 *
	(0.032)	(0.030)
d*River	0.0007 ***	0.00004 *
	(0.00002)	(0.00002)
Yearj*d*U	POSITIVE	POSITIVE
Power spread		0.501 ***
		(0.013)
Direct administration		0.057 **
		(0.029)
Area		−6.39 × 10−7
		(1.52 × 10−6)
Constant	0.537 ***	0.371 ***
	(0.083)	(0.078)
Observations	9378	9290
No. of group	11	11
R-squared	0.30	0.33

Notes: Figures in the parentheses are standard errors; *** denotes significance at the 1%; ** denote significance at the 5%; * denotes significance at the 10%; ‘Yearj*d*U’ are a series of variables for 13 periods (j = 1, 2, …, 13). ‘POSITIVE’ suggests the parameters are significantly positive.

adds a series of intercept terms of the year dummy variable multiple by distance and the upstream dummy variable. Beside the same results obtained before, i.e., the upstream term was positive and the intercept term ‘d*U’ was negative, it is clear to observe that the intercept terms ‘Year_j*d*U’ turned out to be positive on GDP per capita (the parameters for each year of the ‘Year_j*d*U’ are not listed in Table 8; instead, they are described by a simple ‘POSITIVE’). The above outcomes are interesting in terms of illustrating that counties located far away earned long-lasting benefits in economic development with time. In other words, the dam negatively affects those regions closer to it.

It is also important to examine the effects of damming by just keeping the treatment group and separating it into different groups of distance, such as 500 km/1000 km/1500 km. The relevant figures and estimated results are shown in the Appendix (see Figure A1/

Table A1. The robust results—The economic impact of building the TGD by different groups of distance (The upstream regions).

Variable	≤500 km	≤1000 km	≤1500 km
U	0.685 **	0.770 ***	0.283 ***
	(0.294)	(0.252)	(0.161)
d	−0.003 *	0.001	0.0005 *
	(0.0018)	(0.0007)	(0.0003)
d2	5.95 × 106 **	−7.02 × 10−7	−3.43 × 10−7
	(2.97 × 10−6)	(7.39 × 10−7)	(2.47 × 10−7)
d*U	−0.00005	−0.0006 *	−0.0003
	(0.0008)	(0.0003)	(0.0004)
Year	0.035 ***	0.042 ***	0.046 ***
	(0.009)	(0.007)	(0.006)
Year2	0.001	0.004 ***	0.004 ***
	(0.0007)	(0.0006)	(0.0005)
Power spread	−0.089 ***	−0.165 ***	−0.203 ***
	(0.032)	(0.025)	(0.021)
Direct administration	0.120 ***	0.102 **	0.086 **
	(0.039)	(0.042)	(0.043)
Area	−0.015 **	−0.012 ***	−0.002 *
	(0.006)	(0.004)	(0.001)
Lgdp_ pre	0.105 ***	0.143 ***	0.114 ***
	(0.029)	(0.022)	(0.016)
Lgagri	0.554 ***	0.072	0.067
	(0.084)	(0.054)	(0.043)
Constant	−4.543 ***	−1.823 ***	−1.735 ***
	(0.084)	(0.435)	(0.345)
Province	YES	YES	YES
No. of group	18	36	59
Observations	243	481	765
R-squared	0.83	0.71	0.77

Notes: The figures in the parentheses are standard errors; *** denotes significance at the 1% level; ** denotes significance at the 5% level; * denotes significance at the 10% level. The dummy variables ‘Province’ are provincial fix effects.

/

Table A2. The robust results—The economic impact of building the TGD by different groups of distance (The downstream regions).

Variable	≤500 km	≤1000 km	≤1500 km
Down	0.108	0.073	−0.127 *
	(0.146)	(0.094)	(0.067)
d	−0.004 ***	−0.0001	−0.0001
	(0.0008)	(0.0002)	(0.0002)
d2	8.22 × 10−6 ***	1.16 × 10−7	1.71 × 10−7 *
	(1.41 × 10−6)	(2.52 × 10−7)	(9.79 × 10−8)
d*Down	−0.0008 **	0.0001	−0.0001
	(0.00038)	(0.0001)	(0.0002)
Year	0.069 ***	0.119 ***	0.131 ***
	(0.013)	(0.011)	(0.010)
Year2	−0.003 ***	−0.005 ***	−0.005 ***
	(0.001)	(0.001)	(0.0008)
Power spread	−0.022	−0.106 ***	−0.139 ***
	(0.046)	(0.039)	(0.034)
Direct administration	0.131 **	-0.035	0.086 **
	(0.053)	(0.052)	(0.043)
Area	−0.004	0.001	−0.001
	(0.003)	(0.0015)	(0.061)
Lgdp_pre	0.353 ***	0.553 ***	0.542 ***
	(0.037)	(0.029)	(0.021)
Lgagri	0.604 ***	0.384 ***	0.330 ***
	(0.082)	(0.052)	(0.042)
Constant	1.350 **	0.544 *	1.314 ***
	(0.649)	(0.318)	(0.292)
Province	YES	YES	YES
No. of group	18	37	61
Observations	243	481	781
R-squared	0.87	0.82	0.86

Notes: The figures in the parentheses are standard errors; *** denotes significance at the 1% level; ** denotes significance at the 5% level; * denotes significance at the 10% level; The dummy variables ‘Province’ are provincial fixed effects.

in Appendix A). Figure A1 indicates that counties closer to the dam site (within the distance of 500 km/1000 km/1500 km) generally had lower GDP per capita than those far away. According to this figure, there was a sudden decline after the TGD project began construction. However, a different phenomenon is observed in the last graph, which separates counties by the distance of within 2000 km and above 2000 km. Counties with shorter distances did not show poorer economic performance. There are mainly two reasons for this: one is that only a few samples are above 2000 km from the dam site, and those which are above 2000 km include a couple of the least developed counties in Qinghai and Tibet provinces; hence, they are likely to reduce the average level for economic items of the ‘≥2000 km’ group. Another reason is that the influence of the TGD is possibly limited within the threshold of 2000 km. Consequently, in empirical results, samples are only separated by 500 km, 1000 km and 1500 km, respectively. The main results of Table A1 are consistent with the results of the baseline table.

4.2. Robustness Checks

Table A1 shows that when the treatment sample is divided by the distance of 500 km, the results are more stable and significant than larger distance thresholds, which suggests that the dam site generated greater impacts on counties closer to it. The ‘distance’ term is negative, and the square of the distance is positive in the regressions of within 500 km. These suggested that within 500 km, the relationship between distance of the counties from the dam site and economic development was a U-shape, which decreased first, and then increased when the ‘distance’ grows.

From the results, it is clear to see that the term ‘d*U’ is significantly negative, the upstream term is positive and the downstream is negative. Those terms also suggested that the downstream areas were negatively affected by the dam building, which is also consistent with the expectation that after building the TGD, the downstream lakes would be drained and the water pollution would become worse.

Interestingly, in the downstream regression of within 500 km distance in Table A2, there is one unexpected term: ‘d*Down’. It turns out to be negative, which suggested that the counties located nearer the dam site in the downstream by an extra 1 km would increase 0.08% GDP per capita. In this situation, the possible benefits of the TGD was that it protects the downstream counties that are closer to the dam site from the flood, which were easily destroyed by the floods especially during the rainy season before damming; after damming, they had a higher chance to escape from terrible disasters.

Another important issue of the TGD occurred in 2003, when the dam began to function well. Before that, the dam was just under the process of building and the most important change caused by the project was damming and the rising water level; but after June 2003, the dam began to realize part of its functions, such as generating electricity. In this case, GDP is assumed to be stimulated by providing electricity power to big factories and millions of households in China. Furthermore, there are other effects such as migration and environmental issues that have indirect impacts on the economy and direct impacts on agricultural output.

In order to test for the exact impacts of the dam and compare the results of before and after its operation, another dummy variable ‘Year₂₀₀₃’ is added in the empirical model. It equals to 1 if the year is 2003 or afterward, and it equals to 0 before 2003. In the estimated model, some other intercept terms are created after adding this dummy variable.

Table 9. Robust results—economic impact of building the TGD by setting 2003 as the basis year.

Variable
U	0.121 ***	0.127 ***
	(0.045)	(0.045)
d	−5.06 × 10−6	0.00002
	(0.045)	(0.00007)
Year03	−0.142 ***	−0.096 ***
	(0.023)	(0.025)
d*U	−0.0002 **	−0.0002 ***
	(0.00007)	(0.00007)
Year03*U	0.026	−0.003
	(0.026)	(0.03)
Year03*d	0.0002 ***	0.0001 ***
	(0.00002)	(0.00002)
Year03*d*U	−0.00007 ***	−0.00004 *
	(0.0002)	(0.00002)
River	−0.021	−0.04
	(0.027)	(0.027)
d*River	0.00003 *	0.00004 **
	(0.00002)	(0.00002)
Year03*River	−0.043	−0.042
	(0.031)	(0.031)
Year03*River*d	0.00006 ***	0.00006 ***
	(0.00002)	(0.00002)
Power spread		−0.065 ***
		(0.012)
Direct administration		0.009
		(0.021)
Area		−1.79 × 10−6 *
		(1.07 × 10−6)
Year	0.047 ***	0.046 ***
	(0.003)	(0.003)
Year2	0.004 ***	0.005 ***
	(0.0002)	(0.0002)
Constant	0.075	0.065
	(0.056)	(0.056)
Observations	9378	9290
No. of group	11	11
R-squared	0.62	0.62

Notes: Figures in the parentheses are standard errors; *** denotes significance at the 1% level; ** denotes significance at the 5% level; * denotes significance at the 10% level.

shows the results.

From Table 9, the term ‘Year₀₃’ negatively affects the GDP per capita. This table also shows that the upstream region received economic benefit as the parameter before the term of upstream is positive as before. The intercept term ‘d*U* Year₀₃’ was −0.007% for the upstream region, which showed that counties further away from the dam site after the year 2003 were negative. Furthermore, both of the columns suggested that counties located along the river with further distance had less been hurt in economic growth than the closer ones.

After adding some other control variables, similar results are obtained. Consequently, both the terms ‘*Year₀₃* d’ and ‘Year₀₃*River*d’ turned out to be significantly positive at the 1% level, which exactly reinforced the previous results that the treatment group with longer distances away from the dam were less hurt.

Regarding the results for agricultural productions, by taking 2003 as the shock year, the estimated results in

Table 10. Robust results—agricultural impact of building the TGD by setting 2003 as the basis year.

	(1)	(2)	(3)	(4)
	Lg_cotton	Lg_cotton	Lg_oil	Lg_oil
U	−0.734 **	−0.864 ***	0.225 **	0.183 *
	(0.291)	(0.291)	(0.106)	(0.106)
d	0.000	−0.000	0.000	0.000
	(0.000)	(0.000)	(0.000)	(0.000)
Year03	0.320 **	0.308 **	0.154 ***	0.165 ***
	(0.124)	(0.129)	(0.058)	(0.060)
d*U	0.001 **	0.002 ***	−0.000 **	−0.000
	(0.001)	(0.001)	(0.000)	(0.000)
Year03*U	−0.078	−0.072	−0.088	−0.077
	(0.166)	(0.166)	(0.063)	(0.064)
Year03*d	−0.000 ***	−0.000 ***	−0.000 ***	−0.000 ***
	(0.000)	(0.000)	(0.000)	(0.000)
Year03*d*U	−0.000 **	−0.000 *	0.000 ***	0.000 ***
	(0.000)	(0.000)	(0.000)	(0.000)
River	0.277	0.279	0.000	−0.017
	(0.218)	(0.218)	(0.074)	(0.074)
d*River	−0.000	−0.000	−0.000	0.000
	(0.000)	(0.000)	(0.000)	(0.000)
Year03*River	0.022	0.023	0.231 ***	0.226 ***
	(0.200)	(0.199)	(0.078)	(0.078)
Year03*River*d	0.000 *	0.000 *	−0.000	−0.000
	(0.000)	(0.000)	(0.000)	(0.000)
Year	−27.668 ***	−28.499 ***	−14.602 ***	−14.883 ***
	(10.296)	(10.262)	(3.968)	(3.992)
Year2	0.007 ***	0.007 ***	0.004 ***	0.004 ***
	(0.003)	(0.003)	(0.001)	(0.001)
Agri_power		−0.001 *		−0.001 ***
		(0.001)		(0.000)
Population_agri		0.006 **		0.007 ***
		(0.003)		(0.001)
employ_agri		−0.000 **		−0.000 ***
		(0.000)		(0.000)
employ_nlmy		−0.001		−0.003
		(0.005)		(0.002)
Power_spread		0.036		0.012
		(0.060)		(0.024)
Area		0.002 ***		0.000 ***
		(0.001)		(0.000)
_cons	27,771.940 ***	28,601.225 ***	14,621.076 ***	14,899.882 ***
	(10,327.820)	(10,293.454)	(3980.357)	(4004.271)
Observations	3610.000	3587.000	7702.000	7484.000
R-squared	0.061	0.067	0.052	0.061

Notes: Figures in the parentheses are standard errors; *** denotes significance at the 1% level; ** denotes significance at the 5% level; * denotes significance at the 10% level.

were similar to the baseline results in Table 7. From Table 10, the term ‘Year₀₃’ had significantly positive effects on cotton and oil yields. The intercept terms ‘Year₀₃*d’ in these four columns were significantly negative, indicating that crops planted farther away from the dam were less affected by damming in 2003. However, the terms ‘*Year₀₃* d*U’ in the columns (1) and (2) were significantly negative, and in the columns (3) and (4) were significantly positive. Those results further confirmed that the operation of TGD enhanced functions of water storage, power generation, and flood control. Cotton production in the upstream regions remained significantly negative, indicating the construction of the TGD had a negative impact on cotton cultivation in the upstream regions. However, the impact of the TGD on oil yields was positively significant for the upstream regions. The results suggested that the functioning of the TGD provided rich water and favorable climate for oil cultivation in the upstream regions, and consequently promotes oil yields for the upstream river areas.

5. Concluding Remarks

This paper provided empirical evidence to evaluate the economic and agricultural effects of the Three-Gorges Dam during the building period towards the counties along the main stream of the Yangtze River. The research was conducted by selecting 90 counties along/across the river and 661 counties located in the river areas for 14 years (from the start year of damming in 1997 to its finish in 2010). According to the previous literature, there are a wide variety of advantages and disadvantages both in the upstream and downstream regions of the dam site, and it is not possible to make a simple conclusion on which parts take the dominant position. This work dynamically evaluated and examined the economic and agricultural impacts by applying the ‘Treatment Effect for Treated method’ and comparing the samples in different locations, with different distances and in different periods of time. The results suggest that the short term after damming the TGD did not show any stable improvement in economy, especially as the counties geographically closer to the dam site in the downstream were generally negatively affected by the shock of damming (in the year 1997) and operation (in the year 2003). Regarding its impacts on cotton and oil yields, the results of the two crops turned out to be different. It showed a negative effect on cotton yield in the upstream regions, whereas there was a positive effect on oil yields for the same regions, especially since the operation of the TGD in 2003.

In particular, by recording the data of 14 years, on average, the upstream counties had priority to the downstream ones in GDP per capita and cotton planting. The results suggested that claims of protection from flood disasters were unable to exceed damages in particular regions. For instance, environmental pollution, emigration and drying lakes during the damming period occurred in the downstream regions. The estimated results also show the ‘winners’ and ‘losers’ in economic impact and various crop yields caused by damming, although the results were much more complex than simply an obvious conclusion. Counties with 1 km closer to the dam site in the upstream on average raised GDP per capita by 0.02%. This is in line with the phenomena that the dam site in the upstream areas stimulated the transportation and travel industries. On the contrary, the downstream lakes tended to be dried-up after damming. It also brought different impacts on crop cultivation, as the TGD directly affected water and land resources in the upstream and downstream regions, which affected the growth environment of crops. Based on their different growing needs, the TGD turned out to have different impacts on different crops.

The findings of this paper are important, as they give a detailed systemic evaluation of the TGD project for the first time. The findings illustrate the distribution of the various positive/negative effects caused by the dam in different periods of building. Furthermore, they have significant political implications: first, the analysis guides the government to compensate the counties near the downstream of the dam site by providing them economic compensation. Second, the results suggested that some of the functions such as irrigation and flood defense did not outweigh the losses of the counties in the downstream areas close to the dam site. Additionally, different crops are grown in different environments, and specific policies could be formulated for each type of crop in dealing with the adversely affected yields by damming. Some other practical actions could be taken based on the results of this paper. For instance, the government can instruct famers to apply advanced technologies and provide them subsidies for painting crops in particular regions along the river.

The DID analysis in this paper indicates the direction of follow-up research, such as exploring which channels result in the economic/agricultural outputs. Furthermore, this paper investigates that crop production is differently affected by the dam in different building periods. However, it remains unclear the exact reasons/channels of different impacts on different crops; for example, whether it is caused by water resources, land shrinkage or productivity decline. Due to the unavailability of these specific data, such as temperature, precipitation and sunshine duration, we cannot further analyze the specific reasons for the variation in cotton and oil crop productions, and these investigations are directions of the further research. Regarding the methodology in use, future research can also use machine learning and satellite data for in-depth analysis on this topic.